BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors nonprofit publishers academic institutionsresearch libraries and research funders in the common goal of maximizing access to critical research

The Network Dynamics Hypothesis How Channel Networks Structure RiverineHabitatsAuthor(s) LEE BENDA N LEROY POFF DANIEL MILLER THOMAS DUNNE GORDONREEVES GEORGE PESS and MICHAEL POLLOCKSource BioScience 54(5)413-427 2004Published By American Institute of Biological SciencesDOI httpdxdoiorg1016410006-3568(2004)054[0413TNDHHC]20CO2URL httpwwwbiooneorgdoifull1016410006-3568282004290545B04133ATNDHHC5D20CO3B2

BioOne (wwwbiooneorg) is a nonprofit online aggregation of core research in the biological ecological andenvironmental sciences BioOne provides a sustainable online platform for over 170 journals and books publishedby nonprofit societies associations museums institutions and presses

Your use of this PDF the BioOne Web site and all posted and associated content indicates your acceptance ofBioOnersquos Terms of Use available at wwwbiooneorgpageterms_of_use

Usage of BioOne content is strictly limited to personal educational and non-commercial use Commercialinquiries or rights and permissions requests should be directed to the individual publisher as copyright holder

May 2004 Vol 54 No 5 bull BioScience 413

Articles

Principles of fluvial geomorphology have guidedthe development of much of riverine ecology over the

last half-century A prominent example is the influential rivercontinuum concept (RCCVannote et al 1980) Based on earlyprinciples of fluvial geomorphology (eg Leopold et al1964) the RCC emphasizes spatially and temporally averageddownstream changes in channel morphology over many or-ders of magnitude It predicts gradual adjustments of biotaand ecosystem processes in rivers in accordance with thegeomorphic perspective of gradual downstream changes inhydrologic and geomorphic properties

The linear perspective embodied in the RCC has dominatedriver ecology over the last 20 years (Fisher 1997) althoughdownstream interruptions in channel and valley morphologycaused by alternating canyons and floodplains tributary con-fluences and landslides have long been observed Some haveviewed these interruptions simply as adjustments to the orig-inal RCC (eg Bruns et al 1984 Minshall et al 1985) whereasothers have posited that they serve as the basis for a new view of a river as a ldquodiscontinuumrdquo (eg Perry and Schaeffer 1987 Townsend 1989 Montgomery 1999 Rice etal 2001) In essence river discontinuum perspectives high-light the nonuniform or patchy distribution of habitats andtherefore emphasize habitat heterogeneity expressed at thescale of meters to kilometers Such heterogeneity also arisesbecause of the human perception of scale in which fluviallandforms are hierarchically organized from valley segments

to stream bed particles (Frissell et al 1986) Consequently theidea of patchy and multiscale habitat formation and its relatedheterogeneity has imbued much current thinking in riverineecology (Frissell et al 1986 Naiman et al 1988 Townsend1996 Poff 1997)

Riverine ecology has also recognized the importance of physical disturbance (eg storms fires and floods) in dynamically creating and maintaining certain attributes ofhabitats and thus in influencing ecosystem function (egResh et al 1988 Swanson et al 1988 Townsend 1989 Reeveset al 1995 Poff et al 1997) Just as habitat patches create discontinuities in space disturbances create discontinuities in time Concepts emphasizing disturbance or watershed

Lee Benda (e-mail leebendaaolcom) is a senior scientist with Earth Systems

Institute 310 North Mount Shasta Boulevard Mount Shasta CA 96067

N LeRoy Poff is an associate professor in the Department of Biology Colorado

State University Fort Collins CO 80523 Daniel Miller is a senior scientist with

Earth Systems Institute 3040 NW 57th Street Seattle WA 98107 Thomas

Dunne is a professor in the Donald Bren School of Environmental Science and

Management University of California Santa Barbara CA 93106 Gordon

Reeves is a research fisheries biologist with the US Department of Agriculture

Forest Service Corvallis Forestry Sciences Laboratory 3200 SW Jefferson Way

Corvallis OR 97331 George Pess and Michael Pollock are scientists with the

National Marine Fisheries Service 2725 Montlake Boulevard East Seattle WA

98112 copy 2004 American Institute of Biological Sciences

The Network DynamicsHypothesis How ChannelNetworks Structure Riverine Habitats

LEE BENDA N LEROY POFF DANIEL MILLER THOMAS DUNNE GORDON REEVESGEORGE PESS AND MICHAEL POLLOCK

Hierarchical and branching river networks interact with dynamic watershed disturbances such as fires storms and floods to impose a spatial andtemporal organization on the nonuniform distribution of riverine habitats with consequences for biological diversity and productivity Abruptchanges in water and sediment flux occur at channel confluences in river networks and trigger changes in channel and floodplain morphology Thisobservation when taken in the context of a river network as a population of channels and their confluences allows the development of testable pre-dictions about how basin size basin shape drainage density and network geometry interact to regulate the spatial distribution of physical diversityin channel and riparian attributes throughout a river basin The spatial structure of river networks also regulates how stochastic watershed distur-bances influence the morphology and ages of fluvial features found at confluences

Keywords river ecology landscape ecology fluvial geomorphology river networks disturbance

dynamics are generally applied in the context of a particularlocation within a watershed However recent advances inunderstanding watershed disturbance regimes indicate howdisturbance frequency and magnitude are organized by hier-archical and branching river networks (Benda and Dunne1997a 1997b Gomi et al 2002)

In sum although the RCCrsquos predictions of gradual down-stream change in river attributes and associated biologicalprocesses are valid over orders of magnitude in river size threeother themes have arisen in riverine ecology over the past twodecades in the effort to address how deviations arise from theexpected mean state in physical attributes along a river pro-file These themes are (1) patchiness or heterogeneity (2)stochastic disturbance and (3) hierarchical scaling This suiteof concepts has been used to argue that riverine ecologyshould be guided by principles of landscape ecology a disci-pline that incorporates a similar set of ideas (Schlosser 1991)Such a landscape view of rivers has led to the idea of ldquoriver-scapesrdquo (Ward et al 2002) and hence to an emphasis on theimportance of studying riverine habitats and their patchinessover multikilometer scales (eg Fausch et al 2002 Wiens2002)

In a similar vein a conceptual framework integrating het-erogeneity disturbance and hierarchical scaling has beenadvocated for general ecology under the title of ldquohierarchicalpatch dynamicsrdquo(Wu and Loucks 1995) and this concept mayapply well to riverine ecology (Townsend 1989 Poole 2002)While both hierarchical patch dynamics and the landscapeview of rivers hold promise for advancing the field of river-ine ecology they are presently limited in doing so because theylack a physical basis for understanding or predicting the mor-phological implications of rivers as networks (as opposed toldquolinear featuresrdquo sensu Fisher [1997]) and for understandinghow stochastic watershed disturbances are translated into patterns of physical heterogeneity throughout the rivernetwork

The purpose of this article is to develop a geomorphicframework in support of recent advances in river ecology Tocreate this framework which we call the network dynamicshypothesis we developed testable predictions about how thespatial arrangement of tributaries in a river network interactswith stochastic watershed processes to influence spatiotem-poral patterns of habitat heterogeneity We begin with a gen-eral review of how tributary confluences modify channelmorphology Then we describe how tributary confluence ef-fects vary in terms of the specific attributes of a networkrsquosstructure including basin size basin shape network patternsize difference between confluent channels drainage den-sity confluence density local network geometry and thepower law of stream sizes (figure 1) Next we describe howstochastic watershed disturbances such as floods fire andstorms impose temporal heterogeneity on confluence effectsbut in a predictable fashion that reflects the controls exertedby the underlying network structure Finally we considerhow the general principles developed in our hypothesis could

414 BioScience bull May 2004 Vol 54 No 5

Articles

a

b

c

d

Structure of river networks

(1) Basin size and (2) basinshape

(5) Drainage density and (6) confluence density

(3) Network configurationand (4) size difference between tributary channelsand main stem

(7) Network geometryconfluence angles and separation distance betweengeomorphically significanttributaries

Figure 1 Seven structural indexes of river networks areconsidered in this article (a) Basin size is drainage areaand basin shape is determined by dividing the area of theriver basin by the square of the length of the main stem(b) Confluence effects depend on the overall network pat-tern and on the size difference between confluent chan-nels (ie the size ratio of the tributary basin to the mainstem basin upstream of the confluence) (c) Drainage density and confluence density are related measuresdrainage density is the total length of channels per unitarea of the basin while confluence density is the numberof confluences per unit channel length or basin areaChannel sizes in a basin are distributed according to apower law (d) Local network geometry (in the vicinity ofindividual confluences) depends on the tributaryndashmainstem area ratio on the distance between confluences (Z)and on the confluence intersection angle (θ) Black dotsin panels b c and d represent tributary confluences

vance the coupled disciplines of geo-morphology and riverine biology

Effects of tributary confluences onchannel and valley morphology By definition a tributary is the smaller of twointersecting channels and the larger is themain stem Strictly speaking a tributaryjunction or confluence is defined as thepoint where two different streams meet Inthe broader definition used in this article atributary junction is the valley floor envi-ronment influenced by tributaries and mayinclude alluvial fans terraces secondarychannels and wider floodplains The nu-merous bifurcations and confluences of dis-tributaries in braided channel systems arenot covered here

Three main types of processes are re-sponsible for transporting sediment and organic material down tributaries to con-fluences with the main stem Debris flowstransport an unsorted mixture of sediment(including boulders and logs) and often cre-ate erosion-resistant deposits normal runofffloods transport bed load and suspendedload and create stratified alluvial depositsand flash floods transport extremely highsediment loads and create deposits inter-mediate between debris flows and runofffloods These sediment transport processesoften create depositional fans where tribu-tary channels enter lower-gradient and widerchannels or valleys (Bull 1977)

The interaction of two independent sed-iment transport regimes at channel junc-tions can produce dramatic changes on thereceiving channel and valley floor (see table1 for a listing of these effects) Morpholog-ical effects at confluences including forma-tion of fans may be transient or persistentdepending on the rate at which organic ma-terial and sediment are transported to tributary junctionsand moved by receiving channels Sediment deposits thatform at junctions can impose a topographic impedimentto the main channel often locally constricting valley widthand displacing the main channel across the valley floor (figure 2) These topographic effects induce certain morphologic responses in main stem channels such as alocalized flattening of the channel gradient upstream anda corresponding steepening of the gradient downstream(figure 2) Gradient-induced longitudinal variations insediment transport rate in the vicinity of junctions causeupstream reductions in substrate size increases in chan-nel meandering and increases in floodplain and terracewidth These changes are offset by other tendencies on the

May 2004 Vol 54 No 5 bull BioScience 415

Articles

Tributary confluence effects

Longitudinal profile

Distance

Elev

atio

n

Zone of interference Zone of mixing

Morphological consequences

bull Lower gradient

bull Wider channelfloodplain

bull Increased bank erosion

bull Increased wood recruitment

bull Finer substrates

bull Greater lateral connectivity

bull Higher disturbance magnitude

bull Higher gradient

bull Larger substrate sizes

bull Deeper pools

bull More bars

bull Higher frequency and magnitudeof disturbance

North ForkBoise River

Upstream(zone of interference)

Tributary influxof sediment to

alluvial fan

Downstream(zone of mixing)

Alluvialfan

Figure 2 An alluvial fan enlarged after a fire triggers tributaryjunction effects in the North Fork Boise River (320-square-kilometer drainage area) Junction effects include expanded floodplain and terrace formation increased channel meanderingand side channels upstream of the fan and channel steepeningdownstream (Benda et al 2003b) Confluence effects include up-stream interference in which a lower gradient and wider channelcause a reduction in the transport of sediment and wood and acorresponding increase in channel changes ldquoMixingrdquo effectsdownstream of the confluence including a steeper channel gradi-ent and a higher frequency and magnitude of disturbance arisefrom the abrupt introduction of sediment and wood from the tributary Photograph Steven Toth

downstream side of the junction including coarser substratesand increases in channel width pool depth and occurrenceof bars The same general classes of channel changes occur atconfluences regardless of their location in the river network(see table 1) although certain types of changes such as boul-der accumulations leading to rapids occur predominantly neardebris flow or flash flood deposits The morphological con-ditions near channel junctions differ from those in reaches lo-cated upstream or downstream confluences are agents ofhabitat formation and increased morphological heterogene-ity (figure 2 Rice et al 2001 Benda et al 2003a) In this ar-ticle we concentrate on the morphological effects at junctionslinked to tributary sources of sediment and wood althoughour analysis of the influences of river network geometryshould also apply to more flow-related changes in morphol-

ogy at junctions in less erosion-prone landscapes (eg Best1986 Rhoads 1987)

Effects of river networks on the structure of riverine habitatsThe physical structure of river networks can be defined bybasin size basin shape network pattern size difference betweenconfluent channels the power law of stream sizes (egHorton 1945) drainage density and confluence density andlocal network geometry (figure 1) Our predictions abouthow river network structure influences spatial patterns ofconfluence-related morphology (box 1) apply to a range ofchannel changes (eg changes in substrate size channelwidth or extent of floodplains and terraces) althoughmorphological effects can be broadly stratified according to

416 BioScience bull May 2004 Vol 54 No 5

Articles

Table 1 Fourteen studies documenting tributary confluence effects in 19 nonregulated streams and rivers across thewestern United States and Canada

Climatic Type of sediment Contributing Receiving MorphologicalLocation region transport stream area (km2) stream area (km2) effects Source

Sekiu Olympic Mountains Humid Debris flow 002 to 073 067 to 42 a b d g k Benda et al Washington 2003a

Ash Creek Arizona Arid Flash flood 042 98 d f Wohl and Pear-three 1991

Queen Charlotte Islands Humid Debris flow 011 to 56 03 to 120 k Hogan et al 1998British Columbia

Matheny and Sitkum Olympic Humid Debris flow 037 203 a b d g j k Benda et al Mountains Washington 2003a

Coast Range Oregon Humid Debris flow 008 to 027 08 to 30 a b c d e f Everest and g h j k l Meehan 1981

Sheep Creek Idaho Semiarid Alluvial 266 646 a b c d g Benda et al 2003b

Oregon Cascades Humid Debris flow 011 to 30 51 to 71 e f g Grant and Swanson 1995

Crooked River Idaho Semiarid Alluvial 34 219 a b c d g Benda et al 2003b

Bear Creek Colorado Semiarid Flash flood 59 to 239 193 to 407 e Grimm et al 1995

North Fork Boise River Idaho Semiarid Alluvial 06 to 29 322 to 461 a b c d f g h Benda et al 2003b

Wenaha River Oregon Semiarid Alluvial 18 to 71 446 to 516 f g Baxter 2002

Snoqualmie River Washington Humid Alluvial 85 to 750 712 to 1794 a b c d Booth et al 1991

Pine and Sukunka Rivers Semiarid Alluvial 23 to 203 1579 to 2145 d Rice et al 2001British Columbia

South Fork Payette River Idaho Semiarid Flash flood 055 2470 e Meyer and Pierce 2003

Bella Coola River Humid Alluvial 128 to 285 4779 to 5421 m Church 1983British Columbia

Middle Fork Salmon River Idaho Semiarid Debris flow 25 to 295 1176 to 7096 a b e Meyer and flash flood Leidecker 1999

Grand Ronde Oregon Semiarid Alluvial 764 to 1342 6953 to 7781 f g Baxter 2002

Snake River Oregon Semiarid Alluvial 9137 240765 g f Meyer and Pierce 2003

Colorado River Arid Debris flow 143 to 6076 280000 to 386800 a d e i Melis et al 1995(before dam) Colorado flash flood

a gradient steepening b gradient lowering c upstream sediment deposition d changing substrate size e boulders and rapids f terraces g flood-plains h side channels i midchannel bars j ponds k logjams l meanders m channel instability

Note Sites are arrayed according to increasing drainage area of the main stem (in square kilometers) The studies indicate the dominant type ofsediment transport the drainage areas of tributaries and main stems and the type of morphological effects at confluences indicated by the authors

upstream and downstream position relative to confluences(eg figure 2) We also predict local changes in heterogene-ity at confluences which will usually increase Heterogeneityis defined by the type form and age distribution of fluviallandforms It is not yet possible to develop quantitative pre-dictions about specific morphological changes at conflu-ences because of the low resolution of data (eg table 1)and the complex nature of riverine environments (Rhoads1987)

The role of basin size Consistent flow-related morphologicalchanges (ie in channel width and depth) occur at junc-tions where the ratio between tributary size and main stem

size approaches 06 or 07 (Rhoads 1987) We postulate thatmorphological effects caused by punctuated inputs of sedi-ment and wood at confluences will also scale to the size of thetributary relative to the main stem We anticipate this resultbecause larger basins typically produce larger quantities of sed-iment and because larger tributaries generally have larger fansassociated with them (Bull 1977) Moreover larger and morepowerful rivers are more effective at removing tributary inputs of sediment (Benda et al 2003a)

To evaluate this expectation we analyzed the results of 14published field studies that document a range of confluenceeffects in 19 streams and rivers caused by the abrupt intro-duction of sediment and wood (see table 1 for a listing of these

May 2004 Vol 54 No 5 bull BioScience 417

Articles

Testable predictions from the network dynamics hypothesis are listed below The predictions apply to a range ofmorphological changes at confluences (including morphological heterogeneity) although they can be broadly stratifiedaccording to upstream and downstream position relative to confluences (see eg figure 2) Presently it is not feasible to make quantitative predictions of channel changes at confluences because of the complexity of channel environments(Rhoads 1987) Confluences that have identifiable morphological effects (including those confluences listed in table 1)are referred to as ldquogeomorphically significant confluencesrdquo

Predictions related to network structurebull The likelihood of confluence effects increases with the ratio of tributary to main stem sizes

bull Heart-shaped and pear-shaped basins containing dendritic networks favor increasing tributary size and hence confluenceeffects downstream compared with rectangular basins containing trellis or parallel networks which do not

bull The separation distance between geomorphically significant junctions (and their morphological effects) increases downstream withincreasing basin size particularly in dendritic networks

bull As basin size increases the channel length and area affected by individual confluence-related channel and valley morphologicalmodifications will increase

bull Closely spaced tributaries will yield valley segments of higher physical heterogeneity compared with valley segments that do not contain closely spaced geomorphically significant confluences

bull Basins with higher drainage density and corresponding higher junction density will have a higher degree of morphological heterogeneity Correspondingly basins of high topographic roughness in part related to higher drainage density should contain a higher degree of riverine heterogeneity

Predictions related to watershed disturbances or dynamicsbull Basins with higher punctuated sediment supply and transport will be characterized by greater confluence effects

bull Channelized disturbances (ie floods and accelerated sediment and wood supply) will have greater frequency and magnitude prox-imal to and immediately downstream of confluences leading to greater physical heterogeneity including the age distribution of flu-vial landforms

bull Channelized disturbances will be magnified immediately upstream of geomorphically significant confluences leading to greaterphysical heterogeneity including the age distribution of fluvial landforms

bull In sufficiently large basins (approximately 102 km2) the age distribution of confluence-related landforms (ie fans terracesfloodplains) will be skewed toward older geomorphic features in headwaters and toward younger features in downstream portions of the basin

bull Physical heterogeneity (ie diversity of forms and ages of channels floodplains terraces and logjams) will be concentrated in certain parts of networks

Box 1 The influence of river network structure on spatial patterns of confluence-related morphology

effects)We refer to confluences that have observable morpho-logical effects as ldquogeomorphically significantrdquoTogether thesestudies describe 168 junctions along 560 kilometers (km) ofriver spanning 7 orders of magnitude in drainage area in thewestern United States and Canada (see Benda et al [2004] fora more detailed analysis) These data reveal that as the size ofthe main stem increases geomorphically significant conflu-ences are associated with increasingly larger tributaries (fig-ure 3) For example debris flows that originate from smallbasins (up to 1 km2 in drainage area) create tributary-junction effects in basins of only 1 to 50 km2 By contrasttributary-junction effects in larger rivers (1000 to 300000 km2)are associated with larger tributaries (10 to 10000 km2)

The data in figure 3 also reveal a threshold below whichtributary basins less than approximately 1 km2 do not af-fect main stem rivers larger than approximately 50 km2 Thisis significant because it indicates how the ldquolaw of streamnumbersrdquo (Horton 1945) the universal power law regard-ing the distribution of channel sizes (and by analogy of theconfluences linked to those channels) constrains confluenceeffects at the scale of entire networks In humid landscapesfor example the vast majority (80 to 90) of tributarybasins are first or second order with less than 1 km2 indrainage area (Benda and Dunne 1997a) On the basis ofthe threshold identified in figure 3 only 10 to 20 of trib-utary confluences (ie those of third order or higher) areavailable to significantly affect main stem rivers of a sizegreater than approximately 50 km2 This finding has im-portant ramifications for the increased downstream spac-ing of confluence effects as the size of the main stemincreases

The variability of thedata in figure 3 is probablydue to factors such as basingeology (eg the durabil-ity and size distribution ofsediment) sediment trans-port processes (eg debrisflows and flash floods thattransport boulders and aremore likely to cause con-fluence effects) and localvalley width (eg wide val-leys that limit fan forma-tion and the associatedconfluence effects) Tem-poral variation in the his-tory of storms fires andfloods that create or reju-venate confluence effectsshould also cause the wax-ing and waning of conflu-ence effectsA more detailedanalysis of the data in figure3 allows for probabilisticpredictions of confluence

effects (Benda et al 2004) For example in humid environ-ments a range of 06 to 09 for the probability of a confluenceeffect corresponds to a range of 004 to 08 for tributaryndashmainstem drainage area ratios

The role of network pattern and basin shape The scaling re-lationship between tributaries and main stem channels (fig-ure 3) allows us to consider how the factors that control thespatial distribution of tributary sizes in river networks influ-ence spatial patterns of confluence-related morphology andheterogeneity Downstream trends in junction effects are in-fluenced by network patterns and hence by drainage basinshape Two common types of network patterns are dendriticand trellis networks Dendritic networks which resemblethe hierarchical branching pattern of a tree often form inhomogeneous and gently sloped geologic beds and createheart-shaped or pear-shaped basins In contrast trellis net-works characterized by small tributaries intersecting mainstem channels are often associated with elongate landformsin dipped and folded sedimentary rocks or in areas of paral-lel fractures they create narrow rectangular basins

The spatial configuration of tributaries within a watershedchanges the likelihood of confluence effects downstream inriver networks Since larger tributaries are required to creategeomorphic effects as the size of the main stem increases(eg figure 3) dendritic networks in heart-shaped or pear-shaped basins should promote confluence effects throughoutthe watershed (figure 4 see Benda et al [2004] for more details) These effects occur because increasing the basinwidth downstream promotes the coalescing of hierarchicallybranched channels resulting in larger tributaries forming

418 BioScience bull May 2004 Vol 54 No 5

Articles

Drainage area of main stem river (km2)

Humid debris flowSemiarid debris flowflash floodHumid alluvialSemiarid alluvial

Transition from 1st and 2ndorder to higher-order channels

Dra

ina

ge a

rea

of trib

utar

y (k

m2)

Figure 3 A comparison between tributary basin area and main stem drainage area (in squarekilometers) for 168 tributaries where confluence effects (ie geomorphically significant conflu-ences) have been documented reveals that larger main stem rivers require larger tributaries inorder to experience confluence effects (table 1 Benda et al 2004) The dashed line represents theapproximate tributary drainage threshold for the first- and second-order streams that comprisethe vast majority of stream sizes and for their confluences in networks Power law distributionof channel (confluence) sizes is shown on the left side of the figure P refers to the proportion ofall channels of a given drainage area

downstream In contrast narrow rectangular basins con-taining trellis networks lack the formation of larger tributariesand therefore discourage confluence effects as main stemsize increases Consequently network configuration relatedto basin shape should strongly influence the downstreamsequence of confluence effects (figure 4 see also the testablepredictions in box 1)

In addition to dendritic and trellis networks other networktypes include parallel networks which form in conjunctionwith parallel landforms and rectangular networks whichform where numerous faults and joints converge at high an-gles Region-specific types of geology and hillslope topogra-phy should influence network patterns and hence the spatialdistribution of confluence-related channel morphology andassociated physical heterogeneity For example the youngand porous rocks of the high Cascade Mountains in Oregonare characterized by trellis networks while older and lesspermeable rocks nearby exhibit dendritic networks (Grant1997)

Drainage density and confluence density The cumulativeeffect of confluences within a basin should be proportionalto the total number of geomorphically significant channel con-fluences This number is related to drainage density (definedas channel length per unit watershed area) and to networkshape which either promotes or inhibits the occurrence ofconfluent channels (see figure 5) The corresponding con-fluence density (the number of geomorphically significantconfluences per unit area or per unit channel length) shouldprovide a simple measure of the net morphological effect ofconfluences in rivers (box 1) Drainage densities in humid tosemiarid landscapes range from 2 to 12 km channel length persquare kilometer watershed area primarily reflecting varia-tions in precipitation landscape age and bedrock porosity(Grant 1997) This large range in drainage density translatesto a correspondingly large range in the density of channel con-fluences with implications for the degree of channel hetero-geneity found in different landscapes (figure 5)

Local network geometry A riverrsquos geologic structure and tec-tonic and erosional history create specific patterns of inter-secting tributaries over kilometer scales Local networkgeometry can be used to describe the kilometer-scale varia-tion of tributary effects in rivers including the longitudinalsequence of tributaryndashmain stem size ratios (figures 3 4) trib-utary intersection angles and distance between geomorphi-cally significant confluences (figure 1d) The tributaryndashmainstem intersection angle is the upstream angle formed at a con-fluence Intersection angles are almost always acute and as theybecome less so and approach 90deg the likelihood of a geo-morphic effect at a confluence increases For instance in a se-ries of river studies and in flume experiments Mosley (1976)and Best (1986) showed how bar size bar location and scourdepth vary with confluence angle Lateral bars are more likelyto form at acute angles scour depth by contrast increases withincreasing confluence angle and approaches an asymptote at

angles approaching 90deg Moreover in headwater areas con-fluence angles greater than 70deg encourage debris flow depo-sition and consequent junction effects (table 1) whiledeposition is discouraged at junctions with more acute an-gles (Benda and Cundy 1990)

Network geometry also describes the distance separatinggeomorphically significant confluences For example a concentration of large tributaries in a central sediment-producing region is characteristic of watersheds along moun-tain fronts abutting depositional plains (figure 6a) Alterna-tively tributaries that are separated by canyons can lead toclumped distributions of geomorphically significant tribu-taries and associated physical heterogeneity (figure 6b)Large tributary junctions that are closely spaced may have con-

May 2004 Vol 54 No 5 bull BioScience 419

Articles

Distance along main stem channel

Rectangular

a

bHigh

Like

lihoo

d of

con

fluen

ce e

ffect

s

Low

Pear shaped

Heart shaped

Heart shaped(most compact)

Pear shaped(moderately compact)

Rectangular (most elongate)

Figure 4 (a) Network patterns and the associated basinshapes affect the downstream sequence of confluence effects Common network patterns include dendritic networks within heart-shaped and pear-shaped basinsand trellis networks within narrow rectangular basins(b) The anticipated downstream sequence of confluenceeffects (ranging from high to low likelihood) is based onthe size of the tributaries relative to the main stemsDendritic networks in heart-shaped basins promote thegreatest likelihood of confluence effects downstreamwhile trellis networks in rectangular basins promote thefewest effects downstream Modified from Benda and colleagues (2004)

Downstream sequence of confluence effects

Downstream sequence of confluence effects

fluence effects that overlap particularly during watersheddisturbances such as floods and concentrated basin erosionConfluence effects may also be less pronounced in wide valley floors where fans are isolated from the main stem bybroad terraces or floodplains

Scaling properties of confluence effects The finding that largertributaries are required in order to affect the morphology oflarger rivers reveals two scaling effects on habitat patches

that develop near confluences First geomorphically signifi-cant confluences should be separated by increasing distancesdownstream in watersheds where tributary basins down-stream increase in size (eg dendritic networks in pear-shaped and heart-shaped basins figure 4 box 1) Thisincreasing separation occurs because tributary length is relatedto tributary drainage area (Hack 1957) and thus the lengthand width of a basin increase with increasing tributary lengthThis spacing pattern is reflected in the field data (figure 7) Inthe upper portions of humid drainage basins morphologi-cal effects are spaced on average hundreds of meters apart(Benda 1990 Hogan et al 1998 Benda et al 2003a) reflect-ing the spacing of low-order tributaries By contrast in largerbasins (up to 300000 km2) the distances separating junctioneffects are on the order of several kilometers to tens ofkilometers (Baxter 2002 Benda et al 2003b)

420 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 5 Two basins in northern California illustratevariation in confluence densities Black dots representconfluences with tributaryndashmain stem ratios indicatedby dot size A watershed formed in relatively young andhomogeneous lava flows (top) has a drainage density of 3kilometers (km) channel length per square kilometerdrainage area and a corresponding confluence density of47 junctions per square kilometer A watershed locatedin the older and geologically heterogeneous terrain (bot-tom) has a drainage density of 5 km channel length persquare kilometer and a corresponding confluence densityof 12 junctions per square kilometer On the basis of thedifference in confluence density the second watershed ispredicted to have higher confluence-related habitatheterogeneity (box 1) The probability of confluence effects depends on the increasing size ratio of tributary to main stem drainage area as indicated by the size ofthe dots (Benda et al 2004) Estimation of drainage density or confluence density should be sensitive to mapscale and to the method of depicting channel networks

Figure 6 Local network geometry (kilometer scale) variesin response to the size spacing and confluence angles ofintersecting tributaries which reflect underlying geologicstructure topography and erosion history Different localgeometries are expected to lead to different patterns ofphysical heterogeneity linked to tributary confluences(a) With a number of large tributaries near the center of the network confluence-related heterogeneity is con-centrated in a central sediment-producing region Thispattern is characteristic of watersheds located alongmountain fronts and abutting depositional plains(b) When major tributaries are separated by canyonsthe associated areas of physical heterogeneity are dividedas well

Physical heterogeneity

a b

Higher Lower

Canyon

Tributaryndashmain stemdrainage area ratio

Tributaryndashmain stemdrainage area ratio

302 km2

299 km2

The second scaling property of confluence-related mor-phology is that the size of habitat patches associated with trib-utaries increases downstream (figure 7)Although the data aresparse the length of channels affected by confluences rangesfrom tens or a few hundreds of meters (in basins less than ap-proximately 100 km2) to several kilometers (in basins betweenapproximately 1000 and 300000 km2) We expect the size ofhabitat patches to increase farther downstream because thechannel gradient declines with increasing river size thereforeany vertical obstruction in a channel that backs up sediment(eg fans logjams boulders) should influence a channeldistance upstream at least equivalent to the obstructionheight divided by the tangent of the underlying stream gra-dient

Interaction of river networks with stochastic watershed processes A river network can be thought of as the landscape templatewithin which climatically induced stochastic fluctuations inthe supply and transport of water sediment and organicmaterial occur The structure of a river network (basin size

shape network configuration and so on fig-ure 1) can be used to help understand therole of stochastic watershed processes inshaping riverine environments

Disturbance The role of tributary conflu-ences on longitudinal patterns of riverineheterogeneity and the dependence of thisheterogeneity on basin scale basin shapenetwork pattern drainage density and net-work geometry can be framed as a generalset of testable predictions (see box 1) Thespatial configuration of a network thus pro-vides a template for organizing the transportand deposition of sediment wood and wa-ter through a watershed but the supply andtransport of material that creates confluenceeffects varies dynamically over time becauseof climate-driven events (storms floodsand fires) The watershed disturbanceregime (Miller et al 2003)mdashthe frequencymagnitude and spatial extent of episodic cli-matic and geomorphic processesmdashcan varywithin and among watersheds and thuscontribute to variation in heterogeneity atconfluences including the waxing and wan-ing of effects over time

Moderate to large disturbance-drivenlandscape-scale fluctuations in the supplyand storage of in-channel sediment and or-ganic material create many of the morpho-logical changes observed at confluencesFor example the fluctuating supply of sed-iment storage at confluences creates fansfloodplains terraces logjams secondary

channels and fans (table 1 Small 1973 Benda et al 2003b)The episodic nature of sediment-related disturbances createsthe form and age mosaic of the erosional and depositionallandforms characteristic of valley floors ultimately con-tributing to physical heterogeneity and potentially to biologicaldiversity and increased productivity (Swanson et al 1988)Al-though biological recovery processes such as vegetation suc-cession can also influence patterns of physical heterogeneityin the channel and valley floor (Gregory et al 1991) that topicis beyond the scope of this article For more comprehensivediscussions of the role of disturbances in riverine environ-ments see Resh and colleagues (1988) Reeves and colleagues(1995) Poff and colleagues (1997) and Nakamura and colleagues (2000)

Amplification of disturbances at confluences The channel dis-turbance regime can be altered at tributary confluences Firsttributaries represent abrupt increases in the supply of watersediment and wood and therefore channel responses re-lated to those inputs should have a higher frequency andmagnitude near or immediately downstream of confluences

May 2004 Vol 54 No 5 bull BioScience 421

Articles

Figure 7 Confluence effects in rivers (eg changes in substrate floodplains orchannel morphology) reveal two scaling properties First the distance separatinggeomorphically significant confluences (shown in meters) increases downstreamparticularly in dendritic networks (top) Second the length of altered habitatpatches associated with geomorphically significant confluences (also in meters)increases downstream (bottom)

Leng

th o

f inf

luen

ce (m

)S

paci

ng b

etw

een

geom

orph

ical

lysi

gnifi

cant

con

fluen

ces

(m)

Main stem drainage area (km2)

Main stem drainage area (km2)

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

vance the coupled disciplines of geo-morphology and riverine biology

Effects of tributary confluences onchannel and valley morphology By definition a tributary is the smaller of twointersecting channels and the larger is themain stem Strictly speaking a tributaryjunction or confluence is defined as thepoint where two different streams meet Inthe broader definition used in this article atributary junction is the valley floor envi-ronment influenced by tributaries and mayinclude alluvial fans terraces secondarychannels and wider floodplains The nu-merous bifurcations and confluences of dis-tributaries in braided channel systems arenot covered here

Three main types of processes are re-sponsible for transporting sediment and organic material down tributaries to con-fluences with the main stem Debris flowstransport an unsorted mixture of sediment(including boulders and logs) and often cre-ate erosion-resistant deposits normal runofffloods transport bed load and suspendedload and create stratified alluvial depositsand flash floods transport extremely highsediment loads and create deposits inter-mediate between debris flows and runofffloods These sediment transport processesoften create depositional fans where tribu-tary channels enter lower-gradient and widerchannels or valleys (Bull 1977)

The interaction of two independent sed-iment transport regimes at channel junc-tions can produce dramatic changes on thereceiving channel and valley floor (see table1 for a listing of these effects) Morpholog-ical effects at confluences including forma-tion of fans may be transient or persistentdepending on the rate at which organic ma-terial and sediment are transported to tributary junctionsand moved by receiving channels Sediment deposits thatform at junctions can impose a topographic impedimentto the main channel often locally constricting valley widthand displacing the main channel across the valley floor (figure 2) These topographic effects induce certain morphologic responses in main stem channels such as alocalized flattening of the channel gradient upstream anda corresponding steepening of the gradient downstream(figure 2) Gradient-induced longitudinal variations insediment transport rate in the vicinity of junctions causeupstream reductions in substrate size increases in chan-nel meandering and increases in floodplain and terracewidth These changes are offset by other tendencies on the

May 2004 Vol 54 No 5 bull BioScience 415

Articles

Tributary confluence effects

Longitudinal profile

Distance

Elev

atio

n

Zone of interference Zone of mixing

Morphological consequences

bull Lower gradient

bull Wider channelfloodplain

bull Increased bank erosion

bull Increased wood recruitment

bull Finer substrates

bull Greater lateral connectivity

bull Higher disturbance magnitude

bull Higher gradient

bull Larger substrate sizes

bull Deeper pools

bull More bars

bull Higher frequency and magnitudeof disturbance

North ForkBoise River

Upstream(zone of interference)

Tributary influxof sediment to

alluvial fan

Downstream(zone of mixing)

Alluvialfan

Figure 2 An alluvial fan enlarged after a fire triggers tributaryjunction effects in the North Fork Boise River (320-square-kilometer drainage area) Junction effects include expanded floodplain and terrace formation increased channel meanderingand side channels upstream of the fan and channel steepeningdownstream (Benda et al 2003b) Confluence effects include up-stream interference in which a lower gradient and wider channelcause a reduction in the transport of sediment and wood and acorresponding increase in channel changes ldquoMixingrdquo effectsdownstream of the confluence including a steeper channel gradi-ent and a higher frequency and magnitude of disturbance arisefrom the abrupt introduction of sediment and wood from the tributary Photograph Steven Toth

downstream side of the junction including coarser substratesand increases in channel width pool depth and occurrenceof bars The same general classes of channel changes occur atconfluences regardless of their location in the river network(see table 1) although certain types of changes such as boul-der accumulations leading to rapids occur predominantly neardebris flow or flash flood deposits The morphological con-ditions near channel junctions differ from those in reaches lo-cated upstream or downstream confluences are agents ofhabitat formation and increased morphological heterogene-ity (figure 2 Rice et al 2001 Benda et al 2003a) In this ar-ticle we concentrate on the morphological effects at junctionslinked to tributary sources of sediment and wood althoughour analysis of the influences of river network geometryshould also apply to more flow-related changes in morphol-

ogy at junctions in less erosion-prone landscapes (eg Best1986 Rhoads 1987)

Effects of river networks on the structure of riverine habitatsThe physical structure of river networks can be defined bybasin size basin shape network pattern size difference betweenconfluent channels the power law of stream sizes (egHorton 1945) drainage density and confluence density andlocal network geometry (figure 1) Our predictions abouthow river network structure influences spatial patterns ofconfluence-related morphology (box 1) apply to a range ofchannel changes (eg changes in substrate size channelwidth or extent of floodplains and terraces) althoughmorphological effects can be broadly stratified according to

416 BioScience bull May 2004 Vol 54 No 5

Articles

Table 1 Fourteen studies documenting tributary confluence effects in 19 nonregulated streams and rivers across thewestern United States and Canada

Climatic Type of sediment Contributing Receiving MorphologicalLocation region transport stream area (km2) stream area (km2) effects Source

Sekiu Olympic Mountains Humid Debris flow 002 to 073 067 to 42 a b d g k Benda et al Washington 2003a

Ash Creek Arizona Arid Flash flood 042 98 d f Wohl and Pear-three 1991

Queen Charlotte Islands Humid Debris flow 011 to 56 03 to 120 k Hogan et al 1998British Columbia

Matheny and Sitkum Olympic Humid Debris flow 037 203 a b d g j k Benda et al Mountains Washington 2003a

Coast Range Oregon Humid Debris flow 008 to 027 08 to 30 a b c d e f Everest and g h j k l Meehan 1981

Sheep Creek Idaho Semiarid Alluvial 266 646 a b c d g Benda et al 2003b

Oregon Cascades Humid Debris flow 011 to 30 51 to 71 e f g Grant and Swanson 1995

Crooked River Idaho Semiarid Alluvial 34 219 a b c d g Benda et al 2003b

Bear Creek Colorado Semiarid Flash flood 59 to 239 193 to 407 e Grimm et al 1995

North Fork Boise River Idaho Semiarid Alluvial 06 to 29 322 to 461 a b c d f g h Benda et al 2003b

Wenaha River Oregon Semiarid Alluvial 18 to 71 446 to 516 f g Baxter 2002

Snoqualmie River Washington Humid Alluvial 85 to 750 712 to 1794 a b c d Booth et al 1991

Pine and Sukunka Rivers Semiarid Alluvial 23 to 203 1579 to 2145 d Rice et al 2001British Columbia

South Fork Payette River Idaho Semiarid Flash flood 055 2470 e Meyer and Pierce 2003

Bella Coola River Humid Alluvial 128 to 285 4779 to 5421 m Church 1983British Columbia

Middle Fork Salmon River Idaho Semiarid Debris flow 25 to 295 1176 to 7096 a b e Meyer and flash flood Leidecker 1999

Grand Ronde Oregon Semiarid Alluvial 764 to 1342 6953 to 7781 f g Baxter 2002

Snake River Oregon Semiarid Alluvial 9137 240765 g f Meyer and Pierce 2003

Colorado River Arid Debris flow 143 to 6076 280000 to 386800 a d e i Melis et al 1995(before dam) Colorado flash flood

a gradient steepening b gradient lowering c upstream sediment deposition d changing substrate size e boulders and rapids f terraces g flood-plains h side channels i midchannel bars j ponds k logjams l meanders m channel instability

Note Sites are arrayed according to increasing drainage area of the main stem (in square kilometers) The studies indicate the dominant type ofsediment transport the drainage areas of tributaries and main stems and the type of morphological effects at confluences indicated by the authors

upstream and downstream position relative to confluences(eg figure 2) We also predict local changes in heterogene-ity at confluences which will usually increase Heterogeneityis defined by the type form and age distribution of fluviallandforms It is not yet possible to develop quantitative pre-dictions about specific morphological changes at conflu-ences because of the low resolution of data (eg table 1)and the complex nature of riverine environments (Rhoads1987)

The role of basin size Consistent flow-related morphologicalchanges (ie in channel width and depth) occur at junc-tions where the ratio between tributary size and main stem

size approaches 06 or 07 (Rhoads 1987) We postulate thatmorphological effects caused by punctuated inputs of sedi-ment and wood at confluences will also scale to the size of thetributary relative to the main stem We anticipate this resultbecause larger basins typically produce larger quantities of sed-iment and because larger tributaries generally have larger fansassociated with them (Bull 1977) Moreover larger and morepowerful rivers are more effective at removing tributary inputs of sediment (Benda et al 2003a)

To evaluate this expectation we analyzed the results of 14published field studies that document a range of confluenceeffects in 19 streams and rivers caused by the abrupt intro-duction of sediment and wood (see table 1 for a listing of these

May 2004 Vol 54 No 5 bull BioScience 417

Articles

Testable predictions from the network dynamics hypothesis are listed below The predictions apply to a range ofmorphological changes at confluences (including morphological heterogeneity) although they can be broadly stratifiedaccording to upstream and downstream position relative to confluences (see eg figure 2) Presently it is not feasible to make quantitative predictions of channel changes at confluences because of the complexity of channel environments(Rhoads 1987) Confluences that have identifiable morphological effects (including those confluences listed in table 1)are referred to as ldquogeomorphically significant confluencesrdquo

Predictions related to network structurebull The likelihood of confluence effects increases with the ratio of tributary to main stem sizes

bull Heart-shaped and pear-shaped basins containing dendritic networks favor increasing tributary size and hence confluenceeffects downstream compared with rectangular basins containing trellis or parallel networks which do not

bull The separation distance between geomorphically significant junctions (and their morphological effects) increases downstream withincreasing basin size particularly in dendritic networks

bull As basin size increases the channel length and area affected by individual confluence-related channel and valley morphologicalmodifications will increase

bull Closely spaced tributaries will yield valley segments of higher physical heterogeneity compared with valley segments that do not contain closely spaced geomorphically significant confluences

bull Basins with higher drainage density and corresponding higher junction density will have a higher degree of morphological heterogeneity Correspondingly basins of high topographic roughness in part related to higher drainage density should contain a higher degree of riverine heterogeneity

Predictions related to watershed disturbances or dynamicsbull Basins with higher punctuated sediment supply and transport will be characterized by greater confluence effects

bull Channelized disturbances (ie floods and accelerated sediment and wood supply) will have greater frequency and magnitude prox-imal to and immediately downstream of confluences leading to greater physical heterogeneity including the age distribution of flu-vial landforms

bull Channelized disturbances will be magnified immediately upstream of geomorphically significant confluences leading to greaterphysical heterogeneity including the age distribution of fluvial landforms

bull In sufficiently large basins (approximately 102 km2) the age distribution of confluence-related landforms (ie fans terracesfloodplains) will be skewed toward older geomorphic features in headwaters and toward younger features in downstream portions of the basin

bull Physical heterogeneity (ie diversity of forms and ages of channels floodplains terraces and logjams) will be concentrated in certain parts of networks

Box 1 The influence of river network structure on spatial patterns of confluence-related morphology

effects)We refer to confluences that have observable morpho-logical effects as ldquogeomorphically significantrdquoTogether thesestudies describe 168 junctions along 560 kilometers (km) ofriver spanning 7 orders of magnitude in drainage area in thewestern United States and Canada (see Benda et al [2004] fora more detailed analysis) These data reveal that as the size ofthe main stem increases geomorphically significant conflu-ences are associated with increasingly larger tributaries (fig-ure 3) For example debris flows that originate from smallbasins (up to 1 km2 in drainage area) create tributary-junction effects in basins of only 1 to 50 km2 By contrasttributary-junction effects in larger rivers (1000 to 300000 km2)are associated with larger tributaries (10 to 10000 km2)

The data in figure 3 also reveal a threshold below whichtributary basins less than approximately 1 km2 do not af-fect main stem rivers larger than approximately 50 km2 Thisis significant because it indicates how the ldquolaw of streamnumbersrdquo (Horton 1945) the universal power law regard-ing the distribution of channel sizes (and by analogy of theconfluences linked to those channels) constrains confluenceeffects at the scale of entire networks In humid landscapesfor example the vast majority (80 to 90) of tributarybasins are first or second order with less than 1 km2 indrainage area (Benda and Dunne 1997a) On the basis ofthe threshold identified in figure 3 only 10 to 20 of trib-utary confluences (ie those of third order or higher) areavailable to significantly affect main stem rivers of a sizegreater than approximately 50 km2 This finding has im-portant ramifications for the increased downstream spac-ing of confluence effects as the size of the main stemincreases

The variability of thedata in figure 3 is probablydue to factors such as basingeology (eg the durabil-ity and size distribution ofsediment) sediment trans-port processes (eg debrisflows and flash floods thattransport boulders and aremore likely to cause con-fluence effects) and localvalley width (eg wide val-leys that limit fan forma-tion and the associatedconfluence effects) Tem-poral variation in the his-tory of storms fires andfloods that create or reju-venate confluence effectsshould also cause the wax-ing and waning of conflu-ence effectsA more detailedanalysis of the data in figure3 allows for probabilisticpredictions of confluence

effects (Benda et al 2004) For example in humid environ-ments a range of 06 to 09 for the probability of a confluenceeffect corresponds to a range of 004 to 08 for tributaryndashmainstem drainage area ratios

The role of network pattern and basin shape The scaling re-lationship between tributaries and main stem channels (fig-ure 3) allows us to consider how the factors that control thespatial distribution of tributary sizes in river networks influ-ence spatial patterns of confluence-related morphology andheterogeneity Downstream trends in junction effects are in-fluenced by network patterns and hence by drainage basinshape Two common types of network patterns are dendriticand trellis networks Dendritic networks which resemblethe hierarchical branching pattern of a tree often form inhomogeneous and gently sloped geologic beds and createheart-shaped or pear-shaped basins In contrast trellis net-works characterized by small tributaries intersecting mainstem channels are often associated with elongate landformsin dipped and folded sedimentary rocks or in areas of paral-lel fractures they create narrow rectangular basins

The spatial configuration of tributaries within a watershedchanges the likelihood of confluence effects downstream inriver networks Since larger tributaries are required to creategeomorphic effects as the size of the main stem increases(eg figure 3) dendritic networks in heart-shaped or pear-shaped basins should promote confluence effects throughoutthe watershed (figure 4 see Benda et al [2004] for more details) These effects occur because increasing the basinwidth downstream promotes the coalescing of hierarchicallybranched channels resulting in larger tributaries forming

418 BioScience bull May 2004 Vol 54 No 5

Articles

Drainage area of main stem river (km2)

Humid debris flowSemiarid debris flowflash floodHumid alluvialSemiarid alluvial

Transition from 1st and 2ndorder to higher-order channels

Dra

ina

ge a

rea

of trib

utar

y (k

m2)

Figure 3 A comparison between tributary basin area and main stem drainage area (in squarekilometers) for 168 tributaries where confluence effects (ie geomorphically significant conflu-ences) have been documented reveals that larger main stem rivers require larger tributaries inorder to experience confluence effects (table 1 Benda et al 2004) The dashed line represents theapproximate tributary drainage threshold for the first- and second-order streams that comprisethe vast majority of stream sizes and for their confluences in networks Power law distributionof channel (confluence) sizes is shown on the left side of the figure P refers to the proportion ofall channels of a given drainage area

downstream In contrast narrow rectangular basins con-taining trellis networks lack the formation of larger tributariesand therefore discourage confluence effects as main stemsize increases Consequently network configuration relatedto basin shape should strongly influence the downstreamsequence of confluence effects (figure 4 see also the testablepredictions in box 1)

In addition to dendritic and trellis networks other networktypes include parallel networks which form in conjunctionwith parallel landforms and rectangular networks whichform where numerous faults and joints converge at high an-gles Region-specific types of geology and hillslope topogra-phy should influence network patterns and hence the spatialdistribution of confluence-related channel morphology andassociated physical heterogeneity For example the youngand porous rocks of the high Cascade Mountains in Oregonare characterized by trellis networks while older and lesspermeable rocks nearby exhibit dendritic networks (Grant1997)

Drainage density and confluence density The cumulativeeffect of confluences within a basin should be proportionalto the total number of geomorphically significant channel con-fluences This number is related to drainage density (definedas channel length per unit watershed area) and to networkshape which either promotes or inhibits the occurrence ofconfluent channels (see figure 5) The corresponding con-fluence density (the number of geomorphically significantconfluences per unit area or per unit channel length) shouldprovide a simple measure of the net morphological effect ofconfluences in rivers (box 1) Drainage densities in humid tosemiarid landscapes range from 2 to 12 km channel length persquare kilometer watershed area primarily reflecting varia-tions in precipitation landscape age and bedrock porosity(Grant 1997) This large range in drainage density translatesto a correspondingly large range in the density of channel con-fluences with implications for the degree of channel hetero-geneity found in different landscapes (figure 5)

Local network geometry A riverrsquos geologic structure and tec-tonic and erosional history create specific patterns of inter-secting tributaries over kilometer scales Local networkgeometry can be used to describe the kilometer-scale varia-tion of tributary effects in rivers including the longitudinalsequence of tributaryndashmain stem size ratios (figures 3 4) trib-utary intersection angles and distance between geomorphi-cally significant confluences (figure 1d) The tributaryndashmainstem intersection angle is the upstream angle formed at a con-fluence Intersection angles are almost always acute and as theybecome less so and approach 90deg the likelihood of a geo-morphic effect at a confluence increases For instance in a se-ries of river studies and in flume experiments Mosley (1976)and Best (1986) showed how bar size bar location and scourdepth vary with confluence angle Lateral bars are more likelyto form at acute angles scour depth by contrast increases withincreasing confluence angle and approaches an asymptote at

angles approaching 90deg Moreover in headwater areas con-fluence angles greater than 70deg encourage debris flow depo-sition and consequent junction effects (table 1) whiledeposition is discouraged at junctions with more acute an-gles (Benda and Cundy 1990)

Network geometry also describes the distance separatinggeomorphically significant confluences For example a concentration of large tributaries in a central sediment-producing region is characteristic of watersheds along moun-tain fronts abutting depositional plains (figure 6a) Alterna-tively tributaries that are separated by canyons can lead toclumped distributions of geomorphically significant tribu-taries and associated physical heterogeneity (figure 6b)Large tributary junctions that are closely spaced may have con-

May 2004 Vol 54 No 5 bull BioScience 419

Articles

Distance along main stem channel

Rectangular

a

bHigh

Like

lihoo

d of

con

fluen

ce e

ffect

s

Low

Pear shaped

Heart shaped

Heart shaped(most compact)

Pear shaped(moderately compact)

Rectangular (most elongate)

Figure 4 (a) Network patterns and the associated basinshapes affect the downstream sequence of confluence effects Common network patterns include dendritic networks within heart-shaped and pear-shaped basinsand trellis networks within narrow rectangular basins(b) The anticipated downstream sequence of confluenceeffects (ranging from high to low likelihood) is based onthe size of the tributaries relative to the main stemsDendritic networks in heart-shaped basins promote thegreatest likelihood of confluence effects downstreamwhile trellis networks in rectangular basins promote thefewest effects downstream Modified from Benda and colleagues (2004)

Downstream sequence of confluence effects

Downstream sequence of confluence effects

fluence effects that overlap particularly during watersheddisturbances such as floods and concentrated basin erosionConfluence effects may also be less pronounced in wide valley floors where fans are isolated from the main stem bybroad terraces or floodplains

Scaling properties of confluence effects The finding that largertributaries are required in order to affect the morphology oflarger rivers reveals two scaling effects on habitat patches

that develop near confluences First geomorphically signifi-cant confluences should be separated by increasing distancesdownstream in watersheds where tributary basins down-stream increase in size (eg dendritic networks in pear-shaped and heart-shaped basins figure 4 box 1) Thisincreasing separation occurs because tributary length is relatedto tributary drainage area (Hack 1957) and thus the lengthand width of a basin increase with increasing tributary lengthThis spacing pattern is reflected in the field data (figure 7) Inthe upper portions of humid drainage basins morphologi-cal effects are spaced on average hundreds of meters apart(Benda 1990 Hogan et al 1998 Benda et al 2003a) reflect-ing the spacing of low-order tributaries By contrast in largerbasins (up to 300000 km2) the distances separating junctioneffects are on the order of several kilometers to tens ofkilometers (Baxter 2002 Benda et al 2003b)

420 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 5 Two basins in northern California illustratevariation in confluence densities Black dots representconfluences with tributaryndashmain stem ratios indicatedby dot size A watershed formed in relatively young andhomogeneous lava flows (top) has a drainage density of 3kilometers (km) channel length per square kilometerdrainage area and a corresponding confluence density of47 junctions per square kilometer A watershed locatedin the older and geologically heterogeneous terrain (bot-tom) has a drainage density of 5 km channel length persquare kilometer and a corresponding confluence densityof 12 junctions per square kilometer On the basis of thedifference in confluence density the second watershed ispredicted to have higher confluence-related habitatheterogeneity (box 1) The probability of confluence effects depends on the increasing size ratio of tributary to main stem drainage area as indicated by the size ofthe dots (Benda et al 2004) Estimation of drainage density or confluence density should be sensitive to mapscale and to the method of depicting channel networks

Figure 6 Local network geometry (kilometer scale) variesin response to the size spacing and confluence angles ofintersecting tributaries which reflect underlying geologicstructure topography and erosion history Different localgeometries are expected to lead to different patterns ofphysical heterogeneity linked to tributary confluences(a) With a number of large tributaries near the center of the network confluence-related heterogeneity is con-centrated in a central sediment-producing region Thispattern is characteristic of watersheds located alongmountain fronts and abutting depositional plains(b) When major tributaries are separated by canyonsthe associated areas of physical heterogeneity are dividedas well

Physical heterogeneity

a b

Higher Lower

Canyon

Tributaryndashmain stemdrainage area ratio

Tributaryndashmain stemdrainage area ratio

302 km2

299 km2

The second scaling property of confluence-related mor-phology is that the size of habitat patches associated with trib-utaries increases downstream (figure 7)Although the data aresparse the length of channels affected by confluences rangesfrom tens or a few hundreds of meters (in basins less than ap-proximately 100 km2) to several kilometers (in basins betweenapproximately 1000 and 300000 km2) We expect the size ofhabitat patches to increase farther downstream because thechannel gradient declines with increasing river size thereforeany vertical obstruction in a channel that backs up sediment(eg fans logjams boulders) should influence a channeldistance upstream at least equivalent to the obstructionheight divided by the tangent of the underlying stream gra-dient

Interaction of river networks with stochastic watershed processes A river network can be thought of as the landscape templatewithin which climatically induced stochastic fluctuations inthe supply and transport of water sediment and organicmaterial occur The structure of a river network (basin size

shape network configuration and so on fig-ure 1) can be used to help understand therole of stochastic watershed processes inshaping riverine environments

Disturbance The role of tributary conflu-ences on longitudinal patterns of riverineheterogeneity and the dependence of thisheterogeneity on basin scale basin shapenetwork pattern drainage density and net-work geometry can be framed as a generalset of testable predictions (see box 1) Thespatial configuration of a network thus pro-vides a template for organizing the transportand deposition of sediment wood and wa-ter through a watershed but the supply andtransport of material that creates confluenceeffects varies dynamically over time becauseof climate-driven events (storms floodsand fires) The watershed disturbanceregime (Miller et al 2003)mdashthe frequencymagnitude and spatial extent of episodic cli-matic and geomorphic processesmdashcan varywithin and among watersheds and thuscontribute to variation in heterogeneity atconfluences including the waxing and wan-ing of effects over time

Moderate to large disturbance-drivenlandscape-scale fluctuations in the supplyand storage of in-channel sediment and or-ganic material create many of the morpho-logical changes observed at confluencesFor example the fluctuating supply of sed-iment storage at confluences creates fansfloodplains terraces logjams secondary

channels and fans (table 1 Small 1973 Benda et al 2003b)The episodic nature of sediment-related disturbances createsthe form and age mosaic of the erosional and depositionallandforms characteristic of valley floors ultimately con-tributing to physical heterogeneity and potentially to biologicaldiversity and increased productivity (Swanson et al 1988)Al-though biological recovery processes such as vegetation suc-cession can also influence patterns of physical heterogeneityin the channel and valley floor (Gregory et al 1991) that topicis beyond the scope of this article For more comprehensivediscussions of the role of disturbances in riverine environ-ments see Resh and colleagues (1988) Reeves and colleagues(1995) Poff and colleagues (1997) and Nakamura and colleagues (2000)

Amplification of disturbances at confluences The channel dis-turbance regime can be altered at tributary confluences Firsttributaries represent abrupt increases in the supply of watersediment and wood and therefore channel responses re-lated to those inputs should have a higher frequency andmagnitude near or immediately downstream of confluences

May 2004 Vol 54 No 5 bull BioScience 421

Articles

Figure 7 Confluence effects in rivers (eg changes in substrate floodplains orchannel morphology) reveal two scaling properties First the distance separatinggeomorphically significant confluences (shown in meters) increases downstreamparticularly in dendritic networks (top) Second the length of altered habitatpatches associated with geomorphically significant confluences (also in meters)increases downstream (bottom)

Leng

th o

f inf

luen

ce (m

)S

paci

ng b

etw

een

geom

orph

ical

lysi

gnifi

cant

con

fluen

ces

(m)

Main stem drainage area (km2)

Main stem drainage area (km2)

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

downstream side of the junction including coarser substratesand increases in channel width pool depth and occurrenceof bars The same general classes of channel changes occur atconfluences regardless of their location in the river network(see table 1) although certain types of changes such as boul-der accumulations leading to rapids occur predominantly neardebris flow or flash flood deposits The morphological con-ditions near channel junctions differ from those in reaches lo-cated upstream or downstream confluences are agents ofhabitat formation and increased morphological heterogene-ity (figure 2 Rice et al 2001 Benda et al 2003a) In this ar-ticle we concentrate on the morphological effects at junctionslinked to tributary sources of sediment and wood althoughour analysis of the influences of river network geometryshould also apply to more flow-related changes in morphol-

ogy at junctions in less erosion-prone landscapes (eg Best1986 Rhoads 1987)

Effects of river networks on the structure of riverine habitatsThe physical structure of river networks can be defined bybasin size basin shape network pattern size difference betweenconfluent channels the power law of stream sizes (egHorton 1945) drainage density and confluence density andlocal network geometry (figure 1) Our predictions abouthow river network structure influences spatial patterns ofconfluence-related morphology (box 1) apply to a range ofchannel changes (eg changes in substrate size channelwidth or extent of floodplains and terraces) althoughmorphological effects can be broadly stratified according to

416 BioScience bull May 2004 Vol 54 No 5

Articles

Table 1 Fourteen studies documenting tributary confluence effects in 19 nonregulated streams and rivers across thewestern United States and Canada

Climatic Type of sediment Contributing Receiving MorphologicalLocation region transport stream area (km2) stream area (km2) effects Source

Sekiu Olympic Mountains Humid Debris flow 002 to 073 067 to 42 a b d g k Benda et al Washington 2003a

Ash Creek Arizona Arid Flash flood 042 98 d f Wohl and Pear-three 1991

Queen Charlotte Islands Humid Debris flow 011 to 56 03 to 120 k Hogan et al 1998British Columbia

Matheny and Sitkum Olympic Humid Debris flow 037 203 a b d g j k Benda et al Mountains Washington 2003a

Coast Range Oregon Humid Debris flow 008 to 027 08 to 30 a b c d e f Everest and g h j k l Meehan 1981

Sheep Creek Idaho Semiarid Alluvial 266 646 a b c d g Benda et al 2003b

Oregon Cascades Humid Debris flow 011 to 30 51 to 71 e f g Grant and Swanson 1995

Crooked River Idaho Semiarid Alluvial 34 219 a b c d g Benda et al 2003b

Bear Creek Colorado Semiarid Flash flood 59 to 239 193 to 407 e Grimm et al 1995

North Fork Boise River Idaho Semiarid Alluvial 06 to 29 322 to 461 a b c d f g h Benda et al 2003b

Wenaha River Oregon Semiarid Alluvial 18 to 71 446 to 516 f g Baxter 2002

Snoqualmie River Washington Humid Alluvial 85 to 750 712 to 1794 a b c d Booth et al 1991

Pine and Sukunka Rivers Semiarid Alluvial 23 to 203 1579 to 2145 d Rice et al 2001British Columbia

South Fork Payette River Idaho Semiarid Flash flood 055 2470 e Meyer and Pierce 2003

Bella Coola River Humid Alluvial 128 to 285 4779 to 5421 m Church 1983British Columbia

Middle Fork Salmon River Idaho Semiarid Debris flow 25 to 295 1176 to 7096 a b e Meyer and flash flood Leidecker 1999

Grand Ronde Oregon Semiarid Alluvial 764 to 1342 6953 to 7781 f g Baxter 2002

Snake River Oregon Semiarid Alluvial 9137 240765 g f Meyer and Pierce 2003

Colorado River Arid Debris flow 143 to 6076 280000 to 386800 a d e i Melis et al 1995(before dam) Colorado flash flood

a gradient steepening b gradient lowering c upstream sediment deposition d changing substrate size e boulders and rapids f terraces g flood-plains h side channels i midchannel bars j ponds k logjams l meanders m channel instability

Note Sites are arrayed according to increasing drainage area of the main stem (in square kilometers) The studies indicate the dominant type ofsediment transport the drainage areas of tributaries and main stems and the type of morphological effects at confluences indicated by the authors

upstream and downstream position relative to confluences(eg figure 2) We also predict local changes in heterogene-ity at confluences which will usually increase Heterogeneityis defined by the type form and age distribution of fluviallandforms It is not yet possible to develop quantitative pre-dictions about specific morphological changes at conflu-ences because of the low resolution of data (eg table 1)and the complex nature of riverine environments (Rhoads1987)

The role of basin size Consistent flow-related morphologicalchanges (ie in channel width and depth) occur at junc-tions where the ratio between tributary size and main stem

size approaches 06 or 07 (Rhoads 1987) We postulate thatmorphological effects caused by punctuated inputs of sedi-ment and wood at confluences will also scale to the size of thetributary relative to the main stem We anticipate this resultbecause larger basins typically produce larger quantities of sed-iment and because larger tributaries generally have larger fansassociated with them (Bull 1977) Moreover larger and morepowerful rivers are more effective at removing tributary inputs of sediment (Benda et al 2003a)

To evaluate this expectation we analyzed the results of 14published field studies that document a range of confluenceeffects in 19 streams and rivers caused by the abrupt intro-duction of sediment and wood (see table 1 for a listing of these

May 2004 Vol 54 No 5 bull BioScience 417

Articles

Testable predictions from the network dynamics hypothesis are listed below The predictions apply to a range ofmorphological changes at confluences (including morphological heterogeneity) although they can be broadly stratifiedaccording to upstream and downstream position relative to confluences (see eg figure 2) Presently it is not feasible to make quantitative predictions of channel changes at confluences because of the complexity of channel environments(Rhoads 1987) Confluences that have identifiable morphological effects (including those confluences listed in table 1)are referred to as ldquogeomorphically significant confluencesrdquo

Predictions related to network structurebull The likelihood of confluence effects increases with the ratio of tributary to main stem sizes

bull Heart-shaped and pear-shaped basins containing dendritic networks favor increasing tributary size and hence confluenceeffects downstream compared with rectangular basins containing trellis or parallel networks which do not

bull The separation distance between geomorphically significant junctions (and their morphological effects) increases downstream withincreasing basin size particularly in dendritic networks

bull As basin size increases the channel length and area affected by individual confluence-related channel and valley morphologicalmodifications will increase

bull Closely spaced tributaries will yield valley segments of higher physical heterogeneity compared with valley segments that do not contain closely spaced geomorphically significant confluences

bull Basins with higher drainage density and corresponding higher junction density will have a higher degree of morphological heterogeneity Correspondingly basins of high topographic roughness in part related to higher drainage density should contain a higher degree of riverine heterogeneity

Predictions related to watershed disturbances or dynamicsbull Basins with higher punctuated sediment supply and transport will be characterized by greater confluence effects

bull Channelized disturbances (ie floods and accelerated sediment and wood supply) will have greater frequency and magnitude prox-imal to and immediately downstream of confluences leading to greater physical heterogeneity including the age distribution of flu-vial landforms

bull Channelized disturbances will be magnified immediately upstream of geomorphically significant confluences leading to greaterphysical heterogeneity including the age distribution of fluvial landforms

bull In sufficiently large basins (approximately 102 km2) the age distribution of confluence-related landforms (ie fans terracesfloodplains) will be skewed toward older geomorphic features in headwaters and toward younger features in downstream portions of the basin

bull Physical heterogeneity (ie diversity of forms and ages of channels floodplains terraces and logjams) will be concentrated in certain parts of networks

Box 1 The influence of river network structure on spatial patterns of confluence-related morphology

effects)We refer to confluences that have observable morpho-logical effects as ldquogeomorphically significantrdquoTogether thesestudies describe 168 junctions along 560 kilometers (km) ofriver spanning 7 orders of magnitude in drainage area in thewestern United States and Canada (see Benda et al [2004] fora more detailed analysis) These data reveal that as the size ofthe main stem increases geomorphically significant conflu-ences are associated with increasingly larger tributaries (fig-ure 3) For example debris flows that originate from smallbasins (up to 1 km2 in drainage area) create tributary-junction effects in basins of only 1 to 50 km2 By contrasttributary-junction effects in larger rivers (1000 to 300000 km2)are associated with larger tributaries (10 to 10000 km2)

The data in figure 3 also reveal a threshold below whichtributary basins less than approximately 1 km2 do not af-fect main stem rivers larger than approximately 50 km2 Thisis significant because it indicates how the ldquolaw of streamnumbersrdquo (Horton 1945) the universal power law regard-ing the distribution of channel sizes (and by analogy of theconfluences linked to those channels) constrains confluenceeffects at the scale of entire networks In humid landscapesfor example the vast majority (80 to 90) of tributarybasins are first or second order with less than 1 km2 indrainage area (Benda and Dunne 1997a) On the basis ofthe threshold identified in figure 3 only 10 to 20 of trib-utary confluences (ie those of third order or higher) areavailable to significantly affect main stem rivers of a sizegreater than approximately 50 km2 This finding has im-portant ramifications for the increased downstream spac-ing of confluence effects as the size of the main stemincreases

The variability of thedata in figure 3 is probablydue to factors such as basingeology (eg the durabil-ity and size distribution ofsediment) sediment trans-port processes (eg debrisflows and flash floods thattransport boulders and aremore likely to cause con-fluence effects) and localvalley width (eg wide val-leys that limit fan forma-tion and the associatedconfluence effects) Tem-poral variation in the his-tory of storms fires andfloods that create or reju-venate confluence effectsshould also cause the wax-ing and waning of conflu-ence effectsA more detailedanalysis of the data in figure3 allows for probabilisticpredictions of confluence

effects (Benda et al 2004) For example in humid environ-ments a range of 06 to 09 for the probability of a confluenceeffect corresponds to a range of 004 to 08 for tributaryndashmainstem drainage area ratios

The role of network pattern and basin shape The scaling re-lationship between tributaries and main stem channels (fig-ure 3) allows us to consider how the factors that control thespatial distribution of tributary sizes in river networks influ-ence spatial patterns of confluence-related morphology andheterogeneity Downstream trends in junction effects are in-fluenced by network patterns and hence by drainage basinshape Two common types of network patterns are dendriticand trellis networks Dendritic networks which resemblethe hierarchical branching pattern of a tree often form inhomogeneous and gently sloped geologic beds and createheart-shaped or pear-shaped basins In contrast trellis net-works characterized by small tributaries intersecting mainstem channels are often associated with elongate landformsin dipped and folded sedimentary rocks or in areas of paral-lel fractures they create narrow rectangular basins

The spatial configuration of tributaries within a watershedchanges the likelihood of confluence effects downstream inriver networks Since larger tributaries are required to creategeomorphic effects as the size of the main stem increases(eg figure 3) dendritic networks in heart-shaped or pear-shaped basins should promote confluence effects throughoutthe watershed (figure 4 see Benda et al [2004] for more details) These effects occur because increasing the basinwidth downstream promotes the coalescing of hierarchicallybranched channels resulting in larger tributaries forming

418 BioScience bull May 2004 Vol 54 No 5

Articles

Drainage area of main stem river (km2)

Humid debris flowSemiarid debris flowflash floodHumid alluvialSemiarid alluvial

Transition from 1st and 2ndorder to higher-order channels

Dra

ina

ge a

rea

of trib

utar

y (k

m2)

Figure 3 A comparison between tributary basin area and main stem drainage area (in squarekilometers) for 168 tributaries where confluence effects (ie geomorphically significant conflu-ences) have been documented reveals that larger main stem rivers require larger tributaries inorder to experience confluence effects (table 1 Benda et al 2004) The dashed line represents theapproximate tributary drainage threshold for the first- and second-order streams that comprisethe vast majority of stream sizes and for their confluences in networks Power law distributionof channel (confluence) sizes is shown on the left side of the figure P refers to the proportion ofall channels of a given drainage area

downstream In contrast narrow rectangular basins con-taining trellis networks lack the formation of larger tributariesand therefore discourage confluence effects as main stemsize increases Consequently network configuration relatedto basin shape should strongly influence the downstreamsequence of confluence effects (figure 4 see also the testablepredictions in box 1)

In addition to dendritic and trellis networks other networktypes include parallel networks which form in conjunctionwith parallel landforms and rectangular networks whichform where numerous faults and joints converge at high an-gles Region-specific types of geology and hillslope topogra-phy should influence network patterns and hence the spatialdistribution of confluence-related channel morphology andassociated physical heterogeneity For example the youngand porous rocks of the high Cascade Mountains in Oregonare characterized by trellis networks while older and lesspermeable rocks nearby exhibit dendritic networks (Grant1997)

Drainage density and confluence density The cumulativeeffect of confluences within a basin should be proportionalto the total number of geomorphically significant channel con-fluences This number is related to drainage density (definedas channel length per unit watershed area) and to networkshape which either promotes or inhibits the occurrence ofconfluent channels (see figure 5) The corresponding con-fluence density (the number of geomorphically significantconfluences per unit area or per unit channel length) shouldprovide a simple measure of the net morphological effect ofconfluences in rivers (box 1) Drainage densities in humid tosemiarid landscapes range from 2 to 12 km channel length persquare kilometer watershed area primarily reflecting varia-tions in precipitation landscape age and bedrock porosity(Grant 1997) This large range in drainage density translatesto a correspondingly large range in the density of channel con-fluences with implications for the degree of channel hetero-geneity found in different landscapes (figure 5)

Local network geometry A riverrsquos geologic structure and tec-tonic and erosional history create specific patterns of inter-secting tributaries over kilometer scales Local networkgeometry can be used to describe the kilometer-scale varia-tion of tributary effects in rivers including the longitudinalsequence of tributaryndashmain stem size ratios (figures 3 4) trib-utary intersection angles and distance between geomorphi-cally significant confluences (figure 1d) The tributaryndashmainstem intersection angle is the upstream angle formed at a con-fluence Intersection angles are almost always acute and as theybecome less so and approach 90deg the likelihood of a geo-morphic effect at a confluence increases For instance in a se-ries of river studies and in flume experiments Mosley (1976)and Best (1986) showed how bar size bar location and scourdepth vary with confluence angle Lateral bars are more likelyto form at acute angles scour depth by contrast increases withincreasing confluence angle and approaches an asymptote at

angles approaching 90deg Moreover in headwater areas con-fluence angles greater than 70deg encourage debris flow depo-sition and consequent junction effects (table 1) whiledeposition is discouraged at junctions with more acute an-gles (Benda and Cundy 1990)

Network geometry also describes the distance separatinggeomorphically significant confluences For example a concentration of large tributaries in a central sediment-producing region is characteristic of watersheds along moun-tain fronts abutting depositional plains (figure 6a) Alterna-tively tributaries that are separated by canyons can lead toclumped distributions of geomorphically significant tribu-taries and associated physical heterogeneity (figure 6b)Large tributary junctions that are closely spaced may have con-

May 2004 Vol 54 No 5 bull BioScience 419

Articles

Distance along main stem channel

Rectangular

a

bHigh

Like

lihoo

d of

con

fluen

ce e

ffect

s

Low

Pear shaped

Heart shaped

Heart shaped(most compact)

Pear shaped(moderately compact)

Rectangular (most elongate)

Figure 4 (a) Network patterns and the associated basinshapes affect the downstream sequence of confluence effects Common network patterns include dendritic networks within heart-shaped and pear-shaped basinsand trellis networks within narrow rectangular basins(b) The anticipated downstream sequence of confluenceeffects (ranging from high to low likelihood) is based onthe size of the tributaries relative to the main stemsDendritic networks in heart-shaped basins promote thegreatest likelihood of confluence effects downstreamwhile trellis networks in rectangular basins promote thefewest effects downstream Modified from Benda and colleagues (2004)

Downstream sequence of confluence effects

Downstream sequence of confluence effects

fluence effects that overlap particularly during watersheddisturbances such as floods and concentrated basin erosionConfluence effects may also be less pronounced in wide valley floors where fans are isolated from the main stem bybroad terraces or floodplains

Scaling properties of confluence effects The finding that largertributaries are required in order to affect the morphology oflarger rivers reveals two scaling effects on habitat patches

that develop near confluences First geomorphically signifi-cant confluences should be separated by increasing distancesdownstream in watersheds where tributary basins down-stream increase in size (eg dendritic networks in pear-shaped and heart-shaped basins figure 4 box 1) Thisincreasing separation occurs because tributary length is relatedto tributary drainage area (Hack 1957) and thus the lengthand width of a basin increase with increasing tributary lengthThis spacing pattern is reflected in the field data (figure 7) Inthe upper portions of humid drainage basins morphologi-cal effects are spaced on average hundreds of meters apart(Benda 1990 Hogan et al 1998 Benda et al 2003a) reflect-ing the spacing of low-order tributaries By contrast in largerbasins (up to 300000 km2) the distances separating junctioneffects are on the order of several kilometers to tens ofkilometers (Baxter 2002 Benda et al 2003b)

420 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 5 Two basins in northern California illustratevariation in confluence densities Black dots representconfluences with tributaryndashmain stem ratios indicatedby dot size A watershed formed in relatively young andhomogeneous lava flows (top) has a drainage density of 3kilometers (km) channel length per square kilometerdrainage area and a corresponding confluence density of47 junctions per square kilometer A watershed locatedin the older and geologically heterogeneous terrain (bot-tom) has a drainage density of 5 km channel length persquare kilometer and a corresponding confluence densityof 12 junctions per square kilometer On the basis of thedifference in confluence density the second watershed ispredicted to have higher confluence-related habitatheterogeneity (box 1) The probability of confluence effects depends on the increasing size ratio of tributary to main stem drainage area as indicated by the size ofthe dots (Benda et al 2004) Estimation of drainage density or confluence density should be sensitive to mapscale and to the method of depicting channel networks

Figure 6 Local network geometry (kilometer scale) variesin response to the size spacing and confluence angles ofintersecting tributaries which reflect underlying geologicstructure topography and erosion history Different localgeometries are expected to lead to different patterns ofphysical heterogeneity linked to tributary confluences(a) With a number of large tributaries near the center of the network confluence-related heterogeneity is con-centrated in a central sediment-producing region Thispattern is characteristic of watersheds located alongmountain fronts and abutting depositional plains(b) When major tributaries are separated by canyonsthe associated areas of physical heterogeneity are dividedas well

Physical heterogeneity

a b

Higher Lower

Canyon

Tributaryndashmain stemdrainage area ratio

Tributaryndashmain stemdrainage area ratio

302 km2

299 km2

The second scaling property of confluence-related mor-phology is that the size of habitat patches associated with trib-utaries increases downstream (figure 7)Although the data aresparse the length of channels affected by confluences rangesfrom tens or a few hundreds of meters (in basins less than ap-proximately 100 km2) to several kilometers (in basins betweenapproximately 1000 and 300000 km2) We expect the size ofhabitat patches to increase farther downstream because thechannel gradient declines with increasing river size thereforeany vertical obstruction in a channel that backs up sediment(eg fans logjams boulders) should influence a channeldistance upstream at least equivalent to the obstructionheight divided by the tangent of the underlying stream gra-dient

Interaction of river networks with stochastic watershed processes A river network can be thought of as the landscape templatewithin which climatically induced stochastic fluctuations inthe supply and transport of water sediment and organicmaterial occur The structure of a river network (basin size

shape network configuration and so on fig-ure 1) can be used to help understand therole of stochastic watershed processes inshaping riverine environments

Disturbance The role of tributary conflu-ences on longitudinal patterns of riverineheterogeneity and the dependence of thisheterogeneity on basin scale basin shapenetwork pattern drainage density and net-work geometry can be framed as a generalset of testable predictions (see box 1) Thespatial configuration of a network thus pro-vides a template for organizing the transportand deposition of sediment wood and wa-ter through a watershed but the supply andtransport of material that creates confluenceeffects varies dynamically over time becauseof climate-driven events (storms floodsand fires) The watershed disturbanceregime (Miller et al 2003)mdashthe frequencymagnitude and spatial extent of episodic cli-matic and geomorphic processesmdashcan varywithin and among watersheds and thuscontribute to variation in heterogeneity atconfluences including the waxing and wan-ing of effects over time

Moderate to large disturbance-drivenlandscape-scale fluctuations in the supplyand storage of in-channel sediment and or-ganic material create many of the morpho-logical changes observed at confluencesFor example the fluctuating supply of sed-iment storage at confluences creates fansfloodplains terraces logjams secondary

channels and fans (table 1 Small 1973 Benda et al 2003b)The episodic nature of sediment-related disturbances createsthe form and age mosaic of the erosional and depositionallandforms characteristic of valley floors ultimately con-tributing to physical heterogeneity and potentially to biologicaldiversity and increased productivity (Swanson et al 1988)Al-though biological recovery processes such as vegetation suc-cession can also influence patterns of physical heterogeneityin the channel and valley floor (Gregory et al 1991) that topicis beyond the scope of this article For more comprehensivediscussions of the role of disturbances in riverine environ-ments see Resh and colleagues (1988) Reeves and colleagues(1995) Poff and colleagues (1997) and Nakamura and colleagues (2000)

Amplification of disturbances at confluences The channel dis-turbance regime can be altered at tributary confluences Firsttributaries represent abrupt increases in the supply of watersediment and wood and therefore channel responses re-lated to those inputs should have a higher frequency andmagnitude near or immediately downstream of confluences

May 2004 Vol 54 No 5 bull BioScience 421

Articles

Figure 7 Confluence effects in rivers (eg changes in substrate floodplains orchannel morphology) reveal two scaling properties First the distance separatinggeomorphically significant confluences (shown in meters) increases downstreamparticularly in dendritic networks (top) Second the length of altered habitatpatches associated with geomorphically significant confluences (also in meters)increases downstream (bottom)

Leng

th o

f inf

luen

ce (m

)S

paci

ng b

etw

een

geom

orph

ical

lysi

gnifi

cant

con

fluen

ces

(m)

Main stem drainage area (km2)

Main stem drainage area (km2)

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

upstream and downstream position relative to confluences(eg figure 2) We also predict local changes in heterogene-ity at confluences which will usually increase Heterogeneityis defined by the type form and age distribution of fluviallandforms It is not yet possible to develop quantitative pre-dictions about specific morphological changes at conflu-ences because of the low resolution of data (eg table 1)and the complex nature of riverine environments (Rhoads1987)

The role of basin size Consistent flow-related morphologicalchanges (ie in channel width and depth) occur at junc-tions where the ratio between tributary size and main stem

size approaches 06 or 07 (Rhoads 1987) We postulate thatmorphological effects caused by punctuated inputs of sedi-ment and wood at confluences will also scale to the size of thetributary relative to the main stem We anticipate this resultbecause larger basins typically produce larger quantities of sed-iment and because larger tributaries generally have larger fansassociated with them (Bull 1977) Moreover larger and morepowerful rivers are more effective at removing tributary inputs of sediment (Benda et al 2003a)

To evaluate this expectation we analyzed the results of 14published field studies that document a range of confluenceeffects in 19 streams and rivers caused by the abrupt intro-duction of sediment and wood (see table 1 for a listing of these

May 2004 Vol 54 No 5 bull BioScience 417

Articles

Testable predictions from the network dynamics hypothesis are listed below The predictions apply to a range ofmorphological changes at confluences (including morphological heterogeneity) although they can be broadly stratifiedaccording to upstream and downstream position relative to confluences (see eg figure 2) Presently it is not feasible to make quantitative predictions of channel changes at confluences because of the complexity of channel environments(Rhoads 1987) Confluences that have identifiable morphological effects (including those confluences listed in table 1)are referred to as ldquogeomorphically significant confluencesrdquo

Predictions related to network structurebull The likelihood of confluence effects increases with the ratio of tributary to main stem sizes

bull Heart-shaped and pear-shaped basins containing dendritic networks favor increasing tributary size and hence confluenceeffects downstream compared with rectangular basins containing trellis or parallel networks which do not

bull The separation distance between geomorphically significant junctions (and their morphological effects) increases downstream withincreasing basin size particularly in dendritic networks

bull As basin size increases the channel length and area affected by individual confluence-related channel and valley morphologicalmodifications will increase

bull Closely spaced tributaries will yield valley segments of higher physical heterogeneity compared with valley segments that do not contain closely spaced geomorphically significant confluences

bull Basins with higher drainage density and corresponding higher junction density will have a higher degree of morphological heterogeneity Correspondingly basins of high topographic roughness in part related to higher drainage density should contain a higher degree of riverine heterogeneity

Predictions related to watershed disturbances or dynamicsbull Basins with higher punctuated sediment supply and transport will be characterized by greater confluence effects

bull Channelized disturbances (ie floods and accelerated sediment and wood supply) will have greater frequency and magnitude prox-imal to and immediately downstream of confluences leading to greater physical heterogeneity including the age distribution of flu-vial landforms

bull Channelized disturbances will be magnified immediately upstream of geomorphically significant confluences leading to greaterphysical heterogeneity including the age distribution of fluvial landforms

bull In sufficiently large basins (approximately 102 km2) the age distribution of confluence-related landforms (ie fans terracesfloodplains) will be skewed toward older geomorphic features in headwaters and toward younger features in downstream portions of the basin

bull Physical heterogeneity (ie diversity of forms and ages of channels floodplains terraces and logjams) will be concentrated in certain parts of networks

Box 1 The influence of river network structure on spatial patterns of confluence-related morphology

effects)We refer to confluences that have observable morpho-logical effects as ldquogeomorphically significantrdquoTogether thesestudies describe 168 junctions along 560 kilometers (km) ofriver spanning 7 orders of magnitude in drainage area in thewestern United States and Canada (see Benda et al [2004] fora more detailed analysis) These data reveal that as the size ofthe main stem increases geomorphically significant conflu-ences are associated with increasingly larger tributaries (fig-ure 3) For example debris flows that originate from smallbasins (up to 1 km2 in drainage area) create tributary-junction effects in basins of only 1 to 50 km2 By contrasttributary-junction effects in larger rivers (1000 to 300000 km2)are associated with larger tributaries (10 to 10000 km2)

The data in figure 3 also reveal a threshold below whichtributary basins less than approximately 1 km2 do not af-fect main stem rivers larger than approximately 50 km2 Thisis significant because it indicates how the ldquolaw of streamnumbersrdquo (Horton 1945) the universal power law regard-ing the distribution of channel sizes (and by analogy of theconfluences linked to those channels) constrains confluenceeffects at the scale of entire networks In humid landscapesfor example the vast majority (80 to 90) of tributarybasins are first or second order with less than 1 km2 indrainage area (Benda and Dunne 1997a) On the basis ofthe threshold identified in figure 3 only 10 to 20 of trib-utary confluences (ie those of third order or higher) areavailable to significantly affect main stem rivers of a sizegreater than approximately 50 km2 This finding has im-portant ramifications for the increased downstream spac-ing of confluence effects as the size of the main stemincreases

The variability of thedata in figure 3 is probablydue to factors such as basingeology (eg the durabil-ity and size distribution ofsediment) sediment trans-port processes (eg debrisflows and flash floods thattransport boulders and aremore likely to cause con-fluence effects) and localvalley width (eg wide val-leys that limit fan forma-tion and the associatedconfluence effects) Tem-poral variation in the his-tory of storms fires andfloods that create or reju-venate confluence effectsshould also cause the wax-ing and waning of conflu-ence effectsA more detailedanalysis of the data in figure3 allows for probabilisticpredictions of confluence

effects (Benda et al 2004) For example in humid environ-ments a range of 06 to 09 for the probability of a confluenceeffect corresponds to a range of 004 to 08 for tributaryndashmainstem drainage area ratios

The role of network pattern and basin shape The scaling re-lationship between tributaries and main stem channels (fig-ure 3) allows us to consider how the factors that control thespatial distribution of tributary sizes in river networks influ-ence spatial patterns of confluence-related morphology andheterogeneity Downstream trends in junction effects are in-fluenced by network patterns and hence by drainage basinshape Two common types of network patterns are dendriticand trellis networks Dendritic networks which resemblethe hierarchical branching pattern of a tree often form inhomogeneous and gently sloped geologic beds and createheart-shaped or pear-shaped basins In contrast trellis net-works characterized by small tributaries intersecting mainstem channels are often associated with elongate landformsin dipped and folded sedimentary rocks or in areas of paral-lel fractures they create narrow rectangular basins

The spatial configuration of tributaries within a watershedchanges the likelihood of confluence effects downstream inriver networks Since larger tributaries are required to creategeomorphic effects as the size of the main stem increases(eg figure 3) dendritic networks in heart-shaped or pear-shaped basins should promote confluence effects throughoutthe watershed (figure 4 see Benda et al [2004] for more details) These effects occur because increasing the basinwidth downstream promotes the coalescing of hierarchicallybranched channels resulting in larger tributaries forming

418 BioScience bull May 2004 Vol 54 No 5

Articles

Drainage area of main stem river (km2)

Humid debris flowSemiarid debris flowflash floodHumid alluvialSemiarid alluvial

Transition from 1st and 2ndorder to higher-order channels

Dra

ina

ge a

rea

of trib

utar

y (k

m2)

Figure 3 A comparison between tributary basin area and main stem drainage area (in squarekilometers) for 168 tributaries where confluence effects (ie geomorphically significant conflu-ences) have been documented reveals that larger main stem rivers require larger tributaries inorder to experience confluence effects (table 1 Benda et al 2004) The dashed line represents theapproximate tributary drainage threshold for the first- and second-order streams that comprisethe vast majority of stream sizes and for their confluences in networks Power law distributionof channel (confluence) sizes is shown on the left side of the figure P refers to the proportion ofall channels of a given drainage area

downstream In contrast narrow rectangular basins con-taining trellis networks lack the formation of larger tributariesand therefore discourage confluence effects as main stemsize increases Consequently network configuration relatedto basin shape should strongly influence the downstreamsequence of confluence effects (figure 4 see also the testablepredictions in box 1)

In addition to dendritic and trellis networks other networktypes include parallel networks which form in conjunctionwith parallel landforms and rectangular networks whichform where numerous faults and joints converge at high an-gles Region-specific types of geology and hillslope topogra-phy should influence network patterns and hence the spatialdistribution of confluence-related channel morphology andassociated physical heterogeneity For example the youngand porous rocks of the high Cascade Mountains in Oregonare characterized by trellis networks while older and lesspermeable rocks nearby exhibit dendritic networks (Grant1997)

Drainage density and confluence density The cumulativeeffect of confluences within a basin should be proportionalto the total number of geomorphically significant channel con-fluences This number is related to drainage density (definedas channel length per unit watershed area) and to networkshape which either promotes or inhibits the occurrence ofconfluent channels (see figure 5) The corresponding con-fluence density (the number of geomorphically significantconfluences per unit area or per unit channel length) shouldprovide a simple measure of the net morphological effect ofconfluences in rivers (box 1) Drainage densities in humid tosemiarid landscapes range from 2 to 12 km channel length persquare kilometer watershed area primarily reflecting varia-tions in precipitation landscape age and bedrock porosity(Grant 1997) This large range in drainage density translatesto a correspondingly large range in the density of channel con-fluences with implications for the degree of channel hetero-geneity found in different landscapes (figure 5)

Local network geometry A riverrsquos geologic structure and tec-tonic and erosional history create specific patterns of inter-secting tributaries over kilometer scales Local networkgeometry can be used to describe the kilometer-scale varia-tion of tributary effects in rivers including the longitudinalsequence of tributaryndashmain stem size ratios (figures 3 4) trib-utary intersection angles and distance between geomorphi-cally significant confluences (figure 1d) The tributaryndashmainstem intersection angle is the upstream angle formed at a con-fluence Intersection angles are almost always acute and as theybecome less so and approach 90deg the likelihood of a geo-morphic effect at a confluence increases For instance in a se-ries of river studies and in flume experiments Mosley (1976)and Best (1986) showed how bar size bar location and scourdepth vary with confluence angle Lateral bars are more likelyto form at acute angles scour depth by contrast increases withincreasing confluence angle and approaches an asymptote at

angles approaching 90deg Moreover in headwater areas con-fluence angles greater than 70deg encourage debris flow depo-sition and consequent junction effects (table 1) whiledeposition is discouraged at junctions with more acute an-gles (Benda and Cundy 1990)

Network geometry also describes the distance separatinggeomorphically significant confluences For example a concentration of large tributaries in a central sediment-producing region is characteristic of watersheds along moun-tain fronts abutting depositional plains (figure 6a) Alterna-tively tributaries that are separated by canyons can lead toclumped distributions of geomorphically significant tribu-taries and associated physical heterogeneity (figure 6b)Large tributary junctions that are closely spaced may have con-

May 2004 Vol 54 No 5 bull BioScience 419

Articles

Distance along main stem channel

Rectangular

a

bHigh

Like

lihoo

d of

con

fluen

ce e

ffect

s

Low

Pear shaped

Heart shaped

Heart shaped(most compact)

Pear shaped(moderately compact)

Rectangular (most elongate)

Figure 4 (a) Network patterns and the associated basinshapes affect the downstream sequence of confluence effects Common network patterns include dendritic networks within heart-shaped and pear-shaped basinsand trellis networks within narrow rectangular basins(b) The anticipated downstream sequence of confluenceeffects (ranging from high to low likelihood) is based onthe size of the tributaries relative to the main stemsDendritic networks in heart-shaped basins promote thegreatest likelihood of confluence effects downstreamwhile trellis networks in rectangular basins promote thefewest effects downstream Modified from Benda and colleagues (2004)

Downstream sequence of confluence effects

Downstream sequence of confluence effects

fluence effects that overlap particularly during watersheddisturbances such as floods and concentrated basin erosionConfluence effects may also be less pronounced in wide valley floors where fans are isolated from the main stem bybroad terraces or floodplains

Scaling properties of confluence effects The finding that largertributaries are required in order to affect the morphology oflarger rivers reveals two scaling effects on habitat patches

that develop near confluences First geomorphically signifi-cant confluences should be separated by increasing distancesdownstream in watersheds where tributary basins down-stream increase in size (eg dendritic networks in pear-shaped and heart-shaped basins figure 4 box 1) Thisincreasing separation occurs because tributary length is relatedto tributary drainage area (Hack 1957) and thus the lengthand width of a basin increase with increasing tributary lengthThis spacing pattern is reflected in the field data (figure 7) Inthe upper portions of humid drainage basins morphologi-cal effects are spaced on average hundreds of meters apart(Benda 1990 Hogan et al 1998 Benda et al 2003a) reflect-ing the spacing of low-order tributaries By contrast in largerbasins (up to 300000 km2) the distances separating junctioneffects are on the order of several kilometers to tens ofkilometers (Baxter 2002 Benda et al 2003b)

420 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 5 Two basins in northern California illustratevariation in confluence densities Black dots representconfluences with tributaryndashmain stem ratios indicatedby dot size A watershed formed in relatively young andhomogeneous lava flows (top) has a drainage density of 3kilometers (km) channel length per square kilometerdrainage area and a corresponding confluence density of47 junctions per square kilometer A watershed locatedin the older and geologically heterogeneous terrain (bot-tom) has a drainage density of 5 km channel length persquare kilometer and a corresponding confluence densityof 12 junctions per square kilometer On the basis of thedifference in confluence density the second watershed ispredicted to have higher confluence-related habitatheterogeneity (box 1) The probability of confluence effects depends on the increasing size ratio of tributary to main stem drainage area as indicated by the size ofthe dots (Benda et al 2004) Estimation of drainage density or confluence density should be sensitive to mapscale and to the method of depicting channel networks

Figure 6 Local network geometry (kilometer scale) variesin response to the size spacing and confluence angles ofintersecting tributaries which reflect underlying geologicstructure topography and erosion history Different localgeometries are expected to lead to different patterns ofphysical heterogeneity linked to tributary confluences(a) With a number of large tributaries near the center of the network confluence-related heterogeneity is con-centrated in a central sediment-producing region Thispattern is characteristic of watersheds located alongmountain fronts and abutting depositional plains(b) When major tributaries are separated by canyonsthe associated areas of physical heterogeneity are dividedas well

Physical heterogeneity

a b

Higher Lower

Canyon

Tributaryndashmain stemdrainage area ratio

Tributaryndashmain stemdrainage area ratio

302 km2

299 km2

The second scaling property of confluence-related mor-phology is that the size of habitat patches associated with trib-utaries increases downstream (figure 7)Although the data aresparse the length of channels affected by confluences rangesfrom tens or a few hundreds of meters (in basins less than ap-proximately 100 km2) to several kilometers (in basins betweenapproximately 1000 and 300000 km2) We expect the size ofhabitat patches to increase farther downstream because thechannel gradient declines with increasing river size thereforeany vertical obstruction in a channel that backs up sediment(eg fans logjams boulders) should influence a channeldistance upstream at least equivalent to the obstructionheight divided by the tangent of the underlying stream gra-dient

Interaction of river networks with stochastic watershed processes A river network can be thought of as the landscape templatewithin which climatically induced stochastic fluctuations inthe supply and transport of water sediment and organicmaterial occur The structure of a river network (basin size

shape network configuration and so on fig-ure 1) can be used to help understand therole of stochastic watershed processes inshaping riverine environments

Disturbance The role of tributary conflu-ences on longitudinal patterns of riverineheterogeneity and the dependence of thisheterogeneity on basin scale basin shapenetwork pattern drainage density and net-work geometry can be framed as a generalset of testable predictions (see box 1) Thespatial configuration of a network thus pro-vides a template for organizing the transportand deposition of sediment wood and wa-ter through a watershed but the supply andtransport of material that creates confluenceeffects varies dynamically over time becauseof climate-driven events (storms floodsand fires) The watershed disturbanceregime (Miller et al 2003)mdashthe frequencymagnitude and spatial extent of episodic cli-matic and geomorphic processesmdashcan varywithin and among watersheds and thuscontribute to variation in heterogeneity atconfluences including the waxing and wan-ing of effects over time

Moderate to large disturbance-drivenlandscape-scale fluctuations in the supplyand storage of in-channel sediment and or-ganic material create many of the morpho-logical changes observed at confluencesFor example the fluctuating supply of sed-iment storage at confluences creates fansfloodplains terraces logjams secondary

channels and fans (table 1 Small 1973 Benda et al 2003b)The episodic nature of sediment-related disturbances createsthe form and age mosaic of the erosional and depositionallandforms characteristic of valley floors ultimately con-tributing to physical heterogeneity and potentially to biologicaldiversity and increased productivity (Swanson et al 1988)Al-though biological recovery processes such as vegetation suc-cession can also influence patterns of physical heterogeneityin the channel and valley floor (Gregory et al 1991) that topicis beyond the scope of this article For more comprehensivediscussions of the role of disturbances in riverine environ-ments see Resh and colleagues (1988) Reeves and colleagues(1995) Poff and colleagues (1997) and Nakamura and colleagues (2000)

Amplification of disturbances at confluences The channel dis-turbance regime can be altered at tributary confluences Firsttributaries represent abrupt increases in the supply of watersediment and wood and therefore channel responses re-lated to those inputs should have a higher frequency andmagnitude near or immediately downstream of confluences

May 2004 Vol 54 No 5 bull BioScience 421

Articles

Figure 7 Confluence effects in rivers (eg changes in substrate floodplains orchannel morphology) reveal two scaling properties First the distance separatinggeomorphically significant confluences (shown in meters) increases downstreamparticularly in dendritic networks (top) Second the length of altered habitatpatches associated with geomorphically significant confluences (also in meters)increases downstream (bottom)

Leng

th o

f inf

luen

ce (m

)S

paci

ng b

etw

een

geom

orph

ical

lysi

gnifi

cant

con

fluen

ces

(m)

Main stem drainage area (km2)

Main stem drainage area (km2)

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

effects)We refer to confluences that have observable morpho-logical effects as ldquogeomorphically significantrdquoTogether thesestudies describe 168 junctions along 560 kilometers (km) ofriver spanning 7 orders of magnitude in drainage area in thewestern United States and Canada (see Benda et al [2004] fora more detailed analysis) These data reveal that as the size ofthe main stem increases geomorphically significant conflu-ences are associated with increasingly larger tributaries (fig-ure 3) For example debris flows that originate from smallbasins (up to 1 km2 in drainage area) create tributary-junction effects in basins of only 1 to 50 km2 By contrasttributary-junction effects in larger rivers (1000 to 300000 km2)are associated with larger tributaries (10 to 10000 km2)

The data in figure 3 also reveal a threshold below whichtributary basins less than approximately 1 km2 do not af-fect main stem rivers larger than approximately 50 km2 Thisis significant because it indicates how the ldquolaw of streamnumbersrdquo (Horton 1945) the universal power law regard-ing the distribution of channel sizes (and by analogy of theconfluences linked to those channels) constrains confluenceeffects at the scale of entire networks In humid landscapesfor example the vast majority (80 to 90) of tributarybasins are first or second order with less than 1 km2 indrainage area (Benda and Dunne 1997a) On the basis ofthe threshold identified in figure 3 only 10 to 20 of trib-utary confluences (ie those of third order or higher) areavailable to significantly affect main stem rivers of a sizegreater than approximately 50 km2 This finding has im-portant ramifications for the increased downstream spac-ing of confluence effects as the size of the main stemincreases

The variability of thedata in figure 3 is probablydue to factors such as basingeology (eg the durabil-ity and size distribution ofsediment) sediment trans-port processes (eg debrisflows and flash floods thattransport boulders and aremore likely to cause con-fluence effects) and localvalley width (eg wide val-leys that limit fan forma-tion and the associatedconfluence effects) Tem-poral variation in the his-tory of storms fires andfloods that create or reju-venate confluence effectsshould also cause the wax-ing and waning of conflu-ence effectsA more detailedanalysis of the data in figure3 allows for probabilisticpredictions of confluence

effects (Benda et al 2004) For example in humid environ-ments a range of 06 to 09 for the probability of a confluenceeffect corresponds to a range of 004 to 08 for tributaryndashmainstem drainage area ratios

The role of network pattern and basin shape The scaling re-lationship between tributaries and main stem channels (fig-ure 3) allows us to consider how the factors that control thespatial distribution of tributary sizes in river networks influ-ence spatial patterns of confluence-related morphology andheterogeneity Downstream trends in junction effects are in-fluenced by network patterns and hence by drainage basinshape Two common types of network patterns are dendriticand trellis networks Dendritic networks which resemblethe hierarchical branching pattern of a tree often form inhomogeneous and gently sloped geologic beds and createheart-shaped or pear-shaped basins In contrast trellis net-works characterized by small tributaries intersecting mainstem channels are often associated with elongate landformsin dipped and folded sedimentary rocks or in areas of paral-lel fractures they create narrow rectangular basins

The spatial configuration of tributaries within a watershedchanges the likelihood of confluence effects downstream inriver networks Since larger tributaries are required to creategeomorphic effects as the size of the main stem increases(eg figure 3) dendritic networks in heart-shaped or pear-shaped basins should promote confluence effects throughoutthe watershed (figure 4 see Benda et al [2004] for more details) These effects occur because increasing the basinwidth downstream promotes the coalescing of hierarchicallybranched channels resulting in larger tributaries forming

418 BioScience bull May 2004 Vol 54 No 5

Articles

Drainage area of main stem river (km2)

Humid debris flowSemiarid debris flowflash floodHumid alluvialSemiarid alluvial

Transition from 1st and 2ndorder to higher-order channels

Dra

ina

ge a

rea

of trib

utar

y (k

m2)

Figure 3 A comparison between tributary basin area and main stem drainage area (in squarekilometers) for 168 tributaries where confluence effects (ie geomorphically significant conflu-ences) have been documented reveals that larger main stem rivers require larger tributaries inorder to experience confluence effects (table 1 Benda et al 2004) The dashed line represents theapproximate tributary drainage threshold for the first- and second-order streams that comprisethe vast majority of stream sizes and for their confluences in networks Power law distributionof channel (confluence) sizes is shown on the left side of the figure P refers to the proportion ofall channels of a given drainage area

downstream In contrast narrow rectangular basins con-taining trellis networks lack the formation of larger tributariesand therefore discourage confluence effects as main stemsize increases Consequently network configuration relatedto basin shape should strongly influence the downstreamsequence of confluence effects (figure 4 see also the testablepredictions in box 1)

In addition to dendritic and trellis networks other networktypes include parallel networks which form in conjunctionwith parallel landforms and rectangular networks whichform where numerous faults and joints converge at high an-gles Region-specific types of geology and hillslope topogra-phy should influence network patterns and hence the spatialdistribution of confluence-related channel morphology andassociated physical heterogeneity For example the youngand porous rocks of the high Cascade Mountains in Oregonare characterized by trellis networks while older and lesspermeable rocks nearby exhibit dendritic networks (Grant1997)

Drainage density and confluence density The cumulativeeffect of confluences within a basin should be proportionalto the total number of geomorphically significant channel con-fluences This number is related to drainage density (definedas channel length per unit watershed area) and to networkshape which either promotes or inhibits the occurrence ofconfluent channels (see figure 5) The corresponding con-fluence density (the number of geomorphically significantconfluences per unit area or per unit channel length) shouldprovide a simple measure of the net morphological effect ofconfluences in rivers (box 1) Drainage densities in humid tosemiarid landscapes range from 2 to 12 km channel length persquare kilometer watershed area primarily reflecting varia-tions in precipitation landscape age and bedrock porosity(Grant 1997) This large range in drainage density translatesto a correspondingly large range in the density of channel con-fluences with implications for the degree of channel hetero-geneity found in different landscapes (figure 5)

Local network geometry A riverrsquos geologic structure and tec-tonic and erosional history create specific patterns of inter-secting tributaries over kilometer scales Local networkgeometry can be used to describe the kilometer-scale varia-tion of tributary effects in rivers including the longitudinalsequence of tributaryndashmain stem size ratios (figures 3 4) trib-utary intersection angles and distance between geomorphi-cally significant confluences (figure 1d) The tributaryndashmainstem intersection angle is the upstream angle formed at a con-fluence Intersection angles are almost always acute and as theybecome less so and approach 90deg the likelihood of a geo-morphic effect at a confluence increases For instance in a se-ries of river studies and in flume experiments Mosley (1976)and Best (1986) showed how bar size bar location and scourdepth vary with confluence angle Lateral bars are more likelyto form at acute angles scour depth by contrast increases withincreasing confluence angle and approaches an asymptote at

angles approaching 90deg Moreover in headwater areas con-fluence angles greater than 70deg encourage debris flow depo-sition and consequent junction effects (table 1) whiledeposition is discouraged at junctions with more acute an-gles (Benda and Cundy 1990)

Network geometry also describes the distance separatinggeomorphically significant confluences For example a concentration of large tributaries in a central sediment-producing region is characteristic of watersheds along moun-tain fronts abutting depositional plains (figure 6a) Alterna-tively tributaries that are separated by canyons can lead toclumped distributions of geomorphically significant tribu-taries and associated physical heterogeneity (figure 6b)Large tributary junctions that are closely spaced may have con-

May 2004 Vol 54 No 5 bull BioScience 419

Articles

Distance along main stem channel

Rectangular

a

bHigh

Like

lihoo

d of

con

fluen

ce e

ffect

s

Low

Pear shaped

Heart shaped

Heart shaped(most compact)

Pear shaped(moderately compact)

Rectangular (most elongate)

Figure 4 (a) Network patterns and the associated basinshapes affect the downstream sequence of confluence effects Common network patterns include dendritic networks within heart-shaped and pear-shaped basinsand trellis networks within narrow rectangular basins(b) The anticipated downstream sequence of confluenceeffects (ranging from high to low likelihood) is based onthe size of the tributaries relative to the main stemsDendritic networks in heart-shaped basins promote thegreatest likelihood of confluence effects downstreamwhile trellis networks in rectangular basins promote thefewest effects downstream Modified from Benda and colleagues (2004)

Downstream sequence of confluence effects

Downstream sequence of confluence effects

fluence effects that overlap particularly during watersheddisturbances such as floods and concentrated basin erosionConfluence effects may also be less pronounced in wide valley floors where fans are isolated from the main stem bybroad terraces or floodplains

Scaling properties of confluence effects The finding that largertributaries are required in order to affect the morphology oflarger rivers reveals two scaling effects on habitat patches

that develop near confluences First geomorphically signifi-cant confluences should be separated by increasing distancesdownstream in watersheds where tributary basins down-stream increase in size (eg dendritic networks in pear-shaped and heart-shaped basins figure 4 box 1) Thisincreasing separation occurs because tributary length is relatedto tributary drainage area (Hack 1957) and thus the lengthand width of a basin increase with increasing tributary lengthThis spacing pattern is reflected in the field data (figure 7) Inthe upper portions of humid drainage basins morphologi-cal effects are spaced on average hundreds of meters apart(Benda 1990 Hogan et al 1998 Benda et al 2003a) reflect-ing the spacing of low-order tributaries By contrast in largerbasins (up to 300000 km2) the distances separating junctioneffects are on the order of several kilometers to tens ofkilometers (Baxter 2002 Benda et al 2003b)

420 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 5 Two basins in northern California illustratevariation in confluence densities Black dots representconfluences with tributaryndashmain stem ratios indicatedby dot size A watershed formed in relatively young andhomogeneous lava flows (top) has a drainage density of 3kilometers (km) channel length per square kilometerdrainage area and a corresponding confluence density of47 junctions per square kilometer A watershed locatedin the older and geologically heterogeneous terrain (bot-tom) has a drainage density of 5 km channel length persquare kilometer and a corresponding confluence densityof 12 junctions per square kilometer On the basis of thedifference in confluence density the second watershed ispredicted to have higher confluence-related habitatheterogeneity (box 1) The probability of confluence effects depends on the increasing size ratio of tributary to main stem drainage area as indicated by the size ofthe dots (Benda et al 2004) Estimation of drainage density or confluence density should be sensitive to mapscale and to the method of depicting channel networks

Figure 6 Local network geometry (kilometer scale) variesin response to the size spacing and confluence angles ofintersecting tributaries which reflect underlying geologicstructure topography and erosion history Different localgeometries are expected to lead to different patterns ofphysical heterogeneity linked to tributary confluences(a) With a number of large tributaries near the center of the network confluence-related heterogeneity is con-centrated in a central sediment-producing region Thispattern is characteristic of watersheds located alongmountain fronts and abutting depositional plains(b) When major tributaries are separated by canyonsthe associated areas of physical heterogeneity are dividedas well

Physical heterogeneity

a b

Higher Lower

Canyon

Tributaryndashmain stemdrainage area ratio

Tributaryndashmain stemdrainage area ratio

302 km2

299 km2

The second scaling property of confluence-related mor-phology is that the size of habitat patches associated with trib-utaries increases downstream (figure 7)Although the data aresparse the length of channels affected by confluences rangesfrom tens or a few hundreds of meters (in basins less than ap-proximately 100 km2) to several kilometers (in basins betweenapproximately 1000 and 300000 km2) We expect the size ofhabitat patches to increase farther downstream because thechannel gradient declines with increasing river size thereforeany vertical obstruction in a channel that backs up sediment(eg fans logjams boulders) should influence a channeldistance upstream at least equivalent to the obstructionheight divided by the tangent of the underlying stream gra-dient

Interaction of river networks with stochastic watershed processes A river network can be thought of as the landscape templatewithin which climatically induced stochastic fluctuations inthe supply and transport of water sediment and organicmaterial occur The structure of a river network (basin size

shape network configuration and so on fig-ure 1) can be used to help understand therole of stochastic watershed processes inshaping riverine environments

Disturbance The role of tributary conflu-ences on longitudinal patterns of riverineheterogeneity and the dependence of thisheterogeneity on basin scale basin shapenetwork pattern drainage density and net-work geometry can be framed as a generalset of testable predictions (see box 1) Thespatial configuration of a network thus pro-vides a template for organizing the transportand deposition of sediment wood and wa-ter through a watershed but the supply andtransport of material that creates confluenceeffects varies dynamically over time becauseof climate-driven events (storms floodsand fires) The watershed disturbanceregime (Miller et al 2003)mdashthe frequencymagnitude and spatial extent of episodic cli-matic and geomorphic processesmdashcan varywithin and among watersheds and thuscontribute to variation in heterogeneity atconfluences including the waxing and wan-ing of effects over time

Moderate to large disturbance-drivenlandscape-scale fluctuations in the supplyand storage of in-channel sediment and or-ganic material create many of the morpho-logical changes observed at confluencesFor example the fluctuating supply of sed-iment storage at confluences creates fansfloodplains terraces logjams secondary

channels and fans (table 1 Small 1973 Benda et al 2003b)The episodic nature of sediment-related disturbances createsthe form and age mosaic of the erosional and depositionallandforms characteristic of valley floors ultimately con-tributing to physical heterogeneity and potentially to biologicaldiversity and increased productivity (Swanson et al 1988)Al-though biological recovery processes such as vegetation suc-cession can also influence patterns of physical heterogeneityin the channel and valley floor (Gregory et al 1991) that topicis beyond the scope of this article For more comprehensivediscussions of the role of disturbances in riverine environ-ments see Resh and colleagues (1988) Reeves and colleagues(1995) Poff and colleagues (1997) and Nakamura and colleagues (2000)

Amplification of disturbances at confluences The channel dis-turbance regime can be altered at tributary confluences Firsttributaries represent abrupt increases in the supply of watersediment and wood and therefore channel responses re-lated to those inputs should have a higher frequency andmagnitude near or immediately downstream of confluences

May 2004 Vol 54 No 5 bull BioScience 421

Articles

Figure 7 Confluence effects in rivers (eg changes in substrate floodplains orchannel morphology) reveal two scaling properties First the distance separatinggeomorphically significant confluences (shown in meters) increases downstreamparticularly in dendritic networks (top) Second the length of altered habitatpatches associated with geomorphically significant confluences (also in meters)increases downstream (bottom)

Leng

th o

f inf

luen

ce (m

)S

paci

ng b

etw

een

geom

orph

ical

lysi

gnifi

cant

con

fluen

ces

(m)

Main stem drainage area (km2)

Main stem drainage area (km2)

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

downstream In contrast narrow rectangular basins con-taining trellis networks lack the formation of larger tributariesand therefore discourage confluence effects as main stemsize increases Consequently network configuration relatedto basin shape should strongly influence the downstreamsequence of confluence effects (figure 4 see also the testablepredictions in box 1)

In addition to dendritic and trellis networks other networktypes include parallel networks which form in conjunctionwith parallel landforms and rectangular networks whichform where numerous faults and joints converge at high an-gles Region-specific types of geology and hillslope topogra-phy should influence network patterns and hence the spatialdistribution of confluence-related channel morphology andassociated physical heterogeneity For example the youngand porous rocks of the high Cascade Mountains in Oregonare characterized by trellis networks while older and lesspermeable rocks nearby exhibit dendritic networks (Grant1997)

Drainage density and confluence density The cumulativeeffect of confluences within a basin should be proportionalto the total number of geomorphically significant channel con-fluences This number is related to drainage density (definedas channel length per unit watershed area) and to networkshape which either promotes or inhibits the occurrence ofconfluent channels (see figure 5) The corresponding con-fluence density (the number of geomorphically significantconfluences per unit area or per unit channel length) shouldprovide a simple measure of the net morphological effect ofconfluences in rivers (box 1) Drainage densities in humid tosemiarid landscapes range from 2 to 12 km channel length persquare kilometer watershed area primarily reflecting varia-tions in precipitation landscape age and bedrock porosity(Grant 1997) This large range in drainage density translatesto a correspondingly large range in the density of channel con-fluences with implications for the degree of channel hetero-geneity found in different landscapes (figure 5)

Local network geometry A riverrsquos geologic structure and tec-tonic and erosional history create specific patterns of inter-secting tributaries over kilometer scales Local networkgeometry can be used to describe the kilometer-scale varia-tion of tributary effects in rivers including the longitudinalsequence of tributaryndashmain stem size ratios (figures 3 4) trib-utary intersection angles and distance between geomorphi-cally significant confluences (figure 1d) The tributaryndashmainstem intersection angle is the upstream angle formed at a con-fluence Intersection angles are almost always acute and as theybecome less so and approach 90deg the likelihood of a geo-morphic effect at a confluence increases For instance in a se-ries of river studies and in flume experiments Mosley (1976)and Best (1986) showed how bar size bar location and scourdepth vary with confluence angle Lateral bars are more likelyto form at acute angles scour depth by contrast increases withincreasing confluence angle and approaches an asymptote at

angles approaching 90deg Moreover in headwater areas con-fluence angles greater than 70deg encourage debris flow depo-sition and consequent junction effects (table 1) whiledeposition is discouraged at junctions with more acute an-gles (Benda and Cundy 1990)

Network geometry also describes the distance separatinggeomorphically significant confluences For example a concentration of large tributaries in a central sediment-producing region is characteristic of watersheds along moun-tain fronts abutting depositional plains (figure 6a) Alterna-tively tributaries that are separated by canyons can lead toclumped distributions of geomorphically significant tribu-taries and associated physical heterogeneity (figure 6b)Large tributary junctions that are closely spaced may have con-

May 2004 Vol 54 No 5 bull BioScience 419

Articles

Distance along main stem channel

Rectangular

a

bHigh

Like

lihoo

d of

con

fluen

ce e

ffect

s

Low

Pear shaped

Heart shaped

Heart shaped(most compact)

Pear shaped(moderately compact)

Rectangular (most elongate)

Figure 4 (a) Network patterns and the associated basinshapes affect the downstream sequence of confluence effects Common network patterns include dendritic networks within heart-shaped and pear-shaped basinsand trellis networks within narrow rectangular basins(b) The anticipated downstream sequence of confluenceeffects (ranging from high to low likelihood) is based onthe size of the tributaries relative to the main stemsDendritic networks in heart-shaped basins promote thegreatest likelihood of confluence effects downstreamwhile trellis networks in rectangular basins promote thefewest effects downstream Modified from Benda and colleagues (2004)

Downstream sequence of confluence effects

Downstream sequence of confluence effects

fluence effects that overlap particularly during watersheddisturbances such as floods and concentrated basin erosionConfluence effects may also be less pronounced in wide valley floors where fans are isolated from the main stem bybroad terraces or floodplains

Scaling properties of confluence effects The finding that largertributaries are required in order to affect the morphology oflarger rivers reveals two scaling effects on habitat patches

that develop near confluences First geomorphically signifi-cant confluences should be separated by increasing distancesdownstream in watersheds where tributary basins down-stream increase in size (eg dendritic networks in pear-shaped and heart-shaped basins figure 4 box 1) Thisincreasing separation occurs because tributary length is relatedto tributary drainage area (Hack 1957) and thus the lengthand width of a basin increase with increasing tributary lengthThis spacing pattern is reflected in the field data (figure 7) Inthe upper portions of humid drainage basins morphologi-cal effects are spaced on average hundreds of meters apart(Benda 1990 Hogan et al 1998 Benda et al 2003a) reflect-ing the spacing of low-order tributaries By contrast in largerbasins (up to 300000 km2) the distances separating junctioneffects are on the order of several kilometers to tens ofkilometers (Baxter 2002 Benda et al 2003b)

420 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 5 Two basins in northern California illustratevariation in confluence densities Black dots representconfluences with tributaryndashmain stem ratios indicatedby dot size A watershed formed in relatively young andhomogeneous lava flows (top) has a drainage density of 3kilometers (km) channel length per square kilometerdrainage area and a corresponding confluence density of47 junctions per square kilometer A watershed locatedin the older and geologically heterogeneous terrain (bot-tom) has a drainage density of 5 km channel length persquare kilometer and a corresponding confluence densityof 12 junctions per square kilometer On the basis of thedifference in confluence density the second watershed ispredicted to have higher confluence-related habitatheterogeneity (box 1) The probability of confluence effects depends on the increasing size ratio of tributary to main stem drainage area as indicated by the size ofthe dots (Benda et al 2004) Estimation of drainage density or confluence density should be sensitive to mapscale and to the method of depicting channel networks

Figure 6 Local network geometry (kilometer scale) variesin response to the size spacing and confluence angles ofintersecting tributaries which reflect underlying geologicstructure topography and erosion history Different localgeometries are expected to lead to different patterns ofphysical heterogeneity linked to tributary confluences(a) With a number of large tributaries near the center of the network confluence-related heterogeneity is con-centrated in a central sediment-producing region Thispattern is characteristic of watersheds located alongmountain fronts and abutting depositional plains(b) When major tributaries are separated by canyonsthe associated areas of physical heterogeneity are dividedas well

Physical heterogeneity

a b

Higher Lower

Canyon

Tributaryndashmain stemdrainage area ratio

Tributaryndashmain stemdrainage area ratio

302 km2

299 km2

The second scaling property of confluence-related mor-phology is that the size of habitat patches associated with trib-utaries increases downstream (figure 7)Although the data aresparse the length of channels affected by confluences rangesfrom tens or a few hundreds of meters (in basins less than ap-proximately 100 km2) to several kilometers (in basins betweenapproximately 1000 and 300000 km2) We expect the size ofhabitat patches to increase farther downstream because thechannel gradient declines with increasing river size thereforeany vertical obstruction in a channel that backs up sediment(eg fans logjams boulders) should influence a channeldistance upstream at least equivalent to the obstructionheight divided by the tangent of the underlying stream gra-dient

Interaction of river networks with stochastic watershed processes A river network can be thought of as the landscape templatewithin which climatically induced stochastic fluctuations inthe supply and transport of water sediment and organicmaterial occur The structure of a river network (basin size

shape network configuration and so on fig-ure 1) can be used to help understand therole of stochastic watershed processes inshaping riverine environments

Disturbance The role of tributary conflu-ences on longitudinal patterns of riverineheterogeneity and the dependence of thisheterogeneity on basin scale basin shapenetwork pattern drainage density and net-work geometry can be framed as a generalset of testable predictions (see box 1) Thespatial configuration of a network thus pro-vides a template for organizing the transportand deposition of sediment wood and wa-ter through a watershed but the supply andtransport of material that creates confluenceeffects varies dynamically over time becauseof climate-driven events (storms floodsand fires) The watershed disturbanceregime (Miller et al 2003)mdashthe frequencymagnitude and spatial extent of episodic cli-matic and geomorphic processesmdashcan varywithin and among watersheds and thuscontribute to variation in heterogeneity atconfluences including the waxing and wan-ing of effects over time

Moderate to large disturbance-drivenlandscape-scale fluctuations in the supplyand storage of in-channel sediment and or-ganic material create many of the morpho-logical changes observed at confluencesFor example the fluctuating supply of sed-iment storage at confluences creates fansfloodplains terraces logjams secondary

channels and fans (table 1 Small 1973 Benda et al 2003b)The episodic nature of sediment-related disturbances createsthe form and age mosaic of the erosional and depositionallandforms characteristic of valley floors ultimately con-tributing to physical heterogeneity and potentially to biologicaldiversity and increased productivity (Swanson et al 1988)Al-though biological recovery processes such as vegetation suc-cession can also influence patterns of physical heterogeneityin the channel and valley floor (Gregory et al 1991) that topicis beyond the scope of this article For more comprehensivediscussions of the role of disturbances in riverine environ-ments see Resh and colleagues (1988) Reeves and colleagues(1995) Poff and colleagues (1997) and Nakamura and colleagues (2000)

Amplification of disturbances at confluences The channel dis-turbance regime can be altered at tributary confluences Firsttributaries represent abrupt increases in the supply of watersediment and wood and therefore channel responses re-lated to those inputs should have a higher frequency andmagnitude near or immediately downstream of confluences

May 2004 Vol 54 No 5 bull BioScience 421

Articles

Figure 7 Confluence effects in rivers (eg changes in substrate floodplains orchannel morphology) reveal two scaling properties First the distance separatinggeomorphically significant confluences (shown in meters) increases downstreamparticularly in dendritic networks (top) Second the length of altered habitatpatches associated with geomorphically significant confluences (also in meters)increases downstream (bottom)

Leng

th o

f inf

luen

ce (m

)S

paci

ng b

etw

een

geom

orph

ical

lysi

gnifi

cant

con

fluen

ces

(m)

Main stem drainage area (km2)

Main stem drainage area (km2)

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

fluence effects that overlap particularly during watersheddisturbances such as floods and concentrated basin erosionConfluence effects may also be less pronounced in wide valley floors where fans are isolated from the main stem bybroad terraces or floodplains

Scaling properties of confluence effects The finding that largertributaries are required in order to affect the morphology oflarger rivers reveals two scaling effects on habitat patches

that develop near confluences First geomorphically signifi-cant confluences should be separated by increasing distancesdownstream in watersheds where tributary basins down-stream increase in size (eg dendritic networks in pear-shaped and heart-shaped basins figure 4 box 1) Thisincreasing separation occurs because tributary length is relatedto tributary drainage area (Hack 1957) and thus the lengthand width of a basin increase with increasing tributary lengthThis spacing pattern is reflected in the field data (figure 7) Inthe upper portions of humid drainage basins morphologi-cal effects are spaced on average hundreds of meters apart(Benda 1990 Hogan et al 1998 Benda et al 2003a) reflect-ing the spacing of low-order tributaries By contrast in largerbasins (up to 300000 km2) the distances separating junctioneffects are on the order of several kilometers to tens ofkilometers (Baxter 2002 Benda et al 2003b)

420 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 5 Two basins in northern California illustratevariation in confluence densities Black dots representconfluences with tributaryndashmain stem ratios indicatedby dot size A watershed formed in relatively young andhomogeneous lava flows (top) has a drainage density of 3kilometers (km) channel length per square kilometerdrainage area and a corresponding confluence density of47 junctions per square kilometer A watershed locatedin the older and geologically heterogeneous terrain (bot-tom) has a drainage density of 5 km channel length persquare kilometer and a corresponding confluence densityof 12 junctions per square kilometer On the basis of thedifference in confluence density the second watershed ispredicted to have higher confluence-related habitatheterogeneity (box 1) The probability of confluence effects depends on the increasing size ratio of tributary to main stem drainage area as indicated by the size ofthe dots (Benda et al 2004) Estimation of drainage density or confluence density should be sensitive to mapscale and to the method of depicting channel networks

Figure 6 Local network geometry (kilometer scale) variesin response to the size spacing and confluence angles ofintersecting tributaries which reflect underlying geologicstructure topography and erosion history Different localgeometries are expected to lead to different patterns ofphysical heterogeneity linked to tributary confluences(a) With a number of large tributaries near the center of the network confluence-related heterogeneity is con-centrated in a central sediment-producing region Thispattern is characteristic of watersheds located alongmountain fronts and abutting depositional plains(b) When major tributaries are separated by canyonsthe associated areas of physical heterogeneity are dividedas well

Physical heterogeneity

a b

Higher Lower

Canyon

Tributaryndashmain stemdrainage area ratio

Tributaryndashmain stemdrainage area ratio

302 km2

299 km2

The second scaling property of confluence-related mor-phology is that the size of habitat patches associated with trib-utaries increases downstream (figure 7)Although the data aresparse the length of channels affected by confluences rangesfrom tens or a few hundreds of meters (in basins less than ap-proximately 100 km2) to several kilometers (in basins betweenapproximately 1000 and 300000 km2) We expect the size ofhabitat patches to increase farther downstream because thechannel gradient declines with increasing river size thereforeany vertical obstruction in a channel that backs up sediment(eg fans logjams boulders) should influence a channeldistance upstream at least equivalent to the obstructionheight divided by the tangent of the underlying stream gra-dient

Interaction of river networks with stochastic watershed processes A river network can be thought of as the landscape templatewithin which climatically induced stochastic fluctuations inthe supply and transport of water sediment and organicmaterial occur The structure of a river network (basin size

shape network configuration and so on fig-ure 1) can be used to help understand therole of stochastic watershed processes inshaping riverine environments

Disturbance The role of tributary conflu-ences on longitudinal patterns of riverineheterogeneity and the dependence of thisheterogeneity on basin scale basin shapenetwork pattern drainage density and net-work geometry can be framed as a generalset of testable predictions (see box 1) Thespatial configuration of a network thus pro-vides a template for organizing the transportand deposition of sediment wood and wa-ter through a watershed but the supply andtransport of material that creates confluenceeffects varies dynamically over time becauseof climate-driven events (storms floodsand fires) The watershed disturbanceregime (Miller et al 2003)mdashthe frequencymagnitude and spatial extent of episodic cli-matic and geomorphic processesmdashcan varywithin and among watersheds and thuscontribute to variation in heterogeneity atconfluences including the waxing and wan-ing of effects over time

Moderate to large disturbance-drivenlandscape-scale fluctuations in the supplyand storage of in-channel sediment and or-ganic material create many of the morpho-logical changes observed at confluencesFor example the fluctuating supply of sed-iment storage at confluences creates fansfloodplains terraces logjams secondary

channels and fans (table 1 Small 1973 Benda et al 2003b)The episodic nature of sediment-related disturbances createsthe form and age mosaic of the erosional and depositionallandforms characteristic of valley floors ultimately con-tributing to physical heterogeneity and potentially to biologicaldiversity and increased productivity (Swanson et al 1988)Al-though biological recovery processes such as vegetation suc-cession can also influence patterns of physical heterogeneityin the channel and valley floor (Gregory et al 1991) that topicis beyond the scope of this article For more comprehensivediscussions of the role of disturbances in riverine environ-ments see Resh and colleagues (1988) Reeves and colleagues(1995) Poff and colleagues (1997) and Nakamura and colleagues (2000)

Amplification of disturbances at confluences The channel dis-turbance regime can be altered at tributary confluences Firsttributaries represent abrupt increases in the supply of watersediment and wood and therefore channel responses re-lated to those inputs should have a higher frequency andmagnitude near or immediately downstream of confluences

May 2004 Vol 54 No 5 bull BioScience 421

Articles

Figure 7 Confluence effects in rivers (eg changes in substrate floodplains orchannel morphology) reveal two scaling properties First the distance separatinggeomorphically significant confluences (shown in meters) increases downstreamparticularly in dendritic networks (top) Second the length of altered habitatpatches associated with geomorphically significant confluences (also in meters)increases downstream (bottom)

Leng

th o

f inf

luen

ce (m

)S

paci

ng b

etw

een

geom

orph

ical

lysi

gnifi

cant

con

fluen

ces

(m)

Main stem drainage area (km2)

Main stem drainage area (km2)

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

The second scaling property of confluence-related mor-phology is that the size of habitat patches associated with trib-utaries increases downstream (figure 7)Although the data aresparse the length of channels affected by confluences rangesfrom tens or a few hundreds of meters (in basins less than ap-proximately 100 km2) to several kilometers (in basins betweenapproximately 1000 and 300000 km2) We expect the size ofhabitat patches to increase farther downstream because thechannel gradient declines with increasing river size thereforeany vertical obstruction in a channel that backs up sediment(eg fans logjams boulders) should influence a channeldistance upstream at least equivalent to the obstructionheight divided by the tangent of the underlying stream gra-dient

Interaction of river networks with stochastic watershed processes A river network can be thought of as the landscape templatewithin which climatically induced stochastic fluctuations inthe supply and transport of water sediment and organicmaterial occur The structure of a river network (basin size

shape network configuration and so on fig-ure 1) can be used to help understand therole of stochastic watershed processes inshaping riverine environments

Disturbance The role of tributary conflu-ences on longitudinal patterns of riverineheterogeneity and the dependence of thisheterogeneity on basin scale basin shapenetwork pattern drainage density and net-work geometry can be framed as a generalset of testable predictions (see box 1) Thespatial configuration of a network thus pro-vides a template for organizing the transportand deposition of sediment wood and wa-ter through a watershed but the supply andtransport of material that creates confluenceeffects varies dynamically over time becauseof climate-driven events (storms floodsand fires) The watershed disturbanceregime (Miller et al 2003)mdashthe frequencymagnitude and spatial extent of episodic cli-matic and geomorphic processesmdashcan varywithin and among watersheds and thuscontribute to variation in heterogeneity atconfluences including the waxing and wan-ing of effects over time

Moderate to large disturbance-drivenlandscape-scale fluctuations in the supplyand storage of in-channel sediment and or-ganic material create many of the morpho-logical changes observed at confluencesFor example the fluctuating supply of sed-iment storage at confluences creates fansfloodplains terraces logjams secondary

channels and fans (table 1 Small 1973 Benda et al 2003b)The episodic nature of sediment-related disturbances createsthe form and age mosaic of the erosional and depositionallandforms characteristic of valley floors ultimately con-tributing to physical heterogeneity and potentially to biologicaldiversity and increased productivity (Swanson et al 1988)Al-though biological recovery processes such as vegetation suc-cession can also influence patterns of physical heterogeneityin the channel and valley floor (Gregory et al 1991) that topicis beyond the scope of this article For more comprehensivediscussions of the role of disturbances in riverine environ-ments see Resh and colleagues (1988) Reeves and colleagues(1995) Poff and colleagues (1997) and Nakamura and colleagues (2000)

Amplification of disturbances at confluences The channel dis-turbance regime can be altered at tributary confluences Firsttributaries represent abrupt increases in the supply of watersediment and wood and therefore channel responses re-lated to those inputs should have a higher frequency andmagnitude near or immediately downstream of confluences

May 2004 Vol 54 No 5 bull BioScience 421

Articles

Figure 7 Confluence effects in rivers (eg changes in substrate floodplains orchannel morphology) reveal two scaling properties First the distance separatinggeomorphically significant confluences (shown in meters) increases downstreamparticularly in dendritic networks (top) Second the length of altered habitatpatches associated with geomorphically significant confluences (also in meters)increases downstream (bottom)

Leng

th o

f inf

luen

ce (m

)S

paci

ng b

etw

een

geom

orph

ical

lysi

gnifi

cant

con

fluen

ces

(m)

Main stem drainage area (km2)

Main stem drainage area (km2)

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

Second reductions in channel gradient and expansion ofvalley floors upstream of confluences (eg figure 2) can interfere with the downstream and fluctuating transport ofsediment and wood Where increased sediment and woodstorage occurs upstream of confluences that lead to wider andlower-gradient (ie more responsive) channels the magni-tude of flow-related disturbances can increase These inter-actions have been observed in the field and predicted bysimulation models (Benda and Dunne 1997a) For exampleChurch (1983) noted that the highest rates of channel me-andering occurred in the vicinity of tributary junctions incoastal British Columbia Similarly Jacobson (1995) docu-mented that the greatest fluctuations in sediment storageoccurred at and near confluences

Since the supply and transport of significant amounts ofsediment is episodic depositional areas at confluences in-cluding fans (eg figure 2) should expand and contract overtime in response to storms fires and floods (Benda et al2003b) Consequently the spatial extent of these areasrsquo up-stream and downstream zones of influence should vary overtime (figure 8) Moreover a disturbance that originates from

one tributary (ie a flash flood) is more likely toaffect a downstream area of a confluence if thetwo tributaries are relatively close together

River network organization of disturbanceregimes The locally altered disturbance regimeat a confluence is embedded within a larger pat-tern of disturbance frequency and magnitudedictated by the hierarchical nature of branchingriver networks Periods of flooding (and thus ofaccelerated sediment supply and transport)should increase in frequency and decrease inmagnitude downstream because of several fea-tures of the coupled climatendashlandscape system(figure 9) First the typical inverse relationshipbetween storm size and intensity causes floodhydrographs to be most spiked and erosion eventsto be most concentrated in small subbasins(Church 1998 Miller et al 2003) Second asriver size increases downstream (on the order of100 km2) through the intersection of tributariesthe number of potential sources of erosion risesabruptly at each confluence increasing the tempoof sediment supply and transport (Benda andDunne 1997a) As channels widen and sedimentstorage capacity increases downstream how-ever it becomes more difficult to create largesediment-related disturbances (eg figure 9)Consequently although the frequency of sediment-related disturbances should increase downstreamspecifically at confluences their magnitudeshould decline (Benda and Dunne 1997b)

This pattern of sediment-related disturbancefrequency and magnitude throughout a network(figure 9) has implications for confluence-

related morphology Large-magnitude sediment pulses orig-inating from concentrated floods and erosion in the upperregions of networks affected by large storms and fires havea frequency on the order of many decades to a couple ofcenturies (Swanson et al 1982 Meyer et al 2001) Hence anysnapshot of the age distribution of fluvial landforms in head-waters at the mouths of small basins is likely to be skewed to-ward older eroded features whose effects on main stemchannels are minor and dependent on the time since the lastepisodic input (Benda et al 2004) In contrast at the tribu-tary mouths of larger basins characterized by more frequentand lower-magnitude sediment pulses during floods the flu-vial landforms should have a higher proportion of youngerfeatures with more persistent effects in main stem rivers (box 1)

Concentration of heterogeneity in river networks The increasedmorphological heterogeneity at confluences is controlled bythe size ratios of confluent tributaries by the power law ofstream and confluence sizes by network patterns by local network geometry and by the river networksrsquo organization of

422 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 8 Channel disturbances (from fires storms or floods) are amplifiedat confluences because tributary basins are point sources of water sedi-ment and wood and because the topographic knick points created by confluences (eg shallowing of stream gradient) can interfere with thetransport of sediment and wood from upstream (Black dots represent tributary confluences) Because of the timing of disturbances confluence effects expand and contract over time and these effects could overlap withone another during periods of heightened watershed disturbance

Zone ofconfluence effect

Heighteneddisturbance

Littledisturbance

Disturbed watersheds (burned)

Undisturbed watersheds

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

disturbance regimes Because of these controls itis likely that riverine heterogeneity (linked to con-fluences) is nonuniformly distributed across wa-tersheds and it may even be concentrated in certainregions of networks For example in topographi-cally uniform watersheds there may be a zone ofmaximum physical heterogeneity located in thecentral network Relatively close spacing of geo-morphically significant tributaries (eg figure 6a)interacting with a moderate disturbance frequency(and disturbance magnitude figure 9) could max-imize physical heterogeneity between headwatersand basin mouths (box 1) This has been predictedusing a simulation model (Benda and Dunne1997b) Moreover dendritic networks in heart-shaped basins may promote the highest likelihoodof confluence effects in the central network whereconfluencing tributaries are largest (eg figure 4)

Watersheds that have sharply declining valleygradients as they transition from mountains todepositional plain in combination with large trib-utaries converging near the area of transitioncould create a zone of concentrated heterogeneitybetween the mountain uplands and lowlands Forinstance in the unregulated Queets River basin(1170 km2) the highest rate of channel meander-ing and the largest number of gravel bars and side channels occurred in an area approximatelymidway between the headwaters and the mouth(OrsquoConnor et al 2003) The localized convergenceof large tributaries in watersheds with heightenederosion may also focus channel dynamics and associated heterogeneity In the Ozark Plateau forinstance Jacobson (1995) found that the stream bedelevation changed more frequently in midsize watersheds (1400 to 7000 km2) where ldquosedimentwaves combined additively at confluencesrdquo than inchannels located in smaller (lt 1400 km2) or larger(8000 to 10000 km2) watersheds where the fre-quency and magnitude of perturbation were lower

Ecological implications New concepts in riverine ecology are focusing onriver attributes at landscapes scales (eg river-scapes) on their heterogeneity and on the role ofstochastic disturbances in shaping themYet the prevailing per-spective is of a river network as a linear feature hobbling newconcepts in riverine ecology (Fisher 1997) Viewing rivers asnetworks is fundamental to the new landscape view of rivers

The network dynamics hypothesis A physical foundationin riverine ecology New conceptual frameworks in river-ine ecology emphasize the importance of habitat hetero-geneity stochastic disturbances and scaling issues(Townsend 1989 Schlosser 1991 Fausch et al 2002 Poole2002 Ward et al 2002) Although the new conceptual

frameworks hold the potential to advance the discipline ofriver ecology they are hindered in doing so because they lacka physical basis for predicting how stochastic disturbancesor watershed dynamics interact with the spatial structureof river networks to generate patterns of heterogeneity inthe habitat along river profiles and throughout entire watersheds Framed as a set of testable predictions in box1 the network dynamics hypothesis (NDH) contributes aphysically based framework to underpin new conceptualframeworks in the coupled fields of geomorphology andaquatic biology

May 2004 Vol 54 No 5 bull BioScience 423

Articles

Figure 9 Because watershed dynamics (the frequency and magnitude ofsediment-related disturbances) vary with basin size they influence theage distribution of confluence-related landforms and consequently the effects of these landforms in channels (a) Disturbances are large but rarein headwaters (as indicated by the time series of channel sediment stor-age) leading to a higher proportion (P) of older confluence-related land-forms in headwater channels (b c) Disturbances are more frequent but oflower magnitude farther downstream in a network This should create ahigher proportion of younger confluence-related landforms and hencemore persistent confluence effects The highest frequency of intermediate-size disturbances is predicted to occur in the central network (b) con-tributing to a zone of heightened channel disturbance and maximumphysical heterogeneity The relationship between disturbance frequencyand magnitude can be represented in the form of probability distributions(insets for a b and c modified from Benda and Dunne 1997b) the shapeof which evolves downstream from skewed to more symmetrical formsBlack dots on the map represent confluences

Mean depth (m)

Mean depth (m)

Mean depth (m)

a

b

c

299 km2

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

By focusing on variation rather than on expected meanstates in fluvial geomorphic features the NDH complementsthe emphasis on habitat heterogeneity in landscape views ofrivers and in hierarchical patch dynamics This focus on vari-ation allows for understanding the degree of deviation inchannel geomorphic attributes and habitat heterogeneitycaused by punctuated inputs of water sediment and organicmaterial from tributaries along the longitudinal profile of anystream channel Thus proceeding down a river investigatorscan predict that deviations from the expected mean state(ie patchy heterogeneity) will occur in response to networkgeometry (figure 10) Because the NDH also addresses the sto-chastic dimension of sediment flux and storage in a drainagenetwork it can lay a foundation for emerging paradigms inriver ecology concerning the regulation of local disturbanceregimes according to their position in a network especiallyat confluences

This physical foundation has im-portant implications for much-needed generalization in riverineecology Because of differences inbasin size shape network patternchannel-type sequencing basintopography and disturbance re-gimes the variation in riverine en-vironments is virtually infinite Thishas led some researchers to concludethat all river networks are uniquelyindividual (Poole 2002) Althoughthis is certainly true in some ab-solute sense the NDH indicates thatsome generalities can nonethelessbe deduced on the basis of certainuniversal properties of river net-works

Tributary confluences as biologicalhotspots and their organization byriver networks Tributary junctionsrepresent locations in a networkwhere channel and valley morphol-ogy can change and where local het-erogeneity can be enhanced relativeto the central tendency expected un-der the river continuum concept(figure 10) Spatial and temporalheterogeneity in resources and habi-tat may among other things con-tribute to increased local speciesrichness (Huston 1994) thereforetributary junctions may representbiological hotspots within a rivernetwork

There is limited empirical evi-dence demonstrating the ecologicalimportance of morphologically di-

verse tributary junctions Rice and colleagues (2001) discov-ered changes in the abundance and composition ofmacroinvertebrate species in association with sediment sizedifferences at confluences in British Columbia Kupferberg(1996) found that a native frog species (Rana boylii) focusesits breeding nonrandomly near tributary junctions along a 5-km stretch of a fourth-order stream in a relatively steep Cal-ifornia river This species selected shallower and slower areasof stream bed within this zone Furthermore recent and ex-tensive river surveys in the Delaware River (PennsylvaniaNew York and New Jersey) suggest that the average abundanceof unionid mussels is greater in reaches above tributary junc-tions than in reaches below them a difference that may be dri-ven by differential sediment sorting and bed stability duringhigh river flows (William Lellis Chesapeake Watershed Co-operative Ecosystem Studies Unit University of MarylandFrostburg MD personal communication December 2002)

424 BioScience bull May 2004 Vol 54 No 5

Articles

Figure 10 (a) The river continuum concept (RCC) some predictions of which areshown for a main stem river exemplifies the prevailing linear perspective on riverineecology Linear approaches such as the RCC predict gradual and continuous down-stream change with central tendencies in physical and biological processes (b) In thenonlinear network-variance model the branching character of river networks coupledwith stochastic watershed disturbance interrupts downstream continua of physicaland biological processes to generate hypothetical deviations from downstream centraltendencies in the geomorphic properties along the main stem of the network For somevariables the central tendency shown in panel a may be retained but with elevatedvariance around tributary junctions (eg slope substrate) whereas for others the con-fluence effect of tributaries may eliminate the pattern of central tendency downstream(eg bank erosion width) within certain sizes of drainage basins

a Central tendencies

b Network variance

From linear to network perspectives

Bank erosion(cm per yr)

Dis

tanc

e do

wns

trea

mD

ista

nce

dow

nstr

eam

Stream order

Slope () Substrate(cm) Width (m)

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

Despite the lack of biological studies focusing on conflu-ences general ecological principles allow us to infer other likelyeffects For macroinvertebrates and fish increasing the het-erogeneity of habitat conditions including channel widthand depth bed substrate wood storage and water velocityshould increase total species richness (Allan 1995) For ex-ample increased channel-wide habitat diversity above sig-nificant tributary confluences should create habitatopportunities such as side channels during flood events forspecialized aquatic species or for life stages that require low-energy environments (eg Swales and Levings 1989 Reeveset al 1995) This has been documented in the Oregon CoastRange where increased wood storage and pool formation atlow-order confluences resulted in increased salmonid rear-ing (Everest and Meehan 1981) In the Olympic MountainsWashington correlations among low-order confluencesmdashwhich are prone to debris flows large pools gravel depositsand accumulation of woodmdashpromoted increased availabil-ity of fish habitat Likewise for riparian communities greatertopographic variation in floodplains and terraces should cre-ate local variation in inundation and soil moisture regimesthereby increasing plant diversity (eg Kalliola and Puharta1988 Pollock et al 1998) or provide interannual variation inriparian plant recruitment (Cooper et al 2003) Such enhancedfloodplain and terrace topography has been documentedupstream of alluvial fans (eg figure 2 Small 1973 Benda etal 2003b)

Other ecosystem processes may also be amplified at trib-utary junctions For example the localized flattening ofstream gradient and slowing of water velocity upstream ofa tributary fan can increase hydraulic head and enhance hyporheic flow through wedges of gravel substrate (Edwards1998) In this environment dissolved organic nitrogen can bechemically transformed while in hyporheic transit and emergein surface water as dissolved nitrate which can support in-creased primary productivity (Sedell and Dahm 1984) Sim-ilarly emerging hyporheic flow typically exhibits a smallerrange of temperature extremes which is favored by certain fish(eg see Baxter and Hauer 2000)

Of course tributaries can also modify environmental con-ditions other than the sediment and morphology of receiv-ing rivers The details of these modifications will be complexand will depend on the relative sizes and geochemical char-acteristics of the main stem and tributary streams Higher in-puts of nutrients and invertebrates from tributaries havebeen shown to promote primary and secondary productiv-ity in receiving streams (eg Kiffney and Richardson 2001Wipfli and Gregovich 2002) Fish may also use tributarymouths as thermal refugia (eg Scarnecchia and Roper 2000)or as dispersal corridors that support higher than expectedspecies diversity (Osborne and Wiley 1992)

ConclusionsStarting with the basic observation that the likelihood ofmorphologically significant perturbations to main stem chan-nels increases with the ratio of tributary to main stem size

we deduced a set of predictions relating the degree and spatial distribution of physical heterogeneity in a river systemto general features of branching river networks (figure 1box 1) This set of testable predictionsmdashknown collectivelyas the network dynamics hypothesismdashprovides a new frame-work for considering how the spatial structures of river net-works combined with time-varying watershed disturbancescreate and maintain habitat heterogeneity and thus potentially promote biological diversity and productivity in riverineecosystems

The network dynamics hypothesis can serve as a physicallybased framework for recent advances in watershed-scale geo-morphology and aquatic biology namely hierarchical patchdynamics (Townsend 1989 Wu and Loucks 1995 Poole2002) and the application of landscape ecology to river sys-tems (Schlosser 1991 Fausch et al 2002 Ward et al 2002Wiens 2002) Consequently this hypothesis could provide afoundation for new research in riverine ecology It also hasramifications for land management and for conservation orrestoration strategies For example network maps based onthe location of geomorphically interesting confluences couldbe generated to identify the highest likelihood of physical het-erogeneity (Benda et al 2004) Putative biological hotspotscould thus be identified on the basis of network configura-tion and watershed disturbance regimes Also land man-agement practices that alter the spatial and temporaldistribution of sediment and organic matter (eg damsdikes and forestry practices) could be examined in terms oftheir effects on downstream habitat heterogeneity as medi-ated through confluences in specific network structures Suchanalyses could contribute to the development of strategies fortargeted restoration efforts in a whole-watershed context

AcknowledgmentsThe development of this article was supported by a Bio-complexity in the Environment grant (DEB-0083937) fromthe National Science Foundation The present version was sub-stantially improved by comments made on a previous draftby three anonymous reviewers

References citedAllan JD 1995 Stream Ecology Structure and Function of Running Waters

London Chapman and HallBaxter CV 2002 Fish movement and assemblage dynamics in a Pacific

Northwest riverscape PhD dissertation Oregon State University Cor-vallis

Baxter CV Hauer FR 2000 Geomorphology hyporheic exchange and se-lection of spawning habitat by bull trout (Salvelinus confluentus) Cana-dian Journal of Fisheries and Aquatic Sciences 57 1470ndash1481

Benda L 1990 The influence of debris flows on channels and valley floorsof the Oregon Coast Range USA Earth Surface Processes and Land-forms 15 457ndash466

Benda L Cundy TW 1990 Predicting deposition of debris flows in moun-tain channels Canadian Geotechnical Journal 27 409ndash417

Benda L Dunne T 1997a Stochastic forcing of sediment supply to thechannel network from landsliding and debris flow Water ResourcesResearch 33 2849ndash2863

mdashmdashmdash 1997b Stochastic forcing of sediment routing and storage in chan-nel networks Water Resources Research 33 2865ndash2880

May 2004 Vol 54 No 5 bull BioScience 425

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

Benda LVeldhuisen C Black J 2003a Debris flows as agents of morphologicalheterogeneity at low-order confluences Olympic Mountains Washing-ton Geological Society of America Bulletin 115 1110ndash1121

Benda L Miller D Bigelow P Andrus K 2003b Effects of post-wildfire ero-sion on channel environments Boise River Idaho Journal of ForestEcology and Management 178 105ndash119

Benda L Andras K Miller D Bigelow P 2004 Effects of tributaries in rivernetworks Role of basin scale basin shape network geometry and dis-turbance regimes Water Resources Research Forthcoming

Best JL 1986 Sediment transport and bed morphology at river channel con-fluences Sedimentology 35 481ndash498

Booth DB Bell K Whipple KX 1991 Sediment Transport along the SouthFork and Mainstem of the Snoqualmie River Seattle King County De-partment of Public Works Surface Water Management Division

Bruns DA Minshall GW Cushing CE Cummins KW Brock JTVannote RC1984 Tributaries as modifiers of the river continuum concept Analysisby polar ordination and regression models Archiv fuumlr Hydrobiologie 99208ndash220

Bull WB 1977 The alluvial fan environment Progress in Physical Geogra-phy 1 222ndash270

Church M 1983 Pattern of instability in a wandering gravel bed channelSpecial Publications of the International Association of Sedimentologists6 169ndash180

mdashmdashmdash 1998 The landscape of the Pacific Northwest Pages 13ndash22 in HoganDL Tschaplinski PJ Chatwin S eds Carnation Creek and Queen Char-lotte Islands FishForestry Workshop Applying 20 Years of Coastal Re-search to Management SolutionsVictoria (Canada) Ministry of ForestsResearch Program Land Management Handbook no 41

Cooper DJ Andersen DC Chimner RA 2003 Multiple pathways for woodyplant establishment on floodplains at local to regional scales Journal ofEcology 91 182ndash196

Edwards RT 1998 The hyporheic zone Pages 399ndash429 in Naiman R BilbyRE eds River Ecology and Management Lessons from the PacificCoastal Ecosystem New York Springer

Everest FH Meehan WR 1981 Forest management and anadromous fishhabitat productivity Pages 521ndash530 in Transactions of the 46th NorthAmerican Wildlife and Natural Resources Conference Washington(DC) Wildlife Management Institute

Fausch KD Torgersen CE Baxter CV Li HW 2002 Landscapes to riverscapesBridging the gap between research and conservation of stream fishes Bio-Science 52 483ndash498

Fisher SG 1997 Creativity idea generation and the functional morphologyof streams Journal of the North American Benthological Society 16305ndash318

Frissell CA Liss WJ Warren WJ Hurley MD 1986 A hierarchical frameworkfor stream habitat classification Viewing streams in a watershed contextEnvironmental Management 10 199ndash214

Gomi T Sidle RC Richardson JS 2002 Understanding processes and down-stream linkages of headwater systems BioScience 52 905ndash916

Grant G 1997 A geomorphic basis for interpreting the hydrologic behav-ior of large river basins Pages 105ndash116 in Laenen A Dunnette DA edsRiver Quality Dynamics and Restoration Boca Raton (FL) CRC Press

Grant GE Swanson FJ 1995 Morphology and processes of valley floors inmountain streams Western Cascades Oregon Pages 83ndash101 in Costa JEMiller AJ Potter KW Wilcock PR eds Natural and Anthropogenic In-fluences in Fluvial Geomorphology The Wolman Volume Washington(DC) American Geophysical Union Geophysical Monograph no 89

Gregory SV Swanson FJ McKee WA Cummins KW 1991An ecosystem per-spective of riparian zones BioScience 41 540ndash551

Grimm MM Wohl EE Jarrett RD 1995 Coarse-sediment deposition asevidence of an elevation limit for flash flooding Bear Creek ColoradoGeomorphology 14 199ndash210

Hack JT 1957 Studies of Longitudinal Stream Profiles in Virginia andMaryland Washington (DC) US Government Printing Office US Geological Survey Professional Paper no 294-B

Hogan DL Bird SA Hassan MA 1998 Spatial and temporal evolution of smallcoastal gravel-bed streams The influence of forest management onchannel morphology and fish habitats Pages 365ndash392 in Klingeman

PC Beschta RL Komar PD Bradley JB eds Gravel-Bed Rivers in the Environment Highlands Ranch (CO) Water Resources Publications

Horton RE 1945 Hydrophysical approach to the morphology of hillslopesand drainage basins Geological Society of America Bulletin 56 275ndash370

Huston MA 1994 Biological Diversity The Coexistence of Species onChanging Landscapes Cambridge (United Kingdom) Cambridge Uni-versity Press

Jacobson RB 1995 Spatial controls on patterns of land-use induced streamdisturbance at the drainage-basin scalemdashan example from gravel-bedstreams of the Ozark Plateaus Missouri Pages 219ndash239 in Costa JE MillerAJ Potter KW Wilcock PR eds Natural and Anthropogenic Influencesin Fluvial Geomorphology The Wolman Volume Washington (DC)American Geophysical Union Geophysical Monograph no 89

Kalliola R Puharta M 1988 River dynamics and vegetation mosaicism Acase study of the River Kamajohka northernmost Finland Journal of Bio-geography 15 703ndash719

Kiffney PM Richardson JS 2001 Interactions among nutrient periphytonand invertebrate and vertebrate grazers in experimental channels Copeia2001 422ndash429

Kupferberg SJ 1996 Hydrologic and geomorphic factors affecting conser-vation of a river-breeding frog (Rana boylii) Ecological Applications 61332ndash1344

Leopold LB Wolman MG Miller JP 1964 Fluvial Processes in Geo-morphology San Francisco W H Freeman

Melis TS Webb RH Griffiths PG Wise TW 1995 Magnitude and FrequencyData for Historic Debris Flows in Grand Canyon National Park and Vicin-ity Arizona Available from US Geological Survey Information Ser-vices Box 25286 Federal Center Denver CO 80225 Water-ResourcesInvestigations Report no 94-4214

Meyer GA Leidecker ME 1999 Fluvial terraces along the Middle ForkSalmon River Idaho and their relation to glaciation landslide dams andincision rates A preliminary analysis and river-mile guide Pages 219ndash235in Hughes SS Thackray GD eds Guidebook to the Geology of EasternIdaho Pocatello (ID) Idaho Museum of Natural History

Meyer GA Pierce JL 2003 Climatic controls on fire-induced sedimentpulses in Yellowstone National Park and central Idaho A long-termperspective Forest Ecology and Management 178 89ndash104

Meyer GA Pierce JL Wood SH Jull AJT 2001 Fires storms and sedimentyield in the Idaho batholith Hydrological Processes 15 3025ndash3038

Miller DJ Luce CH Benda LE 2003 Time space and episodicity of physi-cal disturbance in streams Forest Ecology and Management 178 121ndash140

Minshall WG Cummins KW Peterson BJ Cushing CE Bruns DA Sedell JRVannote RL 1985 Developments in stream ecosystem theory CanadianJournal of Fisheries and Aquatic Sciences 42 1045ndash1055

Montgomery DR 1999 Process domains and the river continuum Journalof the American Water Resources Association 35 397ndash410

Mosley MP 1976 An experimental study of channel confluences Journal ofGeology 84 535ndash562

Naiman RJ DeCamps H Pastor J Johnston CA 1988 The potential im-portance of boundaries to fluvial ecosystems Journal of the NorthAmerican Benthological Society 7 289ndash306

Nakamura F Swanson FJWondzell SM 2000 Disturbance regimes of streamand riparian systemsmdasha disturbance-cascade perspective HydrologicalProcesses 14 2849ndash2860

OrsquoConnor JE Jones MA Haluska TL 2003 Floodplain and channel dynamicsof the Quinault and Queets Rivers Washington USA Geomorphology51 31ndash59

Osborne LL Wiley MJ 1992 Influence of tributary spatial position on thestructure of warm water fish communities Canadian Journal of Fisheriesand Aquatic Sciences 49 671ndash681

Perry JA Schaeffer DJ 1987 The longitudinal distribution of riverine ben-thos A river discontinuum Hydrobiologia 148 257ndash268

Poff NL 1997 Landscape filters and species traits Towards mechanisticunderstanding and prediction in stream ecology Journal of the NorthAmerican Benthological Society 16 391ndash409

Poff NL Allan JD Bain MB Karr JR Prestegaard KL Richter BD Sparks REStromberg JC 1997 The natural flow regime A paradigm for river con-servation and restoration BioScience 47 769ndash784

426 BioScience bull May 2004 Vol 54 No 5

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles

Pollock MM Naiman RJ Hanley TA 1998 Predicting plant species richnessin forested and emergent wetlandsmdasha test of biodiversity theory Ecol-ogy 79 94ndash105

Poole GC 2002 Fluvial landscape ecology Addressing uniqueness within theriver discontinuum Freshwater Biology 47 641ndash660

Reeves GH Benda LE Burnett KM Bisson PA Sedell JR 1995A disturbance-based ecosystem approach to maintaining and restoring freshwaterhabitats of evolutionarily significant units of anadromous salmonids inthe Pacific Northwest American Fisheries Society Symposium 17334ndash349

Resh VH Brown AV Covich AP Gurtz ME Li HW Minshall GW Reice SRSheldon AL Wallace JB Wissmar R 1988 The role of disturbance instream ecology Journal of the North American Benthological Society 7433ndash455

Rhoads BL 1987 Changes in stream characteristics at tributary junctionsPhysical Geography 8 346ndash361

Rice SP Greenwood MT Joyce CB 2001 Tributaries sediment sources andthe longitudinal organization of macroinvertebrate fauna along river sys-tems Canadian Journal of Fisheries and Aquatic Sciences 58 828ndash840

Scarnecchia DL Roper BB 2000 Large-scale differential summer habitat useof three anadromous salmonids in a large river basin in Oregon USAFisheries Management and Ecology 7 197ndash209

Schlosser IJ 1991 Stream fish ecology A landscape perspective BioScience41 704ndash712

Sedell JR Dahm CN 1984 Catastrophic disturbances to stream ecosys-tems Volcanism and clear-cut logging Pages 531ndash539 in Klug MJ ReddyCA eds Current Perspectives in Microbial Ecology Proceedings of theThird International Symposium on Microbial Ecology Michigan StateUniversity 7ndash12 August 1983 Washington (DC) American Society forMicrobiology

Small RJ 1973 Braiding terraces in the Val drsquoHerens Switzerland Geogra-phy 58 129ndash135

Swales S Levings CD 1989 Role of off-channel ponds in the life cycle of coho

salmon (Oncorhynchus kisutch) and other juvenile salmonids in the

Coldwater River British Columbia Canadian Journal of Fisheries and

Aquatic Sciences 46 232ndash242

Swanson FJ Fredriksen RL McCorison FM 1982 Material transfer in a west-

ern Oregon forested watershed Pages 233ndash266 in Edmonds RL ed

Analysis of Coniferous Forested Ecosystems in the Western United

States Stroudsburg (PA) Dowden Hutchinson and Ross

Swanson FJ Kratz TK Caine N Woodmansee RG 1988 Landform effects

on ecosystem patterns and processes BioScience 38 92ndash98

Townsend CR 1989 The patch dynamics concept of stream community ecol-

ogy Journal of the North American Benthological Society 8 36ndash50

mdashmdashmdash 1996 Concepts in river ecology Pattern and process in the catch-

ment hierarchy Archiv fuumlr Hydrobiologie 113 (suppl) 3ndash21

Vannote RL Minshall GW Cummins KW Sedell JR Cushing CE 1980 The

river continuum concept Canadian Journal of Fisheries and Aquatic Sci-

ences 37 130ndash137

Ward JV Tockner K Arscott DB Claret C 2002 Riverine landscape diver-

sity Freshwater Biology 47 517ndash539

Wiens JA 2002 Riverine landscapes Taking landscape ecology into the

water Freshwater Biology 47 501ndash515

Wipfli MS Gregovich DP 2002 Export of invertebrates and detritus from

fishless headwater streams in southeastern Alaska Implications for

downstream salmonid production Freshwater Biology 47 957ndash969

Wohl EE Pearthree PP 1991 Debris flows as geomorphic agents in the

Huachuca Mountains of southeastern Arizona Geomorphology 4

273ndash292

Wu JG Loucks OL 1995 From balance of nature to hierarchical patch dy-

namics A paradigm shift in ecology Quarterly Review of Biology 70

439ndash466

May 2004 Vol 54 No 5 bull BioScience 427

Articles