Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

download Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

of 17

Transcript of Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    1/17

    J. N. Am. Benthol. Soc., 2008, 27(2):304320 2008 by The North American Benthological SocietyDOI: 10.1899/07-096.1Published online: 25 March 2008

    Evidence for fish dispersal from spatial analysisof stream network topology

    Nathaniel P. Hitt1

    Department of Fisheries and Wildlife Sciences, Virginia Polytechnic Institute and State University,Blacksburg, Virginia 24061-0321 USA

    Paul L. Angermeier2

    Virginia Cooperative Fish and Wildlife Research Unit,3 US Geological Survey, Virginia Polytechnic Instituteand State University, Blacksburg, Virginia 24061-0321 USA

    Abstract. Developing spatially explicit conservation strategies for stream fishes requires an understandingof the spatial structure of dispersal within stream networks. We explored spatial patterns of stream fish

    dispersal by evaluating how the size and proximity of connected streams (i.e., stream network topology)explained variation in fish assemblage structure and how this relationship varied with local stream size. Weused data from the US Environmental Protection Agencys Environmental Monitoring and AssessmentProgram in wadeable streams of the Mid-Atlantic Highlands region (n 308 sites). We quantified streamnetwork topology with a continuous analysis based on the rate of downstream flow accumulation from sitesand with a discrete analysis based on the presence of mainstem river confluences (i.e., basin area .250 km2)within 20 fluvial km (fkm) from sites. Continuous variation in stream network topology was related to localspecies richness within a distance of;10 fkm, suggesting an influence of fish dispersal within this spatialgrain. This effect was explained largely by catostomid species, cyprinid species, and riverine species, butwas not explained by zoogeographic regions, ecoregions, sampling period, or spatial autocorrelation. Sitesnear mainstem river confluences supported greater species richness and abundance of catostomid, cyprinid,and ictalurid fishes than did sites .20 fkm from such confluences. Assemblages at sites on the smalleststreams were not related to stream network topology, consistent with the hypothesis that local stream sizeregulates the influence of regional dispersal. These results demonstrate that the size and proximity of

    connected streams influence the spatial distribution of fish and suggest that these influences can beincorporated into the designs of stream bioassessments and reserves to enhance management efficacy.

    Key words: stream network topology, fish assemblage structure, dispersal, conservation, bioassess-ment, Mantel test, Mid-Atlantic Highlands.

    Development of spatially explicit conservation strat-

    egies requires an understanding of the distances overwhich dispersal influences local population and com-

    munity structure (Poiani et al. 2000). However, analyses

    of stream fish dispersal have been limited by the

    complex spatial structure of stream networks and thelow probability of detecting long-distance movements

    via markrecapture studies. Herein, we explore spatial

    patterns of fish dispersal by evaluating how the size and

    proximity of connected streams (i.e., stream network

    topology) explain variation in fish assemblage structure

    and how this relationship varies with local stream size.

    Our analysis offers new insight into the spatial grain at

    which dispersal affects local fish assemblages.Stream network topology provides a spatial frame-

    work to evaluate lotic dispersal. Whereas the River

    Continuum Concept (RCC; Vannote et al. 1980)

    emphasizes regulation of local conditions by stream

    size, a stream network perspective recognizes that

    connected streams might influence the spatial distribu-

    tion of remote resources, refugia, or immigrant source

    populations (Fausch et al. 2002, Lowe et al. 2006,

    1 E-mail addresses: [email protected] [email protected] The Unit is jointly sponsored by the US Geological

    Survey, Virginia Polytechnic Institute and State University,Virginia Department of Game and Inland Fisheries, andWildlife Management Institute.

    304

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    2/17

    Campbell Grant et al. 2007). For example, a networkperspective might posit that the biotic community in astream flowing into a mainstem river would be distinctfrom the community in a similarly sized stream thatlacks a nearby river confluence. Furthermore, differ-ences between these 2 communities would stem from

    movement by species adapted to live in either stream orriver habitats. In contrast, the RCC predicts that thecommunities in these streams would be similar becauseof their similar size and that species composition wouldlargely reflect adaptations to local conditions.

    Studies of stream fish movements and distributionssuggest that stream network topology has importantconsequences for fish dispersal. For example, non-diadromous stream fishes commonly move longdistances (i.e., .1 fluvial km [fkm]; Gatz and Adams1994, Albanese et al. 2004, Roghair and Dolloff 2005,Gresswell and Hendricks 2007). Across North Amer-ica, fish species richness tends to be greater in rivers

    located within stream networks than in similarly sizedrivers that flow directly into the ocean (Sheldon 1988),suggesting that connected streams provide importantimmigrant source populations. Analogous patternsoccur within basins; streams near mainstem riverstend to support more species than similarly sizedstreams that lack riverine connections (Gorman 1986,Osborne and Wiley 1992, Schaefer and Kerfoot 2004,Smith and Kraft 2005, Hitt and Angermeier 2006).

    Interstream dispersal, which is strongly influencedby connectivity, plays an important role in structuringfish communities. Matthews and Robison (1998) foundthat stream fish assemblages of the Ouachita Moun-

    tains were more similar among sites separated byfewer confluences (i.e., shorter fluvial distance).Angermeier and Winston (1998) demonstrated thatlocal stream fish species richness was better predicted

    by fish diversity within relatively small regions (i.e.,intersections of physiographic regions and river

    basins) than by fish diversity in larger regions (i.e.,river basins), suggesting that interstream connectivityand dispersal affect local assemblage structure. Studiesdemonstrating spatial autocorrelation in fish assem-

    blage structure (Wilkinson and Edds 2001, Hitt et al.2003, Grenouillet et al. 2004) also provide evidence for

    fish dispersal among connected streams.Recognized longitudinal patterns of fish assemblagestructure provide 3 testable predictions about fishdispersal within stream networks. First, streamsflowing directly into rivers (i.e., mainstem tributaries)will support more species than streams that lackriverine connections (i.e., headwater tributaries) be-cause fish species richness typically increases withstream volume (Shelford 1911, Burton and Odum 1945,Sheldon 1968). Second, the effects of dispersal from

    rivers will be weakest in the smallest streams becauseextreme environmental conditions might excludemany potential immigrants (Schlosser 1990). Third,the effects of dispersal from rivers will be more evidentin percid, catostomid, cyprinid, ictalurid, and cen-trarchid fishes, which commonly inhabit rivers, than in

    salmonid or cottid fishes, which rarely inhabit rivers inthe study area (Jenkins and Burkhead 1994).To test our predictions, we developed a spatial

    model of stream network topology based on the rate offlow accumulation downstream from sampled sites.We used this model to characterize the size andproximity of connected streams and to distinguish

    between sites that flow into rivers and streams (i.e.,rapid and slow accumulation of volume per unitdistance, respectively). First, we quantified a surrogateof flow volume at 1-fkm intervals downstream fromsites to a total distance of 20 fkm. Second, we classifiedsites as mainstem tributaries or headwater tributaries

    based on the presence or absence of mainstem riverconfluences, respectively, within 20 fkm. This analysispermitted us to contrast sites based on the availabilityof riverine habitats and potential source populations.

    Our analysis does not measure dispersal empirically but infers dispersal patterns by evaluating how fishassemblages vary among spatial positions withinstream networks. Key limitations of this approach arethat it cannot characterize fish assemblage structure inthe potential source populations (e.g., connected rivers)or identify the timeframe of dispersal. Instead, ourspatial models quantify heterogeneity in the streamnetwork to examine expected differences between

    smallsmall and smalllarge stream interactions.An advantage of our approach is that it permits

    analysis among zoogeographic and physiographicregions over a large geographic extent. In fact, ourstudy provides one of the most spatially extensiveanalyses of stream fish dispersal to date. Such analysesare necessary to test concepts of landscape ecology(Wiens 2001) and to design conservation plans andactions that are effective in freshwater landscapes(Higgins 2003).

    Methods

    Data source

    We used data from the US Environmental ProtectionAgencys (USEPA) Environmental Monitoring andAssessment Program (EMAP) in the Mid-AtlanticHighlands region, USA (n 308 sites; Fig. 1). Thisregion encompasses 205,000 km2 in the mountains ofNew York, Pennsylvania, Maryland, West Virginia, andVirginia. EMAP sites were located using a systematicrandom method (Herlihy et al. 2000) across several

    2008] 305FISH DISPERSAL IN STREAM NETWORKS

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    3/17

    physiographic regions (Blue Ridge, Ridge and Valley,Central Appalachians, and Allegheny Plateau; Omernik1987) and zoogeographic regions (Atlantic Slope basins,Ohio River basin, New River basin [of the Ohio River

    basin], and Tennessee River basin) (Fig. 2; Table 1).In each site, stream fishes were sampled using

    standardized single-pass backpack electrofishingmethods (McCormick and Hughes 1998). Sample reachlengths were 403 the average stream width for eachsite (USEPA 2000). Fishes were identified in the fieldand released, but some problematic specimens werepreserved and identified later by USEPA personnel.We made 78 changes to the EMAP data based onpublished species accounts, under the assumption thatEMAP surveys did not detect new interbasin rangeexpansions (see appendix B, Hitt 2007). Raw data areavailable from http://www.epa.gov/emap.

    We used several criteria to select EMAP data for ouranalysis. First, we restricted sites to wadeable streams(i.e., backpack electrofishing methods) because oflower sampling efficiency for boat electrofishing

    (Cyterski and Barber 2006). Second, we removed sitesthat flow into reservoirs within 20 fkm to removepotential effects of downstream dams and reservoirson upstream assemblages (Winston et al. 1991,Guenther and Spacie 2006). When sites were sampledrepeatedly, we used the sample with the greatestspecies richness or, in the case of ties, the more recentsample. Some EMAP sites included extensive data onphysical habitat and water quality, but many sites withfish data did not. To maximize the use of existing fishdata, we included these sites in our analysis anddeveloped a stream size surrogate of local habitatconditions (see Stream network topology below).

    We calculated taxonomic and habitat-size metrics tocharacterize fish assemblage structure (Table 2). Taxo-nomic metrics included family-level species richnessand abundance. The rarest families were excluded fromthis analysis, and cottid species richness was notcalculated because most sites lacked multiple cottidspecies. Habitat-size metrics distinguished betweenriver and creek specialists. Assignments of river and

    FIG. 1. US Environmental Protection Agency Environmental Monitoring and Assessment Program study site locations withinthe Mid-Atlantic Highlands study area (n 308).

    306 [Volume 27N. P. HITT AND P. L. ANGERMEIER

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    4/17

    creek specialists excluded species reported to inhabit allstream sizes. However, it was necessary to includespecies with intermediate stream-size associations (i.e.,stream species in Jenkins and Burkhead 1994) inassignments of river and creek specialists. For example,

    our classification of river specialists includes speciescategorized by Jenkins and Burkhead (1994) as riverspecies as well as river and stream species. Of 130species in our data set, we classified 41 (32%) as riverspecialists and 27 (21%) creek specialists (Appendix).

    FIG. 2. Zoogeographic regions and ecoregions within the study area. A.Zoogeographic regions include the Ohio River basin(OHI), Atlantic Slope basins (ATL), the New River basin (NEW), and the Tennessee River basin (TEN). B.Ecoregions include theAppalachian Plateau (APPL), the Central Appalachians (CEAP), the Ridge and Valley (RIVA), and the Blue Ridge (BLRI).

    2008] 307FISH DISPERSAL IN STREAM NETWORKS

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    5/17

    Stream network topology

    Stream network topology refers to the size andproximity of connected streams (Benda et al. 2004b)and is a property of the stream system scale ofFrissell et al. (1986). Therefore, properties of streamnetwork topology are not physical features of individ-ual streams but are emergent properties of multipleconnected streams. The first studies of stream networktopology addressed geomorphic processes of erosionand basin evolution (e.g., Shreve 1966), whereas recentapplications have explored animal dispersal andrecolonization dynamics (Fagan 2002, Fausch et al.2002, Lowe et al. 2006).

    We characterized stream networks based on the rateof downstream flow accumulation within 20 fkm. Wechose this grain size for analysis because previouswork has demonstrated significant effects of connectedrivers within this distance (Osborne and Wiley 1992).The large spatial extent of the study area and limited

    computing capacity precluded use of digital elevationmodels to derive flow networks. Consequently, weused the relationship between basin area and streamlength (Hack 1957) to develop a surrogate of streamvolume. We used this modeling approach instead ofstream order to capture continuous variation in streamnetworks (Hughes and Omernik 1981, Matthews 1986,Fausch et al. 2002).

    We calculated upstream cell counts (UCCs) fromstream network raster data as a surrogate of streamvolume (C. D. Heatwole, Virginia Polytechnic andState University, personal communication). First, wedownloaded National Hydrological Data medium-resolution data (1:100,000 scale) and converted streampaths from vector to raster data (30-m2 cells). Raw dataare available at http://nhd.usgs.gov. Second, we usedthe flow network raster to calculate distances fromeach cell to the outlet pour-point of each basin. Third,we converted distances to UCCs and then combinedinflows from confluent basins. We then sampled theUCC raster at EMAP sites and at every kilometer for20 fkm downstream. We did all calculations in

    ARCGIS (version 9.1; Environmental Systems Re-search Institute, Redlands, California).

    We compared the UCC data against independentmeasures of upstream basin area to validate UCCs as asurrogate for stream flow volume. We used simplelinear regression to relate UCCs to upstream basinareas from EMAP calculations in a subset of sites

    where physical habitat data were available (n 198).Log-log plots revealed bivariate linearity and good fit(R2 0.919; Fig. 3); thus, UCC data provide areasonable surrogate of stream flow volume (Hack1957). Based on this relationship, we used a UCC valueof 5000 to indicate the presence a mainstem riverconfluence (i.e., upstream basin area .250 km2;Osborne and Wiley 1992).

    Statistical analysis

    We used 3 approaches to explore relationshipsbetween stream network topology and fish assemblage

    TABLE 1. Numbers of US Environmental Protection Agency Environmental Monitoring and Assessment Program sites withinzoogeographic regions and ecoregions in the Mid-Atlantic Highlands study area. Regions are mapped in Fig. 2.

    Ecoregion

    Zoogeographic region

    TotalsAtlantic Slope basins New River basin Ohio River basin Tennessee River basin

    Appalachian Plateau 19 0 36 0 55

    Blue Ridge 12 6 0 0 18Central Appalachians 2 4 25 1 32Ridge and Valley 150 24 8 21 203Totals 183 34 69 22 308

    TABLE 2. Stream fish metrics, variable codes, and trans-formations used in our analysis of the Mid-AtlanticHighlands region, USA.

    Metric Code Transformation

    Species richness S NoneAbundance N log10(x 1)Cottid abundance COT_N log10(x 1)Salmonid richness SAL_S NoneSalmonid abundance SAL_N log10(x 1)Percid richness PER_S NonePercid abundance PER_N log10(x 1)Centrarchid richness CEN_S None

    Centrarchid abundance CEN_N log10(x 1)Catostomid richness CAT_S NoneCatostomid abundance CAT_N log10(x 1)Cyprinid richness CYP_S NoneCyprinid abundance CYP_N log10(x 1)Ictalurid richness ICT_S NoneIctalurid abundance ICT_N log10(x 1)Creek species richness CRK_S NoneCreek species abundance CRK_N log10(x 1)River species richness RIV_S NoneRiver species abundance RIV_N log10(x 1)

    308 [Volume 27N. P. HITT AND P. L. ANGERMEIER

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    6/17

    structure. First, we evaluated how continuous variationin stream networks explained local fish assemblagemetrics and how these relationships changed withincreasing downstream distances from sample sites(i.e.,increasing analysis grain size). Second, we tested theprediction that the smallest streams would contributeleast to fish assemblagestream network relationships.In each case, the presence of significant relationships

    between stream network topology (i.e., UCC variation)

    and fish assemblage structure would be consistent withthe hypothesis of regional dispersal. Third, we tested foreffects of mainstem river confluences by comparing fishmetrics between mainstem tributaries and headwatertributaries (Osborne and Wiley 1992).

    We used simple and partial Mantel tests to assess therelationships between fish metrics and stream networkstructure (Mantel 1967). Mantel tests are distance-basedmatrix correlations that use permutation procedures tocalculate the probability that observed correlations arerandom. Simple Mantel tests evaluate bivariate rela-tions, whereas partial Mantel tests incorporate blockingfactors to control for potentially confounding variables.These tests are useful for assessing correspondence invariables from unknown statistical distributions andfor quantifying the spatial structure of ecological data(e.g., King et al. 2003, 2005). In this analysis, wecalculated Euclidean dissimilarity matrices from UCCdata arrays to characterize stream network structurealong stream pathways (Fig. 4).

    We used the R programming language with theECODIST library (Goslee and Urban 2007) to calculate

    distance matrices and run Mantel tests. First, we usedsimple Mantel tests to assess potentially confoundingeffects of ecoregion, zoogeographic region, samplingmonth, sampling year, and spatial autocorrelation onfish metrics. To quantify spatial autocorrelation, wecalculated Cartesian distances among sites from

    Universal Transverse Mercator data. Second, we usedpartial Mantel tests to evaluate UCC relations to fishmetrics at increasing analysis grain sizes whilecontrolling for significant variables identified fromsimple Mantel tests. We used Euclidean distances toquantify differences among sites for all variables. Toreduce Type I error probabilities, we chose a high alevel (a 0.10) for inclusion of blocking factors inpartial Mantel tests and a low a level (Bonferroni-corrected a0 0.05/21 tests 0.0024) to assesssignificance of stream network effects. All Mantel testsused 10,000 resampling iterations. We did not usesemivariograms because the long lag distances be-

    tween sites prohibited this type of analysis.We then evaluated the role of local stream size in

    regulating regional dispersal, as inferred from patternsof fish distribution and abundance. We evaluated 5fish metrics that demonstrated significant relationshipswith stream network structure (a0 0.0024 at distances,6 fkm). First, we classified stream sites into 1 of 3sizes (Table 3). Second, we used simple Mantel tests toevaluate potentially confounding effects of ecoregion,zoogeographic region, sampling month, samplingyear, and spatial autocorrelation on fish metrics. Third,we conducted partial Mantel tests to evaluate rela-tionships between stream network structure (i.e., UCC

    variation) and fish metrics across site size classes whilecontrolling for potentially confounding variables.

    Last, we evaluated effects of downstream rivers onlocal fish assemblage structure. Following Osborneand Wiley (1992), we categorized sites based on thepresence or absence of mainstem river confluenceswithin 20 fkm downstream from sites (i.e., UCC .5000; see Fig. 3) as mainstem tributaries and headwa-ter tributaries, respectively. We used MannWhitneytests to compare fish metrics between headwater andmainstem tributaries. We chose nonparametric meth-ods because exploratory analyses revealed significantnonnormality in fish metrics data and transformationsfailed to normalize the data. We identified 248mainstem tributary sites and 60 headwater tributarysites in the study area.

    Results

    Spatial and temporal variation in fish assemblage structure

    The 101,457 individual fish documented in the Mid-Atlantic Highlands data set included 17 families, 53

    FIG. 3. Simple linear regression of upstream basin areas(Environmental Monitoring and Assessment Program

    [EMAP] data) and calculated upstream cell counts (UCCs)for a subset of EMAP sites in the Mid-Atlantic Highlandsstudy area. n 198.

    2008] 309FISH DISPERSAL IN STREAM NETWORKS

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    7/17

    genera, and 130 species. Cyprinids contributed 65% of

    the total abundance. Cottid, catostomid, centrarchid,

    percid, and salmonid fishes contributed 1% to 12%,

    and other families (Lepisosteidae, Umbridae, Clupei-

    dae, Atherinidae, Percopsidae, Sciaenidae, Esocidae,

    Petromyzontidae, Anguillidae, Fundulidae, and Icta-

    luridae) contributed ,1% of the total abundance.

    Total species richness was more strongly related to

    zoogeographic region than to ecoregion, whereas total

    abundance showed the opposite pattern (Table 4).

    Zoogeographic regions and ecoregions showed stron-

    ger relationships to percid species richness and

    abundance than to other families. Creek species

    richness and river species richness were related to

    zoogeographic region and ecoregion, but river species

    showed marginally stronger relationships to regional

    classifications than did creek species (Table 4).

    Variation in the month and year of samples was

    related to taxonomic and functional variation in fish

    assemblage structure (Table 4). Variation among sample

    years (19931998; n 104 [1993], 100 [1994], 2 [1995],4 [1996], 61 [1997], and 37 [1998]) was related to species

    richness and total abundance, centrarchid species

    richness, and percid species richness and abundance

    (Table 4). River species richness and abundance showed

    significant interannual variation, whereas creek species

    richness and abundance did not (Table 4). For example,

    samples taken in 1997 and 1998 tended to contain more

    river species than did samples from 1993 and 1994

    (means 1.4 and 0.9, respectively), but creek species

    FIG. 4. Conceptual diagram of stream network topology model. Euclidean dissimilarity matrices were used to characterizevariation in upstream cell counts (UCCs) at increasing spatial grains along fluvial pathways. The UCC data are hypothetical in thisexample.

    TABLE 3. Assignments of stream size classes. Approximate basin areas are based on the upstream cell count (UCC) relationshipto upstream basin area in Fig. 3. Stream cross-sectional areas were calculated as the product of mean stream width and meanthalweg depth for the subset of Environmental Monitoring and Assessment Program sites with physical habitat data (n198). UCCand upstream basin values are ranges; stream cross-sectional areas are means (SE).

    Site size class n UCC Upstream basin area (ha) Stream cross-sectional area (m2)

    1 102 1192 11000 0.57 (0.11)2 103 193777 10005000 1.88 (0.18)3 103 7904780 500010,000 4.50 (0.35)

    310 [Volume 27N. P. HITT AND P. L. ANGERMEIER

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    8/17

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    9/17

    richness (Fig. 6B), and cyprinid richness (Fig. 6D). Bothmidsized and large sites (i.e., basin areas .1000 ha;Table 3) showed significant stream network relations tocatostomid abundances (Fig. 6C) and river speciesrichness (Fig. 6E). Overall, river species richnessshowed the strongest relationship to stream network

    topology (partial Mantel r . 0.30; Fig. 6E) and showedthe most significant Mantel r values at the largestspatial grain of the analysis (i.e., 20 fkm; Fig. 6E).

    Analysis of stream network effects sorted by streamsize revealed spatial patterns distinct from analysesincluding all sites. When evaluated in stream size

    FIG. 5. Relationships between fish assemblage metrics and stream network topology. A.Total richness and abundance. B.Catostomidae. C.Cyprinidae. D.Centrarchidae. E.Cottidae. F.Ictaluridae. G.Percidae. H.Salmonidae. I.River species.

    J.Creek species. Solid lines indicate species richness and dashed lines indicate abundances. Partial Mantel correlations above thehorizontal dotted line indicate significance at a Bonferroni-corrected error rate, a00.05/21 (21 distance classes)0.0024. Creek andriver species are defined in the Appendix. fkm fluvial kilometer.

    312 [Volume 27N. P. HITT AND P. L. ANGERMEIER

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    10/17

    classes, correlations between fish metrics and networktopology exhibited peak-shaped patterns in totalspecies richness (Fig. 6A), catostomid richness andabundance (Fig. 6B, C), cyprinid richness (Fig. 6D), andriver species richness (Fig. 6E). Midsized streamsshowed peak-shaped patterns for all evaluated metrics.Large-sized streams showed similar peak-shaped pat-terns, but correlations were significant (a00.0024) onlyfor catostomid abundance (Fig. 6C), cyprinid richness(Fig. 6D), andriver species richness (Fig. 6E). In contrast,all metrics except river species richness showeddiminishing effects of stream networks with increasingdownstream distances when analyzed across all sites(i.e., decreasing partial Mantel r coefficients; Fig. 5).

    The presence of mainstem river confluences (i.e.,basin areas.250 km2) had important consequences forlocal fish assemblage structure. Mainstem tributariessupported marginally greater species richness andabundance of catostomid, cyprinid, and ictaluridfishes than did headwater tributaries (Table 6). Totalabundance was also greater in mainstem tributariesthan in headwater tributaries (Table 6). Total speciesrichness tended to be greater in mainstem than inheadwater tributaries, but this difference was notsignificant (MannWhitney, a 0.10). River species

    were more abundant in mainstem tributaries than inheadwater tributaries (Table 6). In contrast, connectiv-ity to rivers did not affect the species richness orabundance of creek species (Table 6).

    Discussion

    We tested the hypotheses that the size and proximityof connected streams influences local fish assemblagestructure through dispersal processes and that local

    stream size regulates the effects of regional dispersal.We predicted that: 1) the smallest streams would beleast likely to exhibit effects of dispersal from rivers

    because of locally unsuitable habitats, and 2) fishfamilies containing many riverine species would showstronger effects of dispersal than families with fewriver species. Overall, our results support bothpredictions and suggest that dispersal among connect-ed streams influences fish assemblage structure andthat these effects vary among fish taxa, local streamsizes, and distances to potential source populationsand remote resources.

    Stream network effects vary among taxa

    Fish families showed distinct relationships to streamnetwork topology. Catostomid and cyprinid fishes weresignificantly related to stream network topology, where-as percid, ictalurid, cottid, and salmonid fishes were not(Fig. 5BH). Moreover, catostomid and cyprinid fishesexhibited greater species richness and abundance inmainstem tributaries than in headwater tributaries(Table 6). The concordance between continuous anddiscrete analyses suggests that river connectivity might

    be more important for local persistence of catostomidsand cyprinids than for other species, even though thesefamilies contributed a minority of the designated riverspecialists (i.e., 17/4142% of river species; Appendix).

    Many catostomids exhibit seasonal spawning migra-tions that extend for tens of kilometers (Scott andCrossman 1973, Jenkins and Burkhead 1994). Therefore,conservation of imperiled catostomids might dependon maintaining stream network connectivity for dis-persal (Fagan et al. 2002, Cooke et al. 2005). Our resultssuggest that streamriver connectivity might influence

    TABLE 5. Relationships of fish metrics, by stream size class, to spatial and temporal variation in the Mid-Atlantic Highlandsstudy area. Cell values are Mantel r correlation coefficients. SpaceCartesian distances among sites. NSnot significant ( p. 0.10),* p , 0.10, ** p , 0.01, *** p , 0.005. Variable codes are given in Table 2. Stream size information is given in Table 3.

    Metric Size class Zoogeographic region Ecoregion Sample year Sample month Space

    S 1 NS 0.081* NS NS NS2 NS NS NS NS NS3 NS NS NS NS NS

    CAT_S 1 NS NS NS NS NS2 0.123* 0.131* 0.097* NS NS3 NS NS NS NS NS

    CAT_N 1 0.079* NS NS NS NS2 0.142*** NS NS NS 0.040*

    3 NS NS NS 0.089** NSCYP_S 1 NS NS NS NS NS

    2 NS NS NS NS NS3 NS NS NS NS NS

    RIV_S 1 NS NS NS NS NS2 NS NS NS NS 0.061*

    3 0.156** 0.088* 0.059* NS NS

    2008] 313FISH DISPERSAL IN STREAM NETWORKS

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    11/17

    catostomid dispersal and that dispersal distances of 6 to8 fkm might be most important for structuring localcatostomid richness and abundance (see also Osborneand Wiley 1992). However, we probably underestimat-ed dispersal distances for some catostomid riverspecialists found outside our study area (e.g., razorback

    sucker [Xyrauchen texanus]; Tyus and Karp 1990).Moreover, Curry and Spacie (1984) concluded that themost abundant catostomids in our analysis (i.e., whitesucker [Catostomus commersoni] and northern hogsucker [ Hypentelium nigricans]) were less mobile thanwere other sucker species (e.g., Moxostoma spp.).

    Our analysis revealed important effects of streamnetwork topology for cyprinid species richness.Matthews and Robison (1998) showed that dispersalamong connected streams influences local distributionof cyprinids. Our analysis supports this notion andsuggests an important role of streamriver connectivityfor cyprinid dispersal. Several cyprinid species exhibit

    seasonal migrations from rivers into connected streams(Gorman 1986). In our analysis, cyprinid speciesrichness and abundance were somewhat greater inmainstem tributaries than in headwater tributaries(Table 6), even though only 14 of the 50 cyprinidspecies were classified as river specialists (Appendix).

    Our results supported the prediction that salmonidand cottid species would show weak effects of streamnetwork topology because of their low species richnessand abundance in rivers within the study area.However, interstream dispersal by salmonids might bemore prevalent where coldwater rivers (i.e., suitabledispersal corridors) connect streams. For example, bulltrout (Salvelinus confluentus) tend to occur in patchesacross the landscape, a distribution that suggests inter-stream colonization (Rieman and McIntyre 1995, Dun-ham and Rieman 1999). Native brook trout (Salvelinusfontinalis) move long distances (i.e.,.1 fkm) in Virginia(Roghair and Dolloff 2005), but we did not evaluatedispersal among connected streams of the same size,and therefore, probably missed most interstreamdispersal relevant to coldwater species in the study area.

    Stream network effects are mediated by local stream size

    Fish assemblage relationships to stream networktopology varied among local stream sizes. Stream

    FIG. 6. Variation among local stream sizes in fishassemblage relationships to stream network topology. A.Total species richness. B.Catostomid richness. C.Catos-tomid abundance. D.Cyprinid richness. E.River speciesrichness. Correlations above the horizontal dotted line are

    statistically significant at a Bonferroni-corrected error rate, a0

    0.05/21 (21 distance classes) 0.0024. Small stream sitesare shown with thin solid lines, medium-sized stream sitesare shown with dashed lines, and the largest stream sites areshown with thick solid lines. See Table 3 for stream sizeclassifications. fkm fluvial kilometer.

    314 [Volume 27N. P. HITT AND P. L. ANGERMEIER

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    12/17

    network effects were weakest in the smallest streams(i.e., basins ,1000 ha; Table 3) and greater in largerstreams (i.e., basins 100010,000 ha; Table 3, Fig. 6).Presumably, harsh conditions in small streams exclud-ed potential immigrants or caused rapid local extirpa-tions (Schlosser 1990), thereby minimizing the

    influence of dispersal from rivers on local assemblagestructure. Similar interactions between connectivityand habitat size regulate the spatial distribution ofChinook salmon (Oncorhynchus tshawytscha) spawningpatches in Idaho (Isaak et al. 2007). Our results supportthe hypothesis that local fish assemblages are regulat-ed in part by an interaction between site suitability andcolonist availability (Angermeier et al. 2002).

    The differences in fish assemblages we observed between small and large streams are consistent withpredictions of the RCC (Vannote et al. 1980). The RCCrecognizes that tributaries might cause discontinuitiesin the river continuum (Vannote et al. 1980), and such

    effects have been shown for spatial patterns of organicmatter (Bruns et al. 1984) and sediment transport(Benda et al. 2004a, Rice et al. 2006). However, ourresults suggest additional effects of stream networkson distribution, abundance, and persistence of fishpopulations. As such, accuracy of RCC predictions forfish assemblage structure might be improved byincorporating measures of the size and proximity ofconnected streams.

    Stream network effects vary with distance

    The absence of information about how streamnetwork effects vary with distance presents a criticalgap for understanding landscape influences on stream

    biota (Wang et al. 2006). Previous studies have adopteda priori distances to test for dispersal from riverine areas

    (e.g., Osborne et al. 1992, Osborne and Wiley 1992),and therefore, were unable to assess the effects ofdistance from potential source populations. By devel-oping a distance-based model of stream networktopology, we provide new insights about the spatialgrain at which dispersal affects local fish assemblages.

    We found that variation in stream network topologywas significantly related to local species richness up toa distance of ;10 fkm downstream from sites (Fig.5A). Similar distance effects were observed forcatostomid abundance (10 fkm; Fig. 5B) and cyprinidspecies richness (8 fkm; Fig. 5C). The longest distance

    effects of stream network topology were detected forriver species richness (.20 fkm; Fig. 5I). However, inmid- and large-sized streams (Table 3), stream networkeffects extended past 20 fkm for total species richness,catostomid species richness and abundance, cyprinidspecies richness, and river species richness (Fig. 6).

    Spatially explicit analysis of distance provided someinsights that were not evident from discrete analyses ofstream networks. For example, total species richnesswas related to stream network topology up to adistance of ;10 fkm (Fig. 5A). However, sitescategorized as mainstem tributaries and headwatertributaries were not significantly different in total

    species richness (a 0.10; Table 6). Although Osborneand Wiley (1992) detected greater differences in speciesrichness between mainstem and headwater tributariesthan we did, they also selected sample sites tomaximize mainstem influences, whereas we evaluatedsites that were located randomly with respect tomainstem confluences.

    Interpretation of dispersal distances from our studyrequires 3 caveats. First, we did not evaluate potentialdispersal among connected streams of the same size orfrom upstream sources. Instead, stream networktopology was quantified as variation in the rate of flowaccumulation downstream from sites. This approachallowed us to quantify continuous variation in streamnetworks (i.e., size and proximity of connected streams)

    but might not have detected signals of dispersal from allpotential source populations. Second, the significanceof distance relationships to stream network structurewas assessed using statistical thresholds, not ecologicalthresholds. We chose a conservative a-level to assessstatistical significance of stream network effects(Bonferroni-corrected a0 0.05/21 comparisons 0.0024),

    TABLE 6. MannWhitney tests for differences in fishmetrics between mainstem tributaries (MT) and headwatertributaries (HT) in the Mid-Atlantic Highlands, USA. See textfor methods used to assign sites to MT and HT classes. NSnot significant (p . 0.10).

    Variable v2 p Direction

    S 2.12 0.146 NSN 2.78 0.095 MT . HTCAT_S 2.90 0.088 MT . HTCAT_N 2.81 0.094 MT . HTCEN_S 0.00 0.988 NSCEN_N 0.03 0.871 NSCOT_N 0.45 0.504 NSCYP_S 3.90 0.048 MT . HTCYP_N 3.17 0.075 MT . HTICT_S 3.29 0.070 MT . HTICT_N 3.74 0.053 MT . HTPER_S 1.17 0.280 NSPER_N 0.77 0.381 NSSAL_S 0.38 0.537 NSSAL_N 0.48 0.490 NS

    CRK_S 0.02 0.896 NSCRK_N 0.02 0.882 NSRIV_S 2.31 0.129 NSRIV_N 2.85 0.091 MT . HT

    2008] 315FISH DISPERSAL IN STREAM NETWORKS

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    13/17

    but different criteria for significance would yielddifferent interpretations of distance effects. Third, ourstudy cannot distinguish between temporary move-ments to access resources (i.e., life-history expression)and interpopulation dispersal (i.e., metapopulationdynamics). We would expect movements for life-

    history expression to encompass smaller distances thanmovements for interpopulation dispersal, but we wereunable to test this notion.

    Implications for managing stream fishes

    The spatial dimensions of community structureremain poorly understood but have fundamentalimplications for managing stream fishes. Our resultssuggest that stream networks influence fish dispersaldynamics and that these effects are mediated by localstream size and the distance to connected streams.Therefore, an understanding of stream network topol-

    ogy might improve the use of fish data in assessingstream quality and in designing conservation areas.Bioassessment studies often treat the spatial grain of

    analysis implicitly by stratifying sites according tostream order (Strahler 1957). Such methods permitextrapolations to unsampled areas under the assump-tion that sample sites of a given stream order representother streams of the same order. In contrast, we foundthat fish dispersal might influence assemblages insimilarly sized streams differently, based on theproximity of sites to rivers. This finding implies that:1) pooling data between mainstem tributaries andheadwater tributaries might decrease the sensitivity of

    fish assemblage responses to environmental stressors,and 2) excluding or including river specialists fromassessment-metric calculations might differentiallyenable local or regional assessments of environmentalquality, respectively. To our knowledge, neither ofthese implications has been incorporated into existing

    bioassessment protocols, and their consequences forbioassessments remain unexamined.

    Recognizing the role of stream network topology infish assemblage organization also might improvefreshwater reserve designs. Reserve design theoryrecommends promoting connectivity for dispersaland recolonization (Groves et al. 2002, Abell et al.2007), but few studies have examined the spatial grainat which dispersal is important for persistence ofstream fishes. Our results suggest that: 1) conservationreserves for lotic fishes will be most effective whenprotecting habitat across !9 fkm; 2) conservation ofriverine specialists, cyprinids, and catostomids mightrequire even greater distances (i.e., .9 fkm); and 3)conservation targets in the smallest streams (i.e., basinareas ,1000 ha) might be less influenced by stream

    network connectivity than targets in larger streams.Additional research will be necessary to understandhow connectivity among similarly sized streamsinfluences fish dispersal within stream networks. Suchknowledge might help optimize the size, configura-tion, and siting of lotic reserves to facilitate conserva-

    tion of freshwater biodiversity.

    Acknowledgements

    We thank A. Dolloff, P. Flebbe, C. Heatwole, R.Voshell, C. Zipper, J. Roberts, E. Frimpong, and R. Betzfor assistance in developing this research and review-ing the manuscripts. We also thank 2 anonymousreferees for their assistance with our manuscript. Thiswork was supported by a Cunningham Fellowship(Virginia Polytechnic Institute and State University),the USEPA National Network for EnvironmentalManagement Studies, the USEPA Office of Water

    (Assessment and Watershed Protection ProgramGrants; X783256601), and the USEPA Science toAchieve Results program (RD-831368010).

    Literature Cited

    ABELL, R., J. D. ALLAN, AND B. LEHNER. 2007. Unlocking thepotential of protected areas for freshwaters. BiologicalConservation 134:4863.

    ALBANESE, B., P. L. ANGERMEIER, AND S. DORAI-RAJ. 2004.Ecological correlates of fish movement in a network ofVirginia streams. Canadian Journal of Fisheries andAquatic Sciences 61:857869.

    ANGERMEIER

    , P. L., K. L. KRUEGER

    ,AND

    C. A. DOLLOFF

    . 2002.Discontinuity in stream-fish distributions: implicationsfor assessing and predicting species occurrence. Pages519527 in J. B. Scott, P. J. Heglund, M. L. Morrison, J. B.Haufler, M. G. Raphael, W. A. Wall, and F. B. Samson(editors). Predicting species occurrences: issues of accu-racy and scale. Island Press, Covelo, California.

    ANGERMEIER, P. L., AND M. R. WINSTON. 1998. Local vs. regionalinfluence on local diversity in stream fish communities ofVirginia. Ecology 79:911927.

    BENDA, L. E., K. ANDRAS, D. MILLER, AND P. BIGELOW. 2004a.Confluence effects in rivers: interactions of basin scale,network geometry, and disturbance regimes. WaterR eso u rc es R esearc h 40: W 05402, do i: 10.1029/

    2003WR002583.BENDA, L. E., N. L. POFF, D. D. MILLER, T. DUNNE, G. H. REEVES,

    G. R. PESS, AND M. M. POLLOCK. 2004b. The networkdynamics hypothesis: how channel networks structureriverine habitats. BioScience 54:413427.

    BRUNS, D. A., G. W. MINSHALL, C. E. CUSHING, K. W. CUMMINS,B. L. BROCK, AND R. L. VANNOTE. 1984. Tributaries asmodifiers of the river continuum concept: analysis bypolar ordination and regression models. Archiv furHydrobiologie 99:208220.

    316 [Volume 27N. P. HITT AND P. L. ANGERMEIER

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    14/17

    BURTON, G. W., AND E. P. ODUM. 1945. The distribution ofstream fish in the vicinity of Mountain Lake, Virginia.Ecology 26:182194.

    CAMPBELL GRANT, E. H., W. H. LOWE, AND W. F. FAGAN. 2007.Living in the branches: population dynamics andecological processes in dendritic networks. EcologyLetters 10:165175.

    COOKE, S. J., C. M. BUNT, S. J. HAMILTON, C. A. JENNINGS, M. P.PEARSON, M . S . COOPERMAN, AND D. F. MARKLE. 2005.Threats, conservation strategies, and prognosis forsuckers (Catostomidae) in North America: insights fromregional case studies of a diverse family of non-gamefishes. Biological Conservation 121:317331.

    CURRY, K. D., AND A. SPACIE. 1984. Differential use of streamhabitat by spawning catostomids. American MidlandNaturalist 111:267279.

    CYTERSKI, M., AND C. BARBER. 2006. Identification andprediction of fish assemblages in streams of the Mid-Atlantic Highlands, USA. Transactions of the AmericanFisheries Society 135:4048.

    DUNHAM, J. B., AND B. E. RIEMAN. 1999. Metapopulationstructure of bull trout: influences of physical, biotic, andgeometrical landscape characteristics. Ecological Appli-cations 9:642655.

    ETNIER, D. A., AND W. C. STARNES. 1993. The fishes ofTennessee. University of Tennessee Press, Knoxville,Tennessee.

    FAGAN, W. F. 2002. Connectivity, fragmentation, and extinc-tion risk in dendritic metapopulations. Ecology 83:32433249.

    FAGAN, W. F., P. J. UNMACK, C. BURGESS, AND W. L. MINCKLEY.2002. Rarity, fragmentation, and extinction risk in desertfishes. Ecology 83:32503256.

    FAUSCH, K. D., C. E. TORGERSEN, C. V. BAXTER, AND H. W. LI.

    2002. Landscapes to riverscapes: bridging the gap between research and conservation of stream fishes.BioScience 52:483498.

    FRISSELL, C. A., W. J. LISS, C. E. WARREN, AND M. D. HURLEY.1986. A hierarchical framework for stream habitatclassification: viewing streams in a watershed context.Environmental Management 10:199214.

    GATZ, A. J., AND S. M. ADAMS. 1994. Patterns of movement ofcentrarchids in two warmwater streams in easternTennessee. Ecology of Freshwater Fish 3:3548.

    GORMAN, O. T. 1986. Assemblage organization of streamfishes: the effect of adventitious streams. AmericanNaturalist 128:611616.

    GOSLEE, S. C., AND D. L. URBAN. 2007. The ecodist package for

    dissimilarity-based analysis of ecological data. Journal ofStatistical Software 22(7). (Available from: http://www.

    jstatsoft.org/v22/i07/paper/)GRENOUILLET, G., D. PONT, AND C. HERISSE. 2004. Within-basin

    fish assemblage structure: the relative influence ofhabitat versus stream spatial position on local speciesrichness. Canadian Journal of Fisheries and AquaticSciences 61:93102.

    GRESSWELL, R. E., AND S. R. HENDRICKS. 2007. Population-scalemovement of coastal cutthroat trout in a naturally

    isolated stream network. Transactions of the AmericanFisheries Society 136:238253.

    GROVES, C. R., D. B. JENSEN, L. L. VALUTIS, K. H. REDFORD, M. L.SHAFFER, J. M. SCOTT, J. V. BAUMGARTNER, J. V. HIGGINS, M.W. BECK, AND M. G. ANDERSON. 2002. Planning for

    biodiversity conservation: putting conservation scienceinto practice. BioScience 52:499512.

    GUENTHER, C. B., AND A. SPACIE. 2006. Changes in fishassemblage structure of impoundments within the upperWabash River basin, Indiana. Transactions of theAmerican Fisheries Society 135:570583.

    HACK, J. T. 1957. Studies of stream profiles in Virginia andMaryland. U.S. Geological Survey Professional Paper294-B. US Geological Survey, Washington, DC.

    HERLIHY, A. T., D. P. LARSEN, S. G. PAULSEN, N. S. URQUHART,AND B . J . ROSENBAUM. 2000. Designing a spatially

    balanced, randomized site selection process for regionalstream surveys: the EMAP Mid-Atlantic pilot study.Environmental Monitoring and Assessment 63:95113.

    HIGGINS, J. V. 2003. Maintaining the ebbs and flows of thelandscape: conservation planning for freshwater ecosys-tems. Pages 292318 in C. Groves (editor). Drafting aconservation blueprint: a practitioners guide to plan-ning for biodiversity. Island Press, Washington, DC.

    HITT, N. P. 2007. Effects of stream network topology on fishassemblage structure and bioassessment sensitivity inthe Mid-Atlantic Highlands, USA. PhD Dissertation,Virginia Polytechnic Institute and State University,Blacksburg, Virginia.

    HITT, N. P., AND P. L. ANGERMEIER. 2006. Effects of adjacentstreams on local fish assemblage structure in westernVirginia: implications for biomonitoring. AmericanFisheries Society Symposium 48:7586.

    HITT, N. P., C. A. FRISSELL, C . C . MUHLFELD, AND F. W.

    ALLENDORF. 2003. Spread of hybridization between nativewestslope cutthroat trout, Oncorhynchus clarki lewisi, andnonnative rainbow trout, Oncorhynchus mykiss. Canadian

    Journal of Fisheries and Aquatic Sciences 60:14401451.HUGHES, R. M., AND J. M. OMERNIK. 1981. Use and misuse of

    the terms watershed and stream order. AmericanFisheries Society Warmwater Streams Symposium 1981:320326.

    ISAAK, D. J., R. F. THUROW, B. E. RIEMAN, AND J. B. DUNHAM.2007. Chinook salmon use of spawning patches: relativeroles of habitat quality, size, and connectivity. EcologicalApplications 17:352364.

    JENKINS, R. E., AND N. M. BURKHEAD. 1994. Freshwater fishes ofVirginia. American Fisheries Society, Bethesda, Mary-

    land.KING, R. S., M. E. BAKER, D. F. WHIGHAM, D. E. WELLER, T. E.

    JORDAN, P. F. KAZYAK, AND M. K. HURD. 2005. Spatialconsiderations for linking watershed land cover toecological indicators in streams. Ecological Applications15:137153.

    KING , R . S . , C . J . RICHARDSON, D . URBAN, AND E. A.ROMANOWICZ. 2003. Spatial dependency of vegetation-environment linkages in an anthropogenically influencedwetland ecosystem. Ecosystems 7:7597.

    2008] 317FISH DISPERSAL IN STREAM NETWORKS

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    15/17

    LEE, D. S., C. R. GILBERT, C. H. HOCUTT, R. E. JENKINS, D. E.MCALLISTER, AND J . C. STAUFFER. 1980. Atlas of NorthAmerican freshwater fishes. North Carolina State Muse-um of Natural History, Raleigh, North Carolina.

    LOWE, W. H., G. E. LIKENS, AND M. E. POWER. 2006. Linkingscales in stream ecology. BioScience 56:591597.

    MANTEL, N. 1967. The detection of disease clustering and a

    generalized regression approach. Cancer Research 27:209220.

    MATTHEWS, W. J. 1986. Fish faunal breaks and stream orderin the eastern and central United States. EnvironmentalBiology of Fishes 17:8192.

    MATTHEWS, W. J., AND H. W. ROBISON. 1998. Influence ofdrainage connectivity, drainage area, and regionalspecies richness on fishes of the interior highlands inArkansas. American Midland Naturalist 139:119.

    MCCORMICK, F. H., AND R. M. HUGHES. 1998. Aquaticvertebrates. Pages 161182 in J. M. Lazorchak, D. J.Klemm, and D. V. Peck (editors). Environmental Mon-itoring and Assessment Programsurface waters: fieldoperations and methods for measuring the ecological

    condition of wadeable streams. EPA 620-R-94004F. USEnvironmental Protection Agency, Washington, DC.

    OMERNIK, J. M. 1987. Ecoregions of the conterminous UnitedStates. Annals of the Association of American Geogra-phers 77:118125.

    OSBORNE, L. L., S. L. KOHLER, P. B. BAYLEY, D. M. DAY, W. A.BERTRAND, M. J. WILEY, AND R. SAUER. 1992. Influence ofstream location in a drainage network on the index of

    biotic integrity. Transactions of the American FisheriesSociety 121:635643.

    OSBORNE, L. L., AND M. J. WILEY. 1992. Influence of tributaryspatial position on the structure of warmwater fishcommunities. Canadian Journal of Fisheries and AquaticSciences 49:671681.

    POIANI, K. A., B. D. RICHTER, M. G. ANDERSON, AND H. E.RICHTER. 2000. Biodiversity conservation at multiplescales: functional sites, landscapes, and networks.BioScience 50:133146.

    RICE, S. P., R. I. FERGUSON, AND T. B. HOEY. 2006. Tributarycontrol of physical heterogeneity and biological diversityat river confluences. Canadian Journal of Fisheries andAquatic Sciences 63:25532566.

    RIEMAN, B. E., AND J. D. MCINTYRE. 1995. Occurrence of bulltrout in naturally fragmented habitat patches of variedsize. Transactions of the American Fisheries Society 124:285296.

    ROGHAIR, C. N., AND C. A. DOLLOFF. 2005. Brook troutmovement during and after recolonization of a naturally

    defaunated stream reach. North American Journal ofFisheries Management 25:785790.

    SCHAEFER, J. F., AND J . R. KERFOOT. 2004. Fish assemblagedynamics in an adventitious stream: a landscapeperspective. American Midland Naturalist 151:134145.

    SCHLOSSER, I. J. 1990. Environmental variation, life historyattributes, and community structure in stream fishes:

    implications for environmental management and assess-ment. Environmental Management 14:621628.

    SCOTT, W. B., AND E. J. CROSSMAN. 1973. Freshwater fishes ofCanada. Bulletin 184. Fisheries Research Board ofCanada, Ottawa, Ontario.

    SHELDON, A. L. 1968. Species diversity and longitudinalsuccession in stream fishes. Ecology 49:193198.

    SHELDON, A. L. 1988. Conservation of stream fishes: patternsof diversity, rarity, and risk. Conservation Biology 2:149156.

    SHELFORD, V. E. 1911. Ecological succession. I. Stream fishesand the method of physiographic analysis. BiologicalBulletin 21:934.

    SHREVE, R. L. 1966. Statistical law of stream numbers. Journalof Geology 74:1737.

    SMITH, T. A., AND C. E. KRAFT. 2005. Stream fish assemblages inrelation to landscape position and local habitat variables.Transactions of the American Fisheries Society 134:430440.

    STRAHLER, A. N. 1957. Quantitative analysis of watershed

    geomorphology. Transactions of the American Geophys-ical Union 38:913920.TYUS, H. M., AND C. A. KARP. 1990. Spawning and movements

    of razorback sucker, Xyrauchen texanus, in the GreenRiver basin of Colorado and Utah. SouthwesternNaturalist 35:427433.

    USEPA (US ENVIRONMENTAL PROTECTION AGENCY). 2000. Eval-uation guidelines for ecological indicators. EPA 620-R-99005. Office of Research and Development, USEnvironmental Protection Agency, Research TrianglePark, North Carolina.

    VANNOTE, R. L., G. W. MINSHALL, K. W. CUMMINS, J. R. SEDELL,AND C. E. CUSHING. 1980. The river continuum concept.Canadian Journal of Fisheries and Aquatic Sciences 37:

    130137.WANG, L., P. SEELBACH, AND R. M. HUGHES. 2006. Introduction

    to landscape influences on stream habitats and biologicalassemblages. American Fisheries Society Symposium 48:123.

    WIENS, J. A. 2001. The landscape context of dispersal. Pages96109 in J. Clobert, E. Danchin, A. A. Dhondt, and J. D.Nichols (editors). Dispersal. Oxford University Press,Oxford, UK.

    WILKINSON, C. D., AND D. EDDS. 2001. Spatial pattern andenvironmental correlates of a midwestern stream fishcommunity: including spatial autocorrelation as a factorin community analyses. American Midland Naturalist146:271289.

    WINSTON, M. R., C. M. TAYLOR, AND J. PIGG. 1991. Upstreamextirpation of four minnow species due to damming of aprairie stream. Transactions of the American FisheriesSociety 120:98105.

    Received: 9 August 2007Accepted: 4 February 2008

    318 [Volume 27N. P. HITT AND P. L. ANGERMEIER

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    16/17

    APPENDIX. Fish species classifications as river and creek habitat specialists. Habitat size assignments are from Jenkins andBurkhead (1994) unless indicated by a footnote.

    Family Scientific name Common name Habitat size

    Atherinidae Labidesthes sicculus Brook silverside RiverCatostomidae Carpiodes cyprinus Quillback River

    Hypentelium roanokense Roanoke hog sucker CreekIctiobus bubalus Smallmouth buffalo Rivera

    Moxostoma macrolepidotum Shorthead redhorse RiverThoburnia rhothoeca Torrent sucker Creek

    Centrarchidae Lepomis megalotis Longear sunfish RiverMicropterus punctulatus Spotted bass RiverPomoxis annularis White crappie River

    Clupeidae Dorosoma cepedianum Gizzard shad RiverCottidae Cottus baileyi Black sculpin Creek

    Cottus bairdi Mottled sculpin CreekCottus cognatus Slimy sculpin Creek

    Cyprinidae Clinostomus elongatus Redside dace Creekb

    Clinostomus funduloides Rosyside dace CreekCyprinella analostana Satinfin shiner RiverCyprinella spiloptera Spotfin shiner RiverCyprinus carpio Common carp River

    Exoglossum laurae Tonguetied minnow CreekLuxilus cerasinus Crescent shiner CreekLythrurus lirus Mountain shiner Creek

    Margariscus margarita Pearl dace CreekNocomis platyrhynchus Bigmouth chub RiverNocomis raneyi Bull chub RiverNocomis micropogon River chub RiverNotropis bifrenatus Bridle shiner CreekNotropis amoenus Comely shiner RiverNotropis atherinoides Emerald shiner RiverNotropis volucellus Mimic shiner RiverNotropis rubellus Rosyface shiner RiverNotropis rubricroceus Saffron shiner CreekNotropis stramineus Sand shiner RiverNotropis photogenis Silver shiner River

    Phenacobius uranops Stargazing minnow RiverPhoxinus oreas Mountain redbelly dace CreekPhoxinus erythrogaster Southern redbelly dace Creeka

    Phoxinus tennesseensis Tennessee dace CreekRhinichthys atratulus Blacknose dace CreekSemotilus atromaculatus Creek chub CreekSemotilus corporalis Fallfish River

    Fundulidae Fundulus catenatus Northern studfish RiverIctaluridae Ictalurus punctatus Channel catfish River

    Noturus flavus Stonecat RiverPylodictis olivaris Flathead catfish River

    Lepisosteidae Lepisosteus osseus Longnose gar RiverPercidae Etheostoma zonale Banded darter River

    Etheostoma camurum Bluebreast darter RiverEtheostoma blennioides Greenside darter River

    Etheostoma longimanum Longfin darter CreekEtheostoma stigmaeum Speckled darter RiverEtheostoma variatum Variegate darter CreekPercina gymnocephala Appalachia darter RiverPercina burtoni Blotchside logperch RiverPercina evides Gilt darter RiverPercina caprodes Logperch RiverPercina rex Roanoke logperch RiverSander vitreus Walleye RiverSander canadense Sauger River

    Percopsidae Percopsis omiscomaycus Trout perch River

    2008] 319FISH DISPERSAL IN STREAM NETWORKS

  • 8/3/2019 Hitt 2008 Evidence Fish Dispersal Spatial Analysis[1]

    17/17

    APPENDIX. Continued.

    Family Scientific name Common name Habitat size

    Petromyzontidae Ichthyomyzon bdellium Ohio lamprey RiverIchthyomyzon greeleyi Mountain brook lamprey CreekLampetra appendix American brook lamprey CreekLampetra aepyptera Least brook lamprey Creek

    Salmonidae Oncorhynchus mykiss Rainbow trout CreekSalmo trutta Brown trout CreekSalvelinus fontinalis Brook trout Creek

    Sciaenidae Aplodinotus grunniens Freshwater drum RiverUmbridae Umbra pygmaea Eastern mudminnow Creek

    a Etnier and Starnes (1993)b Lee et al. (1980)

    320 [Volume 27N. P. HITT AND P. L. ANGERMEIER