Environ Monit AssessDOI 10.1007/s10661-008-0460-x

Linking stream and landscape trajectoriesin the southern Appalachians

Edward P. Gardiner · Andrew B. Sutherland · Rebecca J. Bixby ·Mark C. Scott · Judy L. Meyer · Gene S. Helfman · E. Fred Benfield ·Cathy M. Pringle · Paul V. Bolstad · David N. Wear

Received: 24 December 2007 / Accepted: 26 June 2008© Springer Science + Business Media B.V. 2008

Abstract A proactive sampling strategy was de-signed and implemented in 2000 to documentchanges in streams whose catchment land useswere predicted to change over the next twodecades due to increased building density. Di-atoms, macroinvertebrates, fishes, suspended sed-iment, dissolved solids, and bed composition weremeasured at two reference sites and six sites wherea socioeconomic model suggested new buildingconstruction would influence stream ecosystemsin the future; we label these “hazard sites.”The six hazard sites were located in catchmentswith forested and agricultural land use histories.

E. P. GardinerAmerican Museum of Natural History,New York, NY 10024, USA

E. P. Gardiner (B)Department of Geography,University of Georgia, Athens,GA 30602-2502, USAe-mail: ned@amnh.org

A. B. SutherlandRollins College, 1000 Holt Ave., Winter Park,FL 32789, USA

R. J. BixbyDepartment of Biology,University of New Mexico,167 Castetter Hall, MSC03 2020,Albuquerque, NM 87131, USA

Diatoms were species-poor at reference sites,where riparian forest cover was significantlyhigher than all other sites. Cluster analysis,Wishart’s distance function, non-metric multidi-mensional scaling, indicator species analysis, andt-tests show that macroinvertebrate assemblages,fish assemblages, in situ physical measures, andcatchment land use and land cover were differentbetween streams whose catchments were mostlyforested, relative to those with agricultural landuse histories and varying levels of current andpredicted development. Comparing initial resultswith other regional studies, we predict homoge-

M. C. ScottSouth Carolina Department of Natural Resources,153 Hopewell Rd., Pendleton, SC 29670, USA

J. L. Meyer · G. S. Helfman · C. M. PringleInstitute of Ecology, University of Georgia,Athens, GA 30602-2202, USA

E. F. BenfieldDepartment of Biology, Virginia Polytechnic Institute,Blacksburg, VA 24061, USA

P. V. BolstadDepartment of Forest Resources,University of Minnesota, St. Paul, MN 55108, USA

D. N. WearUSDA Forest Service, P.O. Box 12254,Research Triangle Park, NC 27709, USA

Environ Monit Assess

nization of fauna with increased nutrient inputsand sediment associated with agricultural siteswhere more intense building activities are oc-curring. Based on statistical separability of sam-pled sites, catchment classes were identified andmapped throughout an 8,600 km2 region in west-ern North Carolina’s Blue Ridge physiographicprovince. The classification is a generalized repre-sentation of two ongoing trajectories of land usechange that we suggest will support streams withdiverging biota and physical conditions over thenext two decades.

Keywords Stream ecology · Land use change ·Biotic response · Water chemistry · Watershed ·Catchment classification · Ecological forecasting

Introduction

Because urbanization affects so many streams inthe U.S. and around the world, many authors(Lenat and Crawford 1994; Paul and Meyer 2001;Wang et al. 2003; Allan 2004) point out theneed to examine streams whose catchments arein the early phases of suburban or urban landuse change, i.e., before impervious surfaces dom-inate the hydrologic response, water chemistry,and geomorphology of streams. Such catchmentsare likely to be at the margins of urban centers,for example in suburban areas shifting from lowdensity housing surrounded by forest, pasture, orrow crops to higher density housing, commercial,or industrial land uses. Many biological, chemi-cal, and physical processes in streams change inresponse to land use and land cover transitionsthat occur over large areas. These landscape-level changes include increased sediment loading(Waters 1995; Wood and Armitage 1997), con-taminant loading, and nutrient loading (Delongand Brusven 1998; Hampson et al. 2000), as wellas altered flow regime (Poff et al. 1997), elevatedstream temperature (Scott et al. 2002), increasedchannel incision (Doyle et al. 2000), riparian clear-ing (Jones et al. 1999), loss of large woody debris(Erhman and Lamberti 1992) and habitat homog-enization (Berkman and Rabeni 1987). Landscapechanges in turn manifest themselves through al-tered invertebrate and fish assemblage structure

(Berkman and Rabeni 1987; Sutherland et al.2002).

Agricultural land use negatively impactsstreams by increasing non-point sources ofsediment, nutrients, pesticides and other toxins.The percentage of agricultural land cover isa correlate of change in local-scale fish andinvertebrate abundance and diversity, and large-scale fish assemblage homogenization (Scott2001; Walters et al. 2003). When agricultural landuse exceeded 50% of overall catchment landuse, fish biotic integrity was notably lower inWisconsin streams, relative to that observed inforested catchments; further, agricultural catch-ments supported streams with higher biological in-tegrity than urban catchments (Wang et al. 1997).Increased impervious surface cover, which oftenbrings increased stormflow runoff, channelscouring, and sediment supply, is often associatedwith decreased biological integrity in streams(Waters 1995). When impervious surfaces covermore than approximately 10% of a catchment,hydrologic and geomorphic changes have beenshown to alter stream geomorphology (Boothand Jackson 1997; Doyle et al. 2000), therebyaffecting biological assemblages (Paul and Meyer2001; Roy et al. 2003; Wang et al. 2003). Asland uses change, for example from rural tosuburban or other developed land uses, streamecosystem processes and biota may followpredictable trends. Scientists and managers havea sustained interest in the socioeconomic factorsassociated with catchment land cover and landuse changes and, in turn, their effects on physicaland biological processes in freshwater ecosystems(Carpenter et al. 1999; Strange et al. 1999; Wilsonand Carpenter 1999; Grimm et al. 2000; Gergelet al. 2002). Understanding trajectories of land usechange is necessary in order to forecast associatedimpacts to the physical, chemical and biologicalstructure and function of stream ecosystems.

To address this challenge, we initiated a 30-year study comprised of (a) forecasting buildingactivity, (b) observing catchment land use change,and (c) simultaneously measuring physical, chem-ical, and biological characteristics of streams intarget watersheds. This linked approach will fa-cilitate forecasting stream ecosystem response toland use change, especially in forested headwaters

Environ Monit Assess

of the southern Appalachians. We used an em-pirical model of future land use (Wear andBolstad 1998) to identify sites throughout alarge study area that are likely to support higherdensity housing in the future. Catchments in dif-ferent regions had different land use histories (de-scribed below), and those land use legacies arelikely reflected in current biological and physicalproperties of streams (Harding et al. 1998; Scottet al. 2002; Allan 2004). Using past research as aguide, we designed this study to infer how inver-tebrate and fish assemblages will diverge amongcatchments with different land use trajectories,where “trajectories” refers to the past, present,and expected future land use characteristics ofthose catchments.

In this paper, we document physical, chemi-cal, and biological conditions among streams withdistinct catchment land use histories at the be-ginning of a long-term study. We hypothesizethat catchments with differing land use historiessupport streams with differing physical templets(sensu Southwood 1977) and therefore biota. Wetest this hypothesis through whole-assemblageand landscape-level analyses. First, ordination ofdiatom, macroinvertebrate, and fish assemblagesprovides groupings of sites with distinct taxa. Wecompare landscape-level descriptors as well asphysical characteristics of streams among groupsof sites, as defined by the ordinations. Second,we describe the biological changes that are likelyto accompany two common trajectories of landuse and land cover change occurring in westernNorth Carolina: forested land cover giving wayto low density residential land use; and agricul-tural land use yielding to higher density subur-ban. Third, we use cartographic modeling to mapthe spatial arrangement of catchments through-out a large study area that are likely to havesimilar flora and fauna, relative to the collectionsdescribed here.

Study area and methods

The study area includes most of western NorthCarolina’s major river systems in the Blue Ridgephysiographic province in four sub-catchments of

the upper Tennessee River: the upper LittleTennessee (LT; 1,154 km2), Tuckaseegee(1,706 km2), Pigeon (1,403 km2), and UpperFrench Broad (FB; 4,309 km2) basins (Fig. 1).The backwater reaches of Lake Fontana, a41 km2 reservoir on the Little Tennessee River,define the downstream limits of the LT andTuckaseegee River basins considered here. TheTuckaseegee was historically a major tributaryof the Little Tennessee but now feeds directlyinto the Lake Fontana impoundment. The NorthCarolina border with Tennessee defines thenorthern extent of the Pigeon and FB studybasins. Elevation in the study area ranges from400 to 2,000 m a.s.l. In high gradient catchments,land cover is predominantly forested while landuse is mostly rural and low density residential.Row crops (small farms), pasture, light industry,tourist-oriented businesses, small urban centers,and suburban land uses predominate in thevalleys. The Coweeta Long Term EcologicalResearch (LTER) site is near Franklin, N.C.in the LT basin. Asheville (population approx.70,000), located in the north-central portion of theFB basin, is the largest city in the region. Franklinis a small urban center (population approx. 3,500)in the LT basin. Canton (population approx.4,000) is a comparably sized city in the PigeonRiver basin.

Two basins in our study area, LT and CaneCreek (a sub-basin within the FB), typify twodistinct land use histories and trajectories (seeWear and Bolstad 1998). In the LT, the most com-mon land cover change between 1950 and 1990was the conversion of non-forested to forestedland cover and from forested without buildings toforested with buildings. Wear and Bolstad (1998)concluded that agricultural, timber production,and recreational land uses in the LT had shiftedtoward low density residential development as-sociated with vacation homes. The Cane Creekcatchment had higher proportions of agriculturalland use in both 1950 and 1990 as well as substan-tially higher housing density (occasionally exceed-ing eight building ha−1) than the LT watershed.In the Cane Creek basin, land use transitionsbetween 1950 and 1990 were split equally be-tween forest clearing and reforestation, but bothland cover transitions supported equal increases

Environ Monit Assess

Fig. 1 Sampling catchments and study area in the southern Blue Ridge province of western North Carolina

in housing density. Whereas rural second homedevelopment in forested catchments was evidentin the LT, suburban land use was replacing agri-cultural land use in Cane Creek at the outset ofthis study.

Site selection

We used projected growth in building den-sity from the LT and Cane Creek (Wear andBolstad 1998) to establish a set of stream research

Environ Monit Assess

sites where ecosystem changes will likely occur incoming years. In the southern Blue Ridge physio-graphic province, streams in forested catchmentsare typically cool, clear, low in nutrients and pri-mary productivity, and have relatively low fishdiversity (Wallace et al. 1992; Scott and Helfman2001). Headwater streams may be important refu-gia for endemic fish species as well as sensitiveriverine taxa that migrate to small streams tospawn, or for thermal refugia. Decreased forestcover in a catchment can raise temperatures, in-crease nutrient supply (Scott and Helfman 2001;Scott et al. 2002), and lead to increased suspendedand bedload sediment yield in streams (Lenat andCrawford 1994; Harding et al. 1998; Sutherlandet al. 2002). We therefore expect increases insediment flux, nutrient supply, and runoff whenforested land with no buildings is converted tonon-forested land with buildings. In streams thatdrain watersheds that change from agricultural tourban land cover we expect to see a long-termreduction in sediment, increased metal concentra-tion, increased temperature due to riparian loss,channel incision, and lower baseflow accompa-nied by greater discharge immediately followingstorms.

Two metrics were used to identify sites witha high likelihood of increased building density.The first metric, the difference between buildingdensity projected for 1990 and building densityobserved in 1990 (Wear and Bolstad 1998), wasa measure of the uncapitalized value inherent inland parcels that were accessible and in desirablelocations in 1990. The second metric, the differ-ence between building density projected for 2030and observed building density for 1990 (Wearand Bolstad 1998), was a measure of expectedfuture trends in land use. Where both indicesexceeded three buildings in a 9 ha neighborhood,we inferred a high likelihood that building activitywould proceed by 2030. Observable changes tostreams are expected where building activities areconcentrated near streams, so pixels meeting eachof these criteria were mapped if they lay within100 m of streams on previously undeveloped landparcels. The 100 m distance was chosen in orderto include construction activities near streams butnot confined to active floodplains. Our definitionof “undeveloped” was a land parcel that had been

forested since 1950 and that had fewer than onebuilding per 9 ha in both 1950 and 1990. This 40-year horizon was intended to limit the confound-ing influence of the “ghost of land use past,” aconcept that posits biota may differ in catchmentsthat have similar present-day catchment land usebut different historic land use (Harding et al.1998; Allan 2004). Pixels that met all the criteriadescribed above were included in the CompositeHazard Index used henceforth.

Potential study catchments were restricted tosizes between 10 and 40 km2 in area and outletelevations between 550 and 720 m a.s.l. becausethese size and site elevation ranges were associ-ated with shifts in assemblage structure for fishesin this region (Scott 2001). Watersheds containingrelatively large numbers of Composite HazardIndex pixels were identified and visited. At thebeginning of this study, buildings were alreadybeing constructed in many of those catchments, in-dicating our land use trajectory model had merit.Six catchments with numerous Composite HazardIndex pixels but which did not show evidence ofrecent building activity near streams were selectedfor long-term study (Fig. 1, Table 1). These arehenceforth termed “hazard sites.” Two referencestreams were located on land owned and managedby the US Forest Service (USFS) where land usewas not expected to change over the projected 30-year duration of the study. The reference streamlocated in the LT basin is Coweeta Creek andis at the entrance of the USFS Coweeta Hydro-logic Laboratory. Further evidence for the efficacyof Wear and Bolstad’s (1998) empirical modelwas that the Composite Hazard Index predictedthe location of forest service buildings erected atCoweeta in 2002, well after the publication date ofthe model. The reference stream located in the FBbasin is Avery Creek, and is in the Pisgah NationalForest near Brevard, N.C. Both reference sitesare on public lands supporting recreation and landmanagement research.

Ecosystem assessments

Biological assessments

Biological and physical assessments were con-ducted during the summer of 2000 at the six

Environ Monit Assess

Tab

le1

Lan

dus

e,%

com

posi

teha

zard

inde

xpi

xels

wit

hin

100

mof

stre

ams,

alga

l,an

dge

omor

phic

fact

ors

amon

gth

eei

ght

“haz

ard”

site

sch

osen

for

long

-ter

mre

sear

ch

Upp

erL

ittl

eT

enne

ssee

Riv

erba

sin

Upp

erF

renc

hB

road

Riv

erba

sin

Dar

nell

Way

ahW

atau

gaC

owee

taa

Ave

rya

Gap

Rob

inso

nH

oope

rs

Lan

dus

ean

dpr

ojec

ted

deve

lopm

ent

%F

ores

t199

398

9885

9810

072

6572

%D

evel

oped

1993

0.0

0.11

5.3

00.

020.

866.

20.

63%

Agr

icul

ture

1993

1.5

2.2

8.8

1.2

0.44

2728

28%

For

est—

10m

Buf

fer,

1993

80.3

56.6

44.9

88.9

83.6

46.9

36.6

67.4

%F

ores

t197

099

9884

100

9970

6473

%F

ores

t195

097

9876

9810

064

6468

%C

ompo

site

Haz

ard

Inde

xin

100

mSt

ream

Buf

fer

<1

2.2

2.0

<1

<1

7.7

7.6

3.4

Alg

alan

dge

omor

phic

cond

itio

nsA

FD

M(m

gm

−2;s

.d.;

n=

3)1.

1(1

.3)

4.8

(11)

1.7

(0.8

4)0.

85(1

.1)

0.88

(0.2

9)2.

2(0

.90)

0.8

(0.8

4)1.

2(0

.98)

chl-

a(μ

gm

−2;s

.d.;

n=

3)11

5.7

(112

.4)

89.1

(42.

8)10

9.1

(119

.7)

35.1

(21.

2)60

.4(4

7.4)

299.

6(2

26.9

)42

.4(2

9.6)

48.7

(18.

1)A

vera

geφ

(n=

100)

−4.4

−5.3

−3.8

−4.6

−5.5

−5.1

−2.9

−2.6

D90

(mm

)20

020

012

015

028

040

072

22%

Fin

es(n

=3

per

site

)9.

610

.015

.53.

313

.713

.118

.436

.1T

SS(m

gl−

1 ;s.

d.;n

=3)

14.7

(22.

7)13

.3(6

.1)

132.

3(1

4.0)

6.1

(4.7

)3.

2(0

.8)

14.7

(2.8

)8.

0(2

.8)

19.2

(7.6

)a R

efer

ence

site

Environ Monit Assess

hazard sites and two reference sites. Benthic pe-riphyton was quantitatively sampled with a mod-ified Loeb sampler (Loeb 1981). Three sampleswere collected and composited from representa-tive rocks at 10-m intervals along a 100-m tran-sect above the downstream end of each samplingreach. Samples were preserved on ice, bulked,and later sub-sampled to determine chlorophyll aconcentration and ash-free dry mass (AFDM) us-ing standard methods (Wetzel and Likens 1991).Subsamples were processed using standard meth-ods (Lowe and LaLiberte 1996) and preparedas permanent slides to enumerate diatom speciescomposition and densities to the lowest taxonomicunit possible using algal floras from the south-eastern USA (Camburn et al. 1978; Kociolek andKingston 1999). Only diatoms were quantified be-cause they comprise the majority of periphyton insouthern Appalachian streams (Lowe et al. 1986;Greenwood and Rosemond 2005). Macroinver-tebrates were collected by quantitative kick netsampling within each sample reach (see Hardinget al. 1998). Macroinvertebrates were identified togenus, and densities were recorded. At each site,fishes were collected and identified to species fromrepresentative 50-m reaches between April 16and July 6, 2000 using a backpack electroshocker,seines, and dip nets. Fishes were enumeratedand classified as “cosmopolitan” or “highlandendemics” which generally refer to widespreadspecies and those restricted to high gradient, head-water streams in the southern Appalachians (seeScott and Helfman 2001).

Specimens for all taxa have been accessioned atthe Georgia Museum of Natural History for futurereference. For diatoms, this includes permanentslides and the remainder of samples (preserved informalin). All macroinvertebrates were preservedin alcohol and archived. Voucher fish specimenshave previously been catalogued. These museumrecords establish an important reference collec-tion for future efforts.

Physical assessments

Sediments were sampled using a 60 cm high ×25 cm diameter stainless steel cylinder in threeriffles and three pools at each site to a depthof 10 cm wherever bedrock or boulders were

not reached first. Large sediment (>64 mm) wasweighed in the field; remaining sediment was driedand weighed in a laboratory to calculate % byweight of coarse (2 to 64 mm) and fine (

Environ Monit Assess

catchment-wide % forested area, and catchment-wide % agricultural land cover in the 1970s and1990s. To assess land cover differences amonggroups of sites identified through ordination (asdescribed in the following section), PC1 wasused in lieu of those individual, multicollinearmeasures.

Statistical analyses

Differences among diatom, macroinvertebrate,and fish assemblages at all sites were analyzedusing cluster analysis (Wishart 1969; Greig-Smith1983), Wishart’s objective function (Wishart1969), and nonmetric multidimensional scaling(NMS; Kruskall and Wish 1978) in PC-Ord(McCune and Mefford 1999). Both cluster analy-sis and NMS require a measure of similar-ity or dissimilarity between all possible pairsof sites, so a Bray–Curtis dissimilarity matrixwas calculated from fourth-root transformationsof diatom density (cells/ml), macroinvertebratedensity (#/m2), and fish abundance (#/sample).This transformation reduced the influence ofthe most abundant taxa on site scores, allow-ing less common species to contribute to siterankings (Clarke 1993). The group average dis-tance function was used to calculate between-sitedistances in the n-dimensional space defined byBray–Curtis dissimilarities (UPGMA in McCuneand Mefford 1999) so that sites were given equalweight regardless of the number of observationsper site.

Wishart’s (1969) objective function indicateshow much of the total sum of squared distancesbetween sites (based on Bray–Curtis dissimilarity)is captured by that group of sites. The % Infor-mation Remaining statistic indicates the relativedistance, in Bray–Curtis dissimilarity space, be-tween individual sites or between groups of sitesdefined by branching points in a dendrogram. Ifa given group spans a large proportion of theobjective function distance, that group consistsof sites with heterogeneous assemblages. If theobjective function’s span for a group is small,that group is relatively homogeneous. In clusteranalysis, it is customary to prune a cluster den-drogram, a process whereby the analyst defines

unique clusters based on the maximum allowabledifference between sites within a given group. TheWishart objective function (Wishart 1969) pro-vides a standard measure against which to prunedendrograms. If a dendrogram is pruned at so thatthere is a large % Information Remaining, theresulting groups are homogeneous. Conversely, as% Information Remaining decreases for a givendendrogram node, groups of sites in adjacent den-drogram branches are biotically more similar. Ifgroups of sites branch at 0% Information Remain-ing, there is no floral or faunal affinity amonggroups in those branches. Due to small samplesizes, we pruned trees with narrowly defined clus-ters at 75% Information Remaining.

Ordination plots from NMS provided anotherrepresentation of between-site similarity and dis-similarity for each taxonomic group. To ensurean optimal NMS solution, PC-Ord’s (McCuneand Mefford 1999) “autopilot/slow and thorough”mode was used in which 40 iterations were per-formed on the Bray–Curtis matrix; 50 additionalruns were performed with randomized data sothat the final solution could be compared to so-lutions that might be obtained by chance alone.These methods are well represented in ecologicalliterature comparing aquatic assemblages to oneanother and to environmental gradients (McRaeet al. 1998; Ford and Rose 2000; Hawkins et al.2000; Hawkins and Vinson 2000; McCormick et al.2000; Heino et al. 2003; Mykra et al. 2004; Ilmonenand Paasivirta 2005). Clarke (1993) discussed indetail several advantages of the NMS method-ology employed here. Cluster dendrograms andNMS ordination plots visually and quantitativelydemonstrate assemblage differences among sitesin this study.

For all of the ordinations and classifications de-scribed above, indicator species analysis (Dufreneand Legendre 1997) highlighted taxa that con-tributed to similarity and differences within andbetween groups of sites. Monte Carlo tests ofsignificance based on 1,000 randomizations of theoriginal species matrices were used to indicate thelevel of significance of indicator values (McCuneand Mefford 1999). Physical and chemical differ-ences between site groupings were analyzed usingt-tests and ANOVA in S-plus (Mathsoft 1999)

Environ Monit Assess

based on the groups identified through clusteranalysis and NMS ordination.

Inferring landscape and stream trajectories

We compared the fish data already describedwith fish collections from 1995 and 1996 (seeScott 2001) from eight sites meeting the elevationrange and catchment size criteria established atthe outset. We recognize that abundance, pres-ence, and absence of taxa vary at a site year toyear (Freeman et al. 1988), but ordination ex-amines the ranking of species across sites; ourfourth-root transformation (see above) of rawdata further emphasized presence and absencewithin each sites flora and fauna. If only a few taxaare present one year but not another, then ordi-nation should be robust for analyzing landscapecorrelates of entire assemblages. Previous stud-ies suggested that fish assemblage structure maybe considered stable in this context, especiallycompared to macroinvertebrates (Desmond et al.2002). Because land use and land cover are notgenerally changing rapidly in our study area, landuse data used in this study were valid for the datacollected in 1995 and 1996. Scott’s (2001) fish col-lection methods were identical to those describedabove. By viewing more sites, we were able to ex-amine fish assemblage differences in catchmentswith more residential land use but which wereotherwise similar to the hazard sites’ catchments.For heuristic purposes, these data were pooledwith the collections from 2000 and ordinated us-ing methods described above. The additional sitesextended the gradient of development intensitysampled for this study, thus providing a space-for-time substitution that allowed us to explorepossible future biotic states in hazard site catch-ments. Using the methods described above, ordi-nation was used to identify distinct assemblagesof fishes. We identified land use factors that werecorrelated with groups of sites identified throughordination. We classified catchments based on theland use factors that were strong predictors of fishassemblage structure in this pooled data set, thusproviding a whole-landscape set of predictionsabout fish assemblages across a large portion ofwestern North Carolina.

Table 2 Summary of ordination results: most of the inter-site differences in Bray–Curtis space were encapsulated intwo-axis NMS solutions with minimal stress

% Variance in distance matrix represented by NMS axes

Focal group Final stress Axis 1 Axis 2

Diatoms 3.9 45% 40%Macroinvertebrates 3.2 7% 86%Fishes 2.5 18% 79%

Results

Assemblage ordinations

All three NMS ordinations of assemblage dataproduced two-axis solutions with low instability(i.e., instability

Environ Monit Assess

Fig. 2 Nonmetricmultidimensional scalingordination results fordiatoms sampled frombenthic periphyton ateight hazard sites. A plotof site scores (a) showsthat the two referencesites are very distinctfrom the remainder ofsites in the study. Clusteranalysis (b) confirmedthis pattern

Distance (Objective Function)

Information Remaining (%)

1E-01

100

3.8E-01

75

6.5E-01

50

9.3E-01

25

1.2E+00

0

AveryCoweetaDarnellGapWayahRobinsonHooperWatauga

Avery

Coweeta

Darnell

Gap

Hooper

Robinson

Watauga

Wayah

Axis 1

Axi

s 2

a

b

In other words, there was a detectable differencein diatom assemblages in forested vs. non-forestedcatchments, and catchments with intermediateforest cover included species found in both ofthose other groupings, i.e., cosmopolitan taxa.Analysis of variance showed that the gradientfrom high to low riparian forest was statisticallysignificant (p = 0.04) across those three sets ofsites. No other physical descriptor of stream habi-tats or their catchments was statistically differentamong the groups of sites identified through clus-ter analysis of diatom assemblages.

Clustering among macroinvertebrates yieldeddifferent groupings than those observed based ondiatoms. One relatively tight cluster, consisting ofthe reference sites and two others from the LTbasin, is evident in the NMS ordination and clusteranalysis of macroinvertebrate densities (Fig. 3).This cluster spanned less than 20% of the Wishartobjective function (Fig. 3b) owing to the simi-

larity of density and diversity of taxa amongstthese sites. Macroinvertebrate assemblages at theremaining sites were not similar to one anothernor to the tightly grouped sites. Seven macroin-vertebrate genera were found among the tightlygrouped sites and nowhere else (Table 3), a resultthat did not likely occur by chance alone (MonteCarlo permutation test p = 0.033 for each). Thetight cluster of sites was more speciose thanthe remaining sites (59 vs. 33 species on aver-age; one-tailed t-test, assuming unequal variancep = 0.005). The density of individuals in or-ders Ephemeroptera, Plecoptera, and Trichoptera(EPT) was significantly higher in the tight groupof sites than among all other sites (t-test, p = 0.04based on log-transformed data). The land coverand terrain metric PC1 was significantly differ-ent (t-test, p = 0.002) between the clustered andunclustered sites. Greater catchment-wide forestcover (and less agricultural area) was associated

Environ Monit Assess

Tab

le3

Tax

aas

soci

ated

,via

indi

cato

rsp

ecie

san

alys

isM

onte

Car

lope

rmut

atio

nte

sts,

wit

hgr

oups

ofsi

tes

iden

tifie

dby

NM

San

dcl

uste

ran

alys

es(F

igs.

2,3,

and

4)

Dia

tom

sat

Dia

tom

sin

othe

rcl

uste

rM

acro

inve

rteb

rate

sat

No

mac

ro-

Fis

hes

only

atle

ssF

ishe

son

lyin

mor

ere

fere

nce

site

s(G

ap,R

obin

son,

and

Way

ah)

refe

renc

esi

tes

inve

rteb

rate

clus

ters

deve

lope

dsi

tes

deve

lope

dsi

tes

(Gap

,am

ong

othe

rsi

tes

Hoo

per,

Rob

inso

n)

Ach

nant

hes

hung

aric

aA

chna

nthe

ssu

chla

ndtii

Ace

ntre

lla(E

phem

erop

tera

Rhi

nich

thys

cata

ract

aeN

otro

pis

Bae

tida

e)ru

bric

roce

usD

iato

ma

mes

odon

Cym

bella

affi

nis

Par

alep

toph

lebi

a(E

phem

erop

tera

Onc

oryn

chus

myk

iss

Rhi

nich

tys

(Ehr

.)K

ütz.

aL

epto

phle

biid

ae)

atra

tulu

sd

Eun

otia

inci

saW

.Sm

.aF

ragi

lari

acr

oton

ensi

sC

inyg

mul

a(E

phem

erop

tera

Hyp

ente

lium

Hep

tage

niid

ae)

nigr

ican

sd

Eun

otia

pect

inal

isva

r.G

eiss

leri

ade

cuss

is(Ø

stru

p)T

alla

perl

a(P

leco

pter

aE

theo

stom

am

inor

(Küt

z.)

Rab

eha

Lan

ge–B

erta

ndM

etze

ltin

Pel

tope

rlid

ae)

flabe

llare

inL

ange

–Ber

t.c

Mer

idio

nal

ansm

ithii

Mel

osir

ava

rian

sA

gard

hcG

loss

osom

a(T

rich

opte

raSe

mot

ilus

Bra

ntb

Glo

ssos

omat

idae

)at

rom

acul

atus

d

Syne

dra

acus

Nav

icul

acr

ypto

ceph

ala

Lep

idos

tom

a(T

rich

opte

raK

ütz.

cL

epid

osto

mat

idae

)A

ther

ix(D

ipte

raA

ther

icid

ae)

Due

tosa

mpl

esi

zes,

p-va

lues

diff

erfo

rM

onte

Car

lote

sts

diff

erby

taxo

nom

icgr

oup:

diat

oms

(p<

0.1)

;mac

roin

vert

ebra

tes

(p<

0.03

3);a

ndfis

hes

(p<

0.1)

a Tax

aid

enti

fied

byG

reen

woo

dan

dR

osem

ond

(200

5)an

dP

otap

ova

and

Cha

rles

(200

7)as

indi

cati

veof

fore

sted

,olig

otro

phic

head

wat

erst

ream

sbSh

ade-

tole

rant

,end

emic

toso

uthe

rnA

ppal

achi

anst

ream

s(B

rant

2003

)c S

omew

hats

hade

-int

oler

ant,

mes

otro

phic

diat

oms

dC

osm

opol

itan

fishe

slis

ted

bySc

otta

ndH

elfm

an(2

001)

Environ Monit Assess

Fig. 3 Nonmetricmultidimensional scalingordination (a) and clusteranalysis (b) results formacroinvertebrates fromeight hazard sites

Distance (Objective Function)

Information Remaining (%)

4E-02

100

1.8E-01

75

3.2E-01

50

4.6E-01

25

6E-01

0

CwtAveDarWayWatGapRobHoo

Darnell

Wayah

Watauga

Gap

Robinson

Hooper

Axis 1

Axi

s 2

Coweeta

Avery

a

b

with the tight cluster of sites. Nitrate (p < 0.001)and sodium (p = .001) were higher in the tightcluster; cations (K, p < 10−6; Ca, p = 0.001; Mg,p = 0.001), SO4 (p = 0.006), temperature range(p = 0.02), and substrate particle size (φ, p =0.006) were all lower in the tight cluster of sitesrelative to the remaining sites.

Two groups comprising three sites each arereadily apparent in the NMS ordination graph(Fig. 4a) as well as at the 75% Information Re-maining level of the cluster analysis (Fig. 4b) offish observations. Coweeta, Wayah, and Darnellcreeks had more (t-test on angular transforma-tion of proportions; p < 0.001) highland endemicfishes (76% by abundance) than the three sites inthe lower right in Fig. 4a (46%, by abundance).The group in the upper left also had fewer cos-mopolitan fishes (1%, by abundance, of fauna)than the group to the lower right (42%, by abun-dance; t-test, p = 0.0035). The group to the up-per left (Fig. 4a) had significantly fewer speciesthan the group depicted in the lower right (mean

richness 8.3 vs. 13 species; t-test p = 0.024). Theland cover and terrain metric, PC1, was signifi-cantly different between the groups of sites dis-tinguished by fish abundance and diversity (p <0.0001). Nutrients and trace metals had lowerconcentrations for the group with more highlandendemics: NO3(p < 0.005), K (p < 0.0005), Na(p < 0.005), Mg (p < 0.05), and S04 (p < 0.01).Most of the fishes found in the tight cluster ofsites with higher ion concentrations and less forestcover were cosmopoloitan taxa (Table 3; Scottand Helfman 2001). While not endemic high-land species per se (Scott and Helfman 2001),both fish species found in the more forested sites(Coweeta, Wayah, and Darnell) are commonlyassociated with headwater streams with clean sub-strate (Jenkins and Burkhead 1994).

Landscape-level inferences

An NMS ordination (NMS) of fish assemblagesat 16 sites (Fig. 5a)—including eight sites not

Environ Monit Assess

Fig. 4 Nonmetricmultidimensional scalingordination (a) and clusteranalysis (b) results forfishes from eight hazardsites

Distance (Objective Function)

Information Remaining (%)

3.6E-02

100

2E-01

75

3.6E-01

50

5.3E-01

25

6.9E-01

0

AveryCoweetaWayahDarnellGapHooperRobinsonWatauga

Axis 1

Axi

s 2

Wayah

WataugaCoweeta

Darnell

Gap

Robinson

Hooper

Avery

a

b

analyzed in the previous analysis—encapsulatedabout 90% of the variation present in the matrixof Bray–Curtis dissimilarities between all sites.The positions of sites where future developmentis expected, relative to reference sites, was identi-cal to that observed prior to analyzing the eightadditional sites (Fig. 4a). Similarly, the clusteranalysis dendrogram positions for sites werevery similar between analyses (Figs. 4b and 5b).Four clusters are apparent at the 50% informa-tion remaining level (Fig. 5a dotted lines; Fig. 5b).The two groups with positive Axis 2 scores allhad catchments with greater than 85% forestedarea, road density less than 3 km/km2, andcatchment-wide agricultural land use less than3%. Sites with negative Axis 2 scores had lessforested area and greater road density. This dif-ference in land cover was accompanied by a suiteof differences in water chemistry and temperaturerange (Table 4).

The distinctions observed among groups ofsites depicted in Fig. 5 can be used to derive

a decision tree that can be used to hypothesizeecosystem states likely to occur across the region(Fig. 6a). Compared to Forested sites, catchment-wide agricultural land use (p < 0.01) was greateramong sites depicted in the upper left corner ofFig. 5a. These are labeled “Rural.” Sample sizeswere too small to yield statistically powerful re-sults for the sites with negative Axis 2 scores.It appeared there were more roads in the sitesto the lower right of Fig. 5a (hereafter termed“Suburban/Urban”) than in the lower left (here-after termed “Suburbanizing Agriculture”; t-test,p = 0.16). Haw Creek was an outlier in this analy-sis; its channel is dominated by concrete, bouldersused for stabilization, and its catchment consistsmostly of urban land use. There were clear breaksin this data set that separated Forested, Rural,Suburbanizing Agriculture, and Suburban/Urbansites (Fig. 6a) which were used to classify water-sheds of comparable size throughout the studyarea (Fig. 6b). Land use characteristics of eachclassification (Table 5) show that these classes

Environ Monit Assess

Fig. 5 Nonmetricmultidimensional scalingordination (a) and clusteranalysis (b) results forfishes observed at 16 siteswith comparablecatchment size and outletelevation. Dotted lines ina represent the 50%Information Remainingcriterion

Distance (Objective Function)

Information Remaining (%)

1.7E-02

100

3.9E-01

75

7.5E-01

50

1.1E+00

25

1.5E+00

0

UpperDavAveryLookingGCoweetaWayahDarnellCampgrouBettyTellicoWataugaBeaverdaSweetenGapRobinsonHoopersHaw

Upper Davidson

Looking Glass

Campground

TellicoBetty

Beaverdam

Haw

Sweeten

CoweetaDarnell

WayahWatauga

Avery

Gap

RobinsonHooper

Axis 1

Axi

s2

50% Information Remaining

Forested

Rural

Suburbanizing Agricultural

Suburban

% Forest > 85%

% Forest < 85%

Road Density increases

a

b

Table 4 Nutrientchemistry andtemperature rangeANOVA results forstreams in four catchmentclasses identified throughfish ordination

Forested Rural Suburbanizing Urban(n = 6) (n = 4) agriculture (n = 3) (n = 3)

NO3 (mg l−1) 0.044 0.083 0.33 0.69NH4 (mg l−1) 0.011 0.014 0.021 0.059PO4 (mg l−1) 0.018 0.024 0.021 0.025K (mg l−1) 0.53 0.69 1.83 2.01Na (mg l−1) 1.52 1.44 4.85 4.84Ca (mg l−1) 0.98 1.29 5.36 6.59Mg (mg l−1) 0.41 0.48 1.73 2.84SO4 (mg l−1) 0.97 1.14 2.75 4.92Late summer temp. range (◦C) 17.8 19.3 18.9 21.0

Environ Monit Assess

Fig. 6 Catchments withsimilar areas and outletelevations were comparedto sampled streams toinfer possible streamecosystem states throughthe study area. Decisionrules are outlined in (a)while the resultantcatchment classes aremapped onto thelandscape in (b)

GeorgiaS. Carolina

N. Carolina

Virginia

Tennessee

abama

Southeastern United States

InsetArea

±

Pigeon

Tuckaseegee

Little Tennessee

French Broad

Cane Creek

10 0 10 20 30 405

Kilometers

Watershed Classes

Rural

Forested

SuburbanizingAgricultureSuburban/Urban

550 m < outlet elevation < 720 m; 10 km2 < area < 40

km2

≥ 85% Forested < 85% Forest

Catchment-wide Agricultural Land

Use < 3%Forested

Catchment-wide Agricultural Land

Use > 3%Rural

Catchment-wide Development < 10%

Suburbanizing Agriculture

Catchment-wide Development > 10%

Suburban/Urban

b

a

represent a gradient of development intensity. Wemapped these decision rules onto the entire pop-ulation of catchments throughout the study areawhose elevation and size matched those of thestudy sites (Fig. 6b).

Discussion

The distinctiveness of algal flora between theforested reference sites and the remaining catch-ments, all of which had lower riparian forest cover

than the reference sites, points to diatoms’ sen-sitivity to light availability. Low diatom speciesrichness observed at the reference sites set themapart from all other sites and is consistent withthe expectation that highly forested sites with lowlight and nutrient inputs should have lower pri-mary productivity and low diversity of algal taxa(Lowe et al. 1986). This study and others (Loweet al. 1986; Mosisch et al. 2001; Greenwood andRosemond 2005) demonstrate that canopy coverplays a major role in influencing algal commu-nity structure and growth and can be the primary

Environ Monit Assess

Tab

le5

Lan

dco

ver

and

land

use

attr

ibut

esof

clus

ters

ofco

mpa

rabl

ysi

zed

catc

hmen

tsw

ith

sim

ilar

outl

etel

evat

ions

Are

a(%

)w

ith

give

nla

ndco

ver

Roa

dde

nsit

y(k

mkm

−2)

n19

70F

ores

t19

70F

ores

tin

1993

For

est

1993

For

esti

n19

93A

g.19

93A

g.in

1993

%D

evel

oped

Impr

oved

Uni

mpr

oved

Tot

alin

catc

hmen

t10

0m

buff

erin

catc

hmen

t10

0m

buff

er10

0m

buff

erca

tchm

ent

catc

hmen

t

For

este

d37

97.1

95.1

97.1

94.3

2.0

1.2

1.7

1.5

0.46

1.5

Rur

al42

90.7

84.4

92.3

85.6

9.6

5.4

2.4

1.7

0.57

1.7

Subu

rban

izin

gag

ricu

ltur

e45

73.6

65.2

79.6

61.4

30.7

21.4

5.8

2.4

0.45

2.4

Urb

an25

56.6

48.1

59.1

52.6

22.1

19.1

21.8

5.0

0.3

5.0 driver of periphyton biomass and communities

compared to parameters such as temperature ornutrient inputs. Additionally, the combination ofboth increased irradiance and nutrients (whichwe predict will accompany deforestation that ac-companies increased building density) is likely tocause increases in periphyton biomass and shiftsin community structure (Rosemond 1993). Wepredict that assemblages typified by endemic (i.e.,Meridion alansmithii), shade-tolerant, and olig-otrophic taxa in largely forested catchments willinclude more cosmopolitan taxa if those catch-ments become more developed and have less ri-parian forest cover through time.

Macroinvertebrate and fish assemblages alsosuggest broad differences between stream seg-ments in catchments with differing land use, waterchemistry, and physical habitats. Prior research(Scott and Helfman 2001) suggests lower macroin-vertebrate diversity and higher fish diversity willfollow building activities within forested catch-ments. For example, Darnell and Wayah Creeksboth had low Composite Hazard Index val-ues (Table 1) and high forested area in the1950s and in 1993; they also clustered withthe reference sites (Fig. 4). Forested catchmentsthat undergo limited development are expectedto support somewhat higher fish diversity ascosmopolitan taxa (prevalent in low-elevation,warmer streams) invade but do not displace high-land endemic fishes found in headwater streams(Harding et al. 1998; Scott and Helfman 2001;Walters et al. 2003).

Biologists point to the need for cost effectivemeans of quickly assessing whether one streamsegment is likely to support more sensitive taxathan another (Hawkins et al. 2000). To gener-ate a spatial hypothesis about stream conditionsthroughout the study area at the start of thislong-term study, we used simple decision rulesto separate similarly sized catchments with outletstreams having similar elevation ranges through-out the study region (Fig. 6). Details on watersheddatabase construction can be found elsewhere(Gardiner 2002). The largely forested LT basincontains mostly “Forested” and “Rural” catch-ment categories while the FB has more “Sub-urbanizing Agricultural” and “Suburban/Urban”ones.

Environ Monit Assess

Some catchments of comparable size and ele-vation in this study area contain agricultural areasapproaching 50% (data not shown), a thresholdWang et al. (1997) associated with loss of bioticintegrity among streams in southern Wisconsin.These more agricultural catchments are found inareas that are currently undergoing suburbaniza-tion and urbanization, and therefore representuseful targets for sampling efforts. In the Forestedand Rural catchments, road density was belowthe 1.1 km km−2 density threshold associated withsubstantial hydrologic changes by others (Formanand Alexander 1998). Thus these two classes rep-resent a distinct trajectory. The other two classes,which are each beyond that threshold, representanother land use trajectory from agricultural landuse to Suburban and Urban land uses. Ecologistshave emphasized the need for research examin-ing streams whose watersheds consist of 6% to12% impervious surface area (Paul and Meyer2001; Wang et al. 2003) to understand ecosys-tem states below, at, and above the thresholdbeyond which hydrologic impairment and loweredbiotic integrity often occur (Booth and Jackson1997). This study is well-poised to detect and de-scribe physical, chemical, and biological changesin streams whose catchments will likely exceedimpairment thresholds in the near future.

Conclusion

Sampling at the hazard sites in the next twodecades aims to link specific physical and bio-logical changes within streams to accompanyingcatchment-wide land use changes. We infer likelyfuture conditions within catchments and their as-sociated stream biota. Prior research on land usepatterns and socioeconomic drivers of those landuse patterns suggests that the LT and Cane Creekbasins are on different land-use-change trajec-tories (Wear and Bolstad 1998). One trajectorycan be described as starting from highly forestedcatchments (in the mid-twentieth century) withonly a few buildings and transitioning towardmostly forested catchments with some residentialdevelopment over the next two decades. A secondtrajectory, exemplified in the Cane Creek basin, ischaracterized by suburban development in histor-

ically agricultural catchments (Wear and Bolstad1998).

We show that a gradient of catchments—from Forested to Rural and from SuburbanizingFarmland and to Urban—exists in western NorthCarolina. Diatoms were low in species numbersat forested reference sites but increased in di-versity at sites with more open riparian cover.Macroinvertebrate and fish ordinations both sepa-rated forested catchments from agricultural ones.Suburbanizing Farmland catchments, which hadlower EPT richness, fewer highland endemicfishes, and more cosmopolitan fishes, were ecolog-ically distinct from Forested catchments. Urbancatchments supported mostly tolerant fishes andmacroinvertebrates and did not have many sensi-tive taxa.

In addition to summarizing current findingsand anticipating future biological and physicalchanges, the data presented here will contributeto retrospective analyses to be conducted in futuredecades, after time-series observations are avail-able. When low density housing is built in an oth-erwise forested setting, productivity and diversityincrease (Scott and Helfman 2001). Our analysessuggest that this transition would be marked bya sharp increase in diatom diversity. The tran-sition from agricultural to suburban and urbanland uses, however, would likely be typified by in-creases in dissolved solids and decreases in bioticintegrity (see Scott et al. 2002), such as when bothnon-native (Rahel 2000) and native (Scott andHelfman 2001) invasive fish species flourish at theexpense of native ones (also see Wang et al. 2000;Scott et al. 2002; Wang et al. 2003; Walterset al. 2003). Because much greater building den-sity is predicted in the French Broad basin, thehazard sites there (Table 1) are expected to di-verge from the other hazard sites and referencesites. Scott and Helfman (2001) have proposeda mechanism whereby habitat homogenizationwould lead to biotic homogenization, even in theabsence of non-native invasive species.

This study establishes a sampling protocol totrack biological and physical consequences forfreshwater ecosystems due to predicted land usechanges. Examining land use patterns as theychange as well as and their socioeconomic con-text within a large region provides a template for

Environ Monit Assess

understanding the relationships among landscapeposition, catchment land use, water quality, andbiological integrity of streams. In turn, knowl-edge of socioeconomic determinants of land usecan help identify consequent inputs and thereforechanges to river systems in the region (Turneret al. 1996; Grimm et al. 2000; Pickett et al. 2001).We identified a set of sites that currently liesbelow important thresholds of road density andimpervious surface area but where catchment de-velopment is likely to surpass those thresholdsover the coming decades. Land use and landcover data were used to classify all watershedsof comparable size and elevation in the regioninto categories identified through detailed streamassessments. These classes provide a snapshot ofoverall conditions in the study area and shedlight on possible future conditions in streams asthey are influenced by human land use decisions.The consequences of land use change for streamecosystem change depends on the historical con-text of streams and their watersheds. Current landuse and ecosystem states indeed reflect “the ghostof land use past” (Harding et al. 1998), but theyalso suggest the specter of what is to come.

Acknowledgements This work was conducted with sup-port from NSF grant DEB-9632854 to the University ofGeorgia Coweeta Long Term Ecological Research Pro-gram. We appreciate the thoughtful input of reviewersof an earlier version of this manuscript, including DavidStrayer.

References

Allan, J. D. (2004). Landscapes and riverscapes: The influ-ence of land use on stream ecosystems. Annual Reviewof Ecology Evolution and Systematics, 35, 257–284.doi:10.1146/annurev.ecolsys.35.120202.110122.

Berkman, H. E., & Rabeni, C. F. (1987). Effect of siltationon stream fish communities. Environmental Biology ofFishes, 18, 285–294. doi:10.1007/BF00004881.

Booth, D. B., & Jackson, R. (1997). Urbanization ofaquatic systems: Degradation thresholds, stormwaterdetection, and the limits of mitigation. Journal of theAmerican Water Resources Association, 33, 1077–1090.doi:10.1111/j.1752-1688.1997.tb04126.x.

Brant, L. A. (2003). A new species of Meridion (Bacilla-riophyceae) from western North Carolina. South-eastern Naturalist, 2, 409–418. doi:10.1656/1528-7092(2003)002[0409:ANSOMB]2.0.CO;2.

Camburn, K. E., Lowe, R. L., & Stoneburner, D. L. (1978).The haptobenthic diatom flora of Long Branch Creek,South Carolina. Nova Hedwigia, 30, 149–279.

Carpenter, S. R., Ludwig, D., & Brock, W. A. (1999).Management of eutrophication for lakes subjectto potentially irreversible change. Ecological Appli-cations, 9, 751–771. doi:10.1890/1051-0761(1999)009[0751:MOEFLS]2.0.CO;2.

Clarke, K. R. (1993). Non-parametric multivariate analy-ses of changes in community structure. AustralianJournal of Ecology, 18, 117–143. doi:10.1111/j.1442-9993.1993.tb00438.x.

Deal, J. M. (2001). Procedures for chemical analysis:Coweeta hydrologic laboratory. Retrieved from http://coweeta.ecology.uga.edu/webdocs/1/pdf/AnalyticalLabCookbook15Jan2002.pdf.

Delong, M. D., & Brusven, M. A. (1998). Macroinver-tebrate community structure along the longitudinalgradient of an agriculturally impacted stream. En-vironmental Management, 22, 445–457. doi:10.1007/s002679900118.

Desmond, J. S., Deutschman, D. H., & Zedler, J. B. (2002).Spatial and temporal variation in estuarine fish andinvertebrate assemblages: Analysis of an 11-year dataset. Estuaries, 25, 552–569

Doyle, M. W., Harbor, J. M., Rich, C. F. & Spacie, A.(2000). Examining the effects of urbanization onstreams using indicators of geomorphic stability. Phys-ical Geography, 21, 155–181.

Dufrene, M., & Legendre, P. (1997). Species assem-blages and indicator species: The need for a flexibleasymmetrical approach. Ecological Monographs, 67,345–366.

Erhman, T. P., & Lamberti, G. A. (1992). Hydraulic andparticulate matter retention in a 3rd-order Indianastream. Journal of the North American BenthologicalSociety, 11, 341–349. doi:10.2307/1467556.

Ford, J., & Rose, C. E. (2000). Characterizing small sub-basins: A case study from coastal Oregon. Envi-ronmental Monitoring And Assessment, 64, 359–377.doi:10.1023/A:1006451420945.

Forman, R. T. T., & Alexander, L. E. (1998). Roadsand their major ecological effects. Annual Reviewof Ecology and Systematics, 29, 207. doi:10.1146/annurev.ecolsys.29.1.207.

Freeman, M. C., Crawford, M. K., Barrett, J. C., Facey,D. E., Flood, M. G., Hill, J., et al. (1988). Fish as-semblage stability in a southern Appalachian stream.Canadian Journal of Fisheries and Aquatic Sciences,45, 1948–1958.

Gardiner, E. P. (2002). Geospatial techniques for streamresearch in the southern Blue Ridge Mountains. Dis-sertation, University of Georgia.

Gergel, S. E., Turner, M. G., Miller, J. R., Melack, J. M., &Stanley, E. H. (2002). Landscape indicators of humanimpacts to riverine systems. Aquatic Sciences, 64, 118–128. doi:10.1007/s00027-002-8060-2.

http://dx.doi.org/10.1146/annurev.ecolsys.35.120202.110122http://dx.doi.org/10.1007/BF00004881http://dx.doi.org/10.1111/j.1752-1688.1997.tb04126.xhttp://dx.doi.org/10.1656/1528-7092(2003)002[0409:ANSOMB]2.0.CO;2http://dx.doi.org/10.1656/1528-7092(2003)002[0409:ANSOMB]2.0.CO;2http://dx.doi.org/10.1890/1051-0761(1999)009[0751:MOEFLS]2.0.CO;2http://dx.doi.org/10.1890/1051-0761(1999)009[0751:MOEFLS]2.0.CO;2http://dx.doi.org/10.1111/j.1442-9993.1993.tb00438.xhttp://dx.doi.org/10.1111/j.1442-9993.1993.tb00438.xhttp://coweeta.ecology.uga.edu/webdocs/1/pdf/AnalyticalLabCookbook15Jan2002.pdfhttp://coweeta.ecology.uga.edu/webdocs/1/pdf/AnalyticalLabCookbook15Jan2002.pdfhttp://coweeta.ecology.uga.edu/webdocs/1/pdf/AnalyticalLabCookbook15Jan2002.pdfhttp://dx.doi.org/10.1007/s002679900118http://dx.doi.org/10.1007/s002679900118http://dx.doi.org/10.2307/1467556http://dx.doi.org/10.1023/A:1006451420945http://dx.doi.org/10.1146/annurev.ecolsys.29.1.207http://dx.doi.org/10.1146/annurev.ecolsys.29.1.207http://dx.doi.org/10.1007/s00027-002-8060-2

Environ Monit Assess

Gordon, N. D., McMahon, T. A., & Finlayson, B. L. (1992).Stream hydrology: An introduction for ecologists. NewYork: Wiley.

Greenwood, J. L., & Rosemond, A. D. (2005). Periphytonresponse to long-term nutrient enrichment in a shadedheadwater stream. Canadian Journal of Fisheries andAquatic Sciences, 62, 2033–2045. doi:10.1139/f05-117.

Greig-Smith, P. (1983). Quantitative plant ecology.Berkeley: University of California Press.

Grimm, N. B., Grove, J. M., Pickett, S. T. A., &Redman, C. L. (2000). Integrated approaches tolong-term studies of urban ecological systems. Bio-Science, 50, 571–584. doi:10.1641/0006-3568(2000)050[0571:IATLTO]2.0.CO;2.

Hampson, P. S., Treece, J. M. W., Johnson, G. C., Ahlstedt,S. A., & Connell, J. F. (2000). Water quality inthe upper Tennessee River Basin, Tennessee, NorthCarolina, Virginia, and Georgia 1994–98. Circular1205. Washington: USGS.

Harding, J. S., Benfield, E. F., Bolstad, P. V., Helfman,G. S., & Jones III, E. B. D. (1998). Stream biodiversity:The ghost of land use past. Proceedings of the NationalAcademy of Sciences of the United States of America,95, 14843–14847. doi:10.1073/pnas.95.25.14843.

Hawkins, C. P., Norris, R. H., Gerritsen, J., Hughes, R. M.,Jackson, S. K., Johnson, R. K., et al. (2000). Evaluationof the use of landscape classifications for the predic-tion of freshwater biota: Synthesis and recommenda-tions. Journal of the North American BenthologicalSociety, 19, 541–556. doi:10.2307/1468113.

Hawkins, C. P., & Vinson, M. R. (2000). Weak correspon-dence between landscape classifications and streaminvertebrate assemblages: Implications for bioassess-ment. Journal of The North American BenthologicalSociety, 19, 501–517. doi:10.2307/1468111.

Heino, J., Muotka, T., Mykra, H., Paavola, R.,Hamalainen, H., & Koskenniemi, E. (2003). Definingmacroinvertebrate assemblage types of headwaterstreams: Implications for bioassessment and conser-vation. Ecological Applications, 13, 842–852. doi:10.1890/1051-0761(2003)013[0842:DMATOH]2.0.CO;2.

Ilmonen, J., & Paasivirta, L. (2005). Benthic macrocrus-tacean and insect assemblages in relation to springhabitat characteristics: Patterns in abundance anddiversity. Hydrobiologia, 533, 99–113. doi:10.1007/s10750-004-2399-4.

Jenkins, R. E., & Burkhead, N. M. (1994). Freshwater fishesof Virginia. Bethesda: American Fisheries Society.

Jones, E. B. D., Helfman, G. S., Harper, J. O., & Bolstad,P. V. (1999). The effects of riparian forest removalon fish assemblages in southern Appalachian streams.Conservation Biology, 13, 1454–1465. doi:10.1046/j.1523-1739.1999.98172.x.

King, R. S., Baker, M. E., Whigham, D. F., Weller, D.E., Jordan, T. E., Kazyak, P. F., et al. (2005). Spatialconsiderations for linking watershed land cover to eco-logical indicators in streams. Ecological Applications,15, 137–153. doi:10.1890/04-0481.

Kociolek, J. P., & Kingston, J. C. (1999). Taxonomy, ultra-structure, and distribution of some gomphonemoid di-atoms (Bacillariophyceae: Gomphonemataceae) from

rivers in the United States. Canadian Journal ofBotany, 77, 686–705. doi:10.1139/cjb-77-5-686.

Kruskall, J. B., & Wish, M. (1978). Multidimensionalscaling. Beverly Hills: Sage.

Laacke, J. (2000). Life form (software), alpha release0.14. Pacific southwest experiment station. California:Redding.

Lenat, D. R., & Crawford, J. K. (1994). Effects of landuse on water quality and aquatic biota of three NorthCarolina Piedmont streams. Hydrobiologia, 294, 185–199. doi:10.1007/BF00021291.

Loeb, S. L. (1981). An in situ method for measuring theprimary productivity and standing crop of the epilithicperiphyton community in lentic systems. Limnologyand Oceanography, 26, 394–399.

Lowe, R. L., Golladay, S. W., & Webster, J. R. (1986).Periphyton response to nutrient manipulation instreams draining clearcut and forested watersheds.Journal of the North American Benthological Society,5, 221–229. doi:10.2307/1467709.

Lowe, R. L., & LaLiberte, G. D. (1996). Benthic streamalgae: Distribution and structure. In F.R. Hauer &G. A. Lamberti (Eds.), Methods in stream ecology(pp. 269–293). New York: Academic.

Mathsoft (1999). S-Plus 2000 Professional Release 1.Lucent Technologies, Inc.

McCormick, F. H., Peck, D. V., & Larsen, D. P. (2000).Comparison of geographic classification schemes forMid-Atlantic stream fish assemblages. Journal of theNorth American Benthological Society, 19, 385–404.doi:10.2307/1468102.

McCune, B., & Mefford, M. J. (1999). PC-ORD v. 4.10.Multivariate analysis of ecological data. GlenedenBeach: MjM Software.

McRae, G., Camp, D. K., Lyons, W. G., & Dix, T. L.(1998). Relating benthic infaunal community struc-ture to environmental variables in estuaries usingnonmetric multidimensional scaling and similarityanalysis. Environmental Monitoring And Assessment,51, 233–246. doi:10.1023/A:1005943504335.

Mosisch, T. D., Bunn, S. E., & Davies, P. M. (2001). Therelative importance of shading and nutrients on algalproduction in subtropical streams. Freshwater Biology,46, 1269–1278. doi:10.1046/j.1365-2427.2001.00747.x.

Mykra, H., Heino, J., & Muotka, T. (2004). Variability oflotic macroinvertebrate assemblages and stream habi-tat characteristics across hierarchical landscape clas-sifications. Environmental Management, 34, 341–352.doi:10.1007/s00267-004-0236-1.

Paul, M. J., & Meyer, J. L. (2001). Streams in the ur-ban landscape. Annual Review of Ecology and Sys-tematics, 32, 333–365. doi:10.1146/annurev.ecolsys.32.081501.114040.

Pickett, S. T. A., Cadenasso, M. L., Grove, J. M., Nilon,C. H., Pouyat, R. V., Zipperer, W. C., & Costanza, R.(2001). Urban ecological systems: Linking terrestrialecological physical, and socioeconomic componentsof metropolitan areas. Annual Review of Ecologyand Systematics, 32, 127–157. doi:10.1146/annurev.ecolsys.32.081501.114012.

http://dx.doi.org/10.1139/f05-117http://dx.doi.org/10.1641/0006-3568(2000)050[0571:IATLTO]2.0.CO;2http://dx.doi.org/10.1641/0006-3568(2000)050[0571:IATLTO]2.0.CO;2http://dx.doi.org/10.1073/pnas.95.25.14843http://dx.doi.org/10.2307/1468113http://dx.doi.org/10.2307/1468111http://dx.doi.org/10.1890/1051-0761(2003)013[0842:DMATOH]2.0.CO;2http://dx.doi.org/10.1890/1051-0761(2003)013[0842:DMATOH]2.0.CO;2http://dx.doi.org/10.1007/s10750-004-2399-4http://dx.doi.org/10.1007/s10750-004-2399-4http://dx.doi.org/10.1046/j.1523-1739.1999.98172.xhttp://dx.doi.org/10.1046/j.1523-1739.1999.98172.xhttp://dx.doi.org/10.1890/04-0481http://dx.doi.org/10.1139/cjb-77-5-686http://dx.doi.org/10.1007/BF00021291http://dx.doi.org/10.2307/1467709http://dx.doi.org/10.2307/1468102http://dx.doi.org/10.1023/A:1005943504335http://dx.doi.org/10.1046/j.1365-2427.2001.00747.xhttp://dx.doi.org/10.1007/s00267-004-0236-1http://dx.doi.org/10.1146/annurev.ecolsys.32.081501.114040http://dx.doi.org/10.1146/annurev.ecolsys.32.081501.114040http://dx.doi.org/10.1146/annurev.ecolsys.32.081501.114012http://dx.doi.org/10.1146/annurev.ecolsys.32.081501.114012

Environ Monit Assess

Poff, N. L., Allan, J. D., Bain, M. B., Karr, J. R.,Prestegaard, K. L., Richter, B., et al. (1997). The nat-ural flow regime: A new paradigm for riverine con-servation and restoration. BioScience, 47, 769–784.doi:10.2307/1313099.

Potapova, M., & Charles, D. F. (2007). Diatom metricsfor monitoring eutrophication in rivers of the UnitedStates. Ecological Indicators, 7, 48–70. doi:10.1016/j.ecolind.2005.10.001.

Rahel, F. J. (2000). Homogenization of fish faunasacross the United States. Science, 288, 854–856.doi:10.1126/science.288.5467.854.

Rosemond, A. D. (1993). Interactions among irradi-ance, nutrients, and herbivores constrain a streamalgal community. Oecologia, 94, 585–594. doi:10.1007/BF00566976.

Roy, A. H., Rosemond, A. D., Paul, M. J., Leigh, D. S.,& Wallace, J. B. (2003). Stream macroinvertebrate re-sponse to catchment urbanisation (Georgia, U.S.A.).Freshwater Biology, 48, 329–346. doi:10.1046/j.1365-2427.2003.00979.x.

Scott, M. C. (2001). Integrating the stream and its val-ley: Land use change, aquatic habitat, and fish assem-blages. Dissertation, University of Georgia.

Scott, M. C., & Helfman, G. S. (2001). Native invasions,homogenization, and the mismeasure of integrityof fish assemblages. Fisheries, 26, 6–15. doi:10.1577/1548-8446(2001)0262.0.CO;2.

Scott, M. C., Helfman, G. S., McTammany, M. E., Benfield,E. F., & Bolstad, P. V. (2002). Multiscale influenceson physical and chemical stream conditions acrossBlue Ridge landscapes. Water Resources Bulletin, 38,1379–1392.

Southwood, T. R. E. (1977). Habitat, the templet forecological strategies? Journal of Animal Ecology, 46,337–365. doi:10.2307/3817.

Strange, E. M., Fausch, K. D., & Covich, A. P. (1999). Sus-taining ecosystem services in human-dominated wa-tersheds: Biohydrology and ecosystem processes inthe South Platte River Basin. Environmental Manage-ment, 24, 39–54. doi:10.1007/s002679900213.

Sutherland, A. B., Meyer, J. L., & Gardiner, E. P.(2002). Effects of land cover on sediment regimeand fish assemblage structure in four southernAppalachian streams. Freshwater Biology, 47, 1791–1805. doi:10.1046/j.1365-2427.2002.00927.x.

Turner, M. G., Wear, D. N., & Flamm, R. O. (1996).Land ownership and land-cover change in thesouthern Appalachian highlands and the Olympicpeninsula. Ecological Applications, 6, 1150–1172.doi:10.2307/2269599.

Wallace, J. B., Webster, J. R., & Lowe, R. L. (1992).High-gradient streams of the Appalachians. In C. T.Hackney, S. M. Adams, & W. A. Martin (Eds.), Biodi-versity of the Southeastern United States/ aquatic com-munities (pp. 131–191). New York: Wiley.

Walters, D. M., Leigh, D. S., & Bearden, A. B. (2003). Ur-banization, sedimentation, and the homogenization offish assemblages in the Etowah River Basin, USA. Hy-drobiologia, 494, 5–10. doi:10.1023/A:1025412804074.

Wang, L., Lyons, J., Kanehl, P., Bannerman, R., &Emmons, E. (2000). Watershed urbanization andchanges in fish communities in southeastern Wis-consin streams. Journal of the American Water Re-sources Association, 36, 1173–1189. doi:10.1111/j.1752-1688.2000.tb05719.x.

Wang, L. Z., Lyons, J., Kanehl, P., & Gatti, R. (1997).Influences of watershed land use on habitat qualityand biotic integrity in Wisconsin streams. Fisheries,22, 6–12. doi:10.1577/1548-8446(1997)0222.0.CO;2.

Wang, L. Z., Lyons, J., & Kanehl, P. (2003). Impactsof urban land cover on trout streams in Wisconsinand Minnesota. Transactions of the American FisheriesSociety, 132, 825–839. doi:10.1577/T02-099.

Waters, T. F. (1995). Sediment in streams: Sources, biolog-ical effects, and control. American Fisheries SocietyMonograph 7.

Wear, D. N., & Bolstad, P. V. (1998). Land use changesin Southern Appalachian landscapes: Spatial analy-sis and forecast evaluation. Ecosystems, 1, 575–594.doi:10.1007/s100219900052.

Wetzel, R. G., & Likens, G. E. (1991). Limmnologicalanalyses. New York: Springer.

Wilson, M. A., & Carpenter, S. R. (1999). Economic val-uation of freshwater ecosystem services in the UnitedStates: 1971–1997. Ecological Applications, 9, 772–783.

Wishart, D. (1969). An algorithm for hierarchical classifi-cations. Biometrics, 25, 165–170. doi:10.2307/2528688.

Wood, P. J., & Armitage, P. D. (1997). Biological effects offine sediment in the lotic environment. EnvironmentalManagement, 21, 203–217. doi:10.1007/s002679900019.

http://dx.doi.org/10.2307/1313099http://dx.doi.org/10.1016/j.ecolind.2005.10.001http://dx.doi.org/10.1016/j.ecolind.2005.10.001http://dx.doi.org/10.1126/science.288.5467.854http://dx.doi.org/10.1007/BF00566976http://dx.doi.org/10.1007/BF00566976http://dx.doi.org/10.1046/j.1365-2427.2003.00979.xhttp://dx.doi.org/10.1046/j.1365-2427.2003.00979.xhttp://dx.doi.org/10.1577/1548-8446(2001)0262.0.CO;2http://dx.doi.org/10.1577/1548-8446(2001)0262.0.CO;2http://dx.doi.org/10.2307/3817http://dx.doi.org/10.1007/s002679900213http://dx.doi.org/10.1046/j.1365-2427.2002.00927.xhttp://dx.doi.org/10.2307/2269599http://dx.doi.org/10.1023/A:1025412804074http://dx.doi.org/10.1111/j.1752-1688.2000.tb05719.xhttp://dx.doi.org/10.1111/j.1752-1688.2000.tb05719.xhttp://dx.doi.org/10.1577/1548-8446(1997)0222.0.CO;2http://dx.doi.org/10.1577/1548-8446(1997)0222.0.CO;2http://dx.doi.org/10.1577/T02-099http://dx.doi.org/10.1007/s100219900052http://dx.doi.org/10.2307/2528688http://dx.doi.org/10.1007/s002679900019

Linking stream and landscape trajectories in the southern AppalachiansAbstractIntroductionStudy area and methodsSite selectionEcosystem assessmentsBiological assessmentsPhysical assessmentsStatistical analyses

Inferring landscape and stream trajectories

ResultsAssemblage ordinationsLandscape-level inferences

DiscussionConclusionReferences

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /Warning /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 150 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 600 /MonoImageMinResolutionPolicy /Warning /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 600 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/Description >>> setdistillerparams> setpagedevice