Rapid geomorphic and habitat stream assessment techniques inform restoration differently based on levels of stream disturbance

RAPID GEOMORPHIC AND HABITAT STREAM ASSESSMENT TECHNIQUES INFORM

RESTORATION DIFFERENTLY BASED ON LEVELS OF STREAM DISTURBANCE1

Michael W. Habberfield, Stacey Sloan Blersch, Sean J. Bennett, and Joseph F. Atkinson2

ABSTRACT: Visual-based rapid assessment techniques provide an efficient method for characterizing the resto-ration potential of streams, with many focusing on channel stability and instream habitat features. Few studies,however, have compared these techniques to see if they result in differing restoration priorities. Three rapidassessment techniques were contrasted at three wild trout streams in western New York with different amountsof channel disturbance. Two methods focused only on geomorphic stability, whereas the third addressed physicalhabitat condition. Habitat assessment scores were not correlated with scores for either geomorphic assessmentmethod and they varied more between channels with different degrees of disturbance. A model based ondynamic equilibrium concepts best explains the variation among the streams and techniques because it accountsfor a stream’s capacity to maintain ecological integrity despite some inherent instability. Geomorphic indicescan serve as effective proxies for biological indices in highly disturbed systems. Yet, this may not be the case inless disturbed systems, where geomorphic indices cannot differentiate channel adjustments that impact biotafrom those that do not. Dynamically stable streams can include both stable and unstable reaches locally as char-acterized by geomorphic methods and translating these results into restoration priorities may not be appropriateif interpretations are limited to the reach scale.

(KEY TERMS: biotic integrity; rivers/streams; fluvial processes; monitoring; restoration; watershed manage-ment; geomorphology.)

Habberfield, Michael W., Stacey Sloan Blersch, Sean J. Bennett, and Joseph F. Atkinson, 2014. Rapid Geomorphicand Habitat Stream Assessment Techniques Inform Restoration Differently Based on Levels of Stream Disturbance.Journal of the American Water Resources Association (JAWRA) 50(4): 1051-1062. DOI: 10.1111/jawr.12156

INTRODUCTION

Stream restoration, or the action of reestablishingthe structure and function of a degraded stream eco-system to its remaining natural potential, remains anextremely active area for both researchers and practi-tioners, where the number of project records andscientific publications focused on restoration hasincreased exponentially in the last few decades (Bern-

hardt et al., 2005; Hassett et al., 2005; Bernhardtand Palmer, 2011). This was due, in part, by pressurefrom Congress that shifted funding in Clean WaterAct programs away from monitoring and towardimplementation (USEPA, 1998). Methods to identifypreexisting conditions and to prioritize projectswithin stream corridors in an efficient manner weretherefore needed, especially in streams where exist-ing data were sparse or nonexistent. Some of themost common and cost-effective methods used to

1Paper No. JAWRA-13-0003-P of the Journal of the American Water Resources Association (JAWRA). Received January 3, 2013; acceptedNovember 26, 2013. © 2014 American Water Resources Association. Discussions are open until six months from print publication.

2Ph.D. Candidate (Habberfield) and Professor (Bennett), Department of Geography; Ph.D. Candidate (Blersch) and Professor (Atkinson),Department of Civil, Structural, and Environmental Engineering, State University of New York at Buffalo, 202 Jarvis Hall, Buffalo,New York 14260 (E-Mail/Habberfield: [email protected]).

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION JAWRA1051

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

Vol. 50, No. 4 AMERICAN WATER RESOURCES ASSOCIATION August 2014

identify and prioritize stream restoration sites arevisual-based rapid assessment (VBRA) techniquesthat evaluate the quality of the physical conditions ofthe stream (Barbour et al., 1999; Ward et al., 2003;Hughes et al., 2010).

Despite their common usage, few studies havecompared rapid assessment techniques to determineif they result in different conclusions about instreamconditions (Ward et al., 2003; Hughes et al., 2010). Ifdifferent outcomes do occur among VBRA techniqueswhen applied at the same location, selecting anappropriate technique is imperative to ensure thatrestoration is occurring at sites that are both physi-cally and ecologically impaired. The challenge isselecting a technique that is representative of thestream under most conditions, despite the range ofcomplex interactions that occur in a stream ecosys-tem spatially and temporally (Odum, 1994; Allan andCastillo, 2007; Sullivan, 2012). For stream restorationpurposes, the most commonly cited design flow is thechannel-forming discharge that in stable channels isconsidered equivalent to bankfull flow or, on average,the 1.5- to 2-year storm event (Julien, 2002; Shieldset al., 2003). The bankfull flow corresponds spatiallyto changes at the reach scale (Frissell et al., 1986;but see Doyle et al., 2005 for how this might changethroughout a watershed). This study, therefore,focuses on assessment techniques conducted at thereach scale.

Rapid stream assessment techniques emphasizegeomorphic characteristics (e.g., Pfankuch, 1975;Harrelson et al., 1994; Simon and Downs, 1995;Rosgen, 1996) as well as habitat characteristics andbiological potential (Barbour et al., 1999; Sullivan,2012). This emphasis corresponds to the current dom-inant paradigm in stream restoration, where creatingand increasing stable habitat features are the pri-mary design objectives (Hey, 1988; Palmer et al.,2010). Through the use of instream redirective struc-tures, the physical and hydraulic conditions of astream can be manipulated to address geomorphicinstability (i.e., excessive bank erosion, head cuts,widening, entrenchment) and to enhance or recreatespecific habitat structures such as riffle-poolsequences (Simon and Downs, 1995; Bernhardt et al.,2005; Kondolf et al., 2007). Ideally, restoration pro-jects should result in a stream in dynamic equili-brium (Lane, 1955; Rosgen, 2001), providing acontinually changing mosaic of habitat types (Flors-heim et al., 2008). Yet, many restored streams arerendered to a state of perpetual stability and are nolonger able to adjust freely geomorphically (Simonet al., 2007). The underlying ecological processes,such as sediment transport, stream metabolism, andnutrient spiraling, may not be compatible with thepost-restoration stable channel conditions (Darby and

Sear, 2008). Therefore, if VBRAs are used to guidethe selection of restoration projects, it would seemreasonable to question the emphasis placed on chan-nel stability.

To address this issue, VBRA techniques wereapplied and contrasted at three different streams inwestern New York. Sites supporting wild trout wereselected based on perceived drivers of disturbancedue to land-use patterns in each watershed. Threedifferent VBRA techniques were conducted over sev-eral reaches at each stream; two focusing on geomor-phic stability (rapid geomorphic assessments, orRGAs), and a third primarily focused on physicalhabitat characteristics (the U.S. EPA’s 1999 RapidBioassessment Protocol, or RBP). The RBP attemptsto evaluate ecological integrity as the combinedwater quality, habitat, and biological conditionswithin the natural range of variability in an undis-turbed stream (Barbour et al., 1999). The habitatportion used here serves as one component of mea-suring overall ecological integrity and parallels simi-lar models (e.g., Index of Biotic Integrity) (Karr andChu, 2000). It was hypothesized that the two geo-morphic assessments would correlate strongly witheach other, but that neither would correlate with thehabitat assessment. This is because the two RGAsemphasize geomorphic stability, whereas the RBPemphasizes the mosaic of physical habitat and result-ing biological potential that may be derived from nat-ural stream instability (Florsheim et al., 2008).Because the relationship between geomorphic adjust-ments and changes in physical habitat is complex(Allan, 1995; Sullivan et al., 2004), and can varywithin different contexts (Sullivan, 2012), it sug-gested here that varying the levels of disturbancemay also demonstrate differential geomorphic andhabitat conditions.

Three models were tested to best explain the varia-tions among streams and techniques. Two simplemodels, each consisting of a single factor (assessmenttechnique factor and stream factor), correspond tohypotheses suggesting that score variation can beattributable to its respective factor. These simplemodels were compared to a more complex modelbased on the concept of dynamic equilibrium. Thismodel hypothesizes that all streams, regardless ofdisturbance level, will have a mix of geomorphicallystable and unstable reaches due to the dynamics ofequilibrium adjustments. Highly disturbed streamsare very likely to exhibit instability, but the conceptof dynamic equilibrium does not preclude undisturbedor stable streams from also exhibiting instability inactively adjusting reaches. This means that rapidgeomorphic assessment scores may remain similarbetween more and less disturbed systems, while habi-tat assessment scores should respond inversely to

JAWRA JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION1052

HABBERFIELD, BLERSCH, BENNETT, AND ATKINSON

disturbance level. Following from this, a relatedhypothesis is that only relatively undisturbedstreams will show differences between the scores ofthe assessment techniques. This is because activelyadjusting reaches will lower the geomorphic assess-ment scores while habitat assessment scores willapproach their maximum potential. Accordingly, weexpect that habitat assessments will evaluate streamecological integrity independent of geomorphic stabil-ity. That is, streams with high ecological integritymay still exhibit geomorphic instability.

The objective of this study was to compare differ-ent VBRA techniques to determine if these can detecta difference in stream ecological integrity. By choos-ing techniques that emphasize various aspects of thestream ecosystem (geomorphic vs. habitat), this studyalso investigates the ability of these techniques todetermine existing conditions in the context ofdynamic equilibrium. A difference among streams forany of the techniques would support the a priori des-ignation of the three streams having varying distur-bance levels. How that disturbance is characterized,however, depends on which of the techniques areshown to vary across the streams. If all techniquesare different among streams, the idea that geomor-phic and biological methods both characterize streamdisturbance similarly is supported. If the habitatassessment, however, shows a difference amongstreams while the geomorphic methods do not, the

hypothesis predicting that ecological integrity can beindependent of geomorphic stability is supported.Such results would have important implications forhow VBRAs should be used in restoration planning.

METHODS

Study Sites

Three streams were selected for this analysis.These represent a heavily disturbed stream (EltonCreek), a less disturbed reference stream within thesame watershed (Clear Creek), and an undisturbedreference stream within a different watershed (Mc-Intosh Creek). The degree of disturbance is basedon land-cover data and perceived anthropogenic driv-ers from previous site visits. All three creeks arelocated in Cattaraugus County, New York (Figure 1).Climate in Cattaraugus County is humid continentalwith an average annual temperature of 6.9°C andaverage annual precipitation of 1,236 mm. The regionis part of the Appalachian Plateau physiographicprovince resulting in a relatively complex topographycovered primarily by glacial deposits, and a geomor-phology heavily influenced by the underlying bed-rock.

FIGURE 1. Map of the Study Region Showing the Three Streams Investigated—Elton, Clear, andMcIntosh Creeks—and Their Broader Watershed Settings in Cattaraugus County, New York.


RAPID GEOMORPHIC AND HABITAT STREAM ASSESSMENT TECHNIQUES INFORM RESTORATION DIFFERENTLY BASED ON LEVELS OF STREAM DISTURBANCE

The selected reaches of Elton Creek and ClearCreek lie in the middle portions of each creek’s respec-tive subwatershed and represent third-order streams.The general soil types in the area are deep, well-drained gravelly silt loams. Cultivated crops andpastureland contribute significantly to the land coverin both watersheds, and no land-cover differences existat the scale of U.S. Geological Survey watershed units(Table 1). Elton Creek, however, has minimal riparianforest, whereas Clear Creek has substantial riparianforest cover in many areas. The Elton Creek reachesare adjacent to intensive gravel mining operations andhave been subjected to channelization in the past. TheClear Creek reaches do not have a legacy of intensedisturbance and are primarily impacted by agriculturewithin the watershed. McIntosh Creek is a first-orderstream in the Allegheny River watershed. It is locatedwithin Allegany State Park and has a primarilyforested watershed. Soils around this site are silty clayloams with numerous boulders. In 2008, a habitatimprovement project was completed on McIntoshCreek that increased the amount of deep pool habitatfor native trout.

Visual-Based Rapid Assessments

The following VBRA techniques were applied atfour reaches within each stream: RPB (Barbour et al.,1999), Channel Stability Ranking Scheme (CSRS)(Simon and Downs, 1995), and the Pfankuch ChannelStability Evaluation Procedure (PCSEP) (Pfankuch,1975; as updated by Rosgen, 1996). The criteriaassessed with each technique are summarized inFigure 2. Reaches were 232 � 76 m in length(mean � standard deviation) and were specified basedon the guidance provided within the assessment formsand onsite geomorphological controls (e.g., reachesshould be long enough to cover two riffle-poolsequences, in the range of 6-20 channel widths) (Simonand Downs, 1995). One researcher conducted all of theRGA assessments, whereas another conducted all ofthe RBP assessments so as to control any correlationbias. Each researcher was experienced in their respec-tive method. Stream reaches were walked in theirentirety and visually assessed according to the criteriaof each assessment method. Evaluation was donethroughout each reach length with an attempt todetermine an average condition for that reach. TheElton Creek, Clear Creek, and McIntosh Creeksurveys were completed in April 2009, October 2009,and October 2010, respectively.

Rapid Bioassessment Protocol. The quality ofphysical habitat in each creek was assessed using thevisual-based habitat assessment approach outlined by

TABLE

1.Geo

morphic

Characteristics

oftheThreeSurvey

edStrea

msandtheCorresp

ondingWatershed

Disturbance

Fea

tures.

Geomorphic

Characteristics

LevelofDistu

rbance

Stream

Order

Slope

(mm

�1)

Sin

uosity

(SL/V

L)

BedType

(major/m

inor)

USGS

Watersh

edUnit

LandCover(%

)Anth

ropogenic

Influence

Forest

Pasture/

Crop

Developed

Oth

er

Domin

ant

Drivers

Categorization

Elton

Creek

3rd

0.009

1.79

Cob

ble/bou

lder

51.6

38.2

4.7

5.5

Gravel

mining/

channelization

Hea

vily

disturbed

ClearCreek

3rd

0.006

1.20

Cob

ble/bou

lder

51.1

39.0

4.1

5.8

Agricu

lture

Marginally

disturbed

McIntosh

Creek

1st

0.046

1.11

Bou

lder/cob

ble

93.3

0.6

1.5

4.6

State

park

recrea

tion

Undisturbed

Note:

SL/VL,stream

length/valley

length;USGS,U.S.Geo

logicalSurvey

.



Barbour et al. (1999) as part of their broader RBP.This assessment is designed to accompany biologicalsurveys conducted on streams, but the visual-basedapproach provided the only data gathered on the eco-logical state of the stream during this study. Ten cri-teria are used to quantify habitat quality (Figure 2).These criteria combine both micro- and macroscalehabitat features for a full assessment of habitatstructure (Barbour et al., 1999). For example, avail-able cover and embeddedness serve to representniche variety, refugia, and spawning sites for fishesand macroinvertebrates. The channel alteration andriffle/bend frequency criteria provide a larger scaleinfluence of stream character on habitat. Collectively,the assessment emphasizes habitat diversity and theability of the system to recover after disturbance. Thevisual assessment for high gradient (riffle/run preva-lent) streams was used. A low gradient (glide/poolprevalent) assessment also is provided by Barbouret al. (1999) and is very similar in its habitat charac-terization, but accounts for the inherent differencesin form such as sediment size and sinuosity.

Each criterion is scored on a scale of 0-20, withhigher scores representing higher quality habitat,which can be compared to a regional reference condi-tion. For any single survey, a maximum score of 200is possible. It is worth noting that the nominal desig-nations of some criteria are counterintuitive. Forexample, embeddedness and sediment deposition willscore highly when there is a lack of these conditions.

Simon-Downs Channel Stability RankingScheme. The CSRS is part of a modular interdisci-plinary approach to determine the potential, magni-tude, and distribution of channel instability at awatershed scale (Simon and Downs, 1995). This studywas concerned only with a single module: ranking ofrelative channel stability. Nine criteria are used to

provide insight into channel evolution processes andrelative stability. A unique feature of this assessmentform is the channel evolution framework criterion,based on earlier work by Simon and Hupp (1986) thatproposes six stages of channel evolution for predictingfuture potential for instability. While channel form isthe main indicator, the criteria are designed to provideinsight into riparian vegetation changes, instreamhabitat conditions, and sediment transport processes.

Each criterion has a possible score of 4, with atotal possible score of 36 for each reach. Upon com-pletion of the form, a channel stability index is deter-mined, indicating the relative stability of a particularstream reach. Interpretation of the final score isbased on a relative scale. As such, scores represent arelative framework upon which reaches with similarsoils and hydrologic conditions can be ranked. A scoreof 10 or less is representative of a stable reach, scoresbetween 11 and 19 somewhat stable, and scoreshigher than 20 are indicative of unstable reaches(Simon and Klimetz, 2008).

Pfankuch Channel Stability Evaluation Pro-cedure. The PCSEP was developed for the U.S.Forest Service for assessing channel stability inmountain stream channels (Pfankuch, 1975) and waslater adopted by Rosgen (1996) as part of the naturalchannel design process. The variables included in theassessment were selected to provide an indication ofthe dynamic equilibrium of the stream in terms ofchannel adjustment and sediment transport as aresult of changes in the hydraulic forces at work. Dif-ferent classifications for upper, lower, and bottom ofbank are used to help frame the types of variablesevaluated. The upper bank features provide an indi-cation of the valley type and stream access to thefloodplain, as well as the condition of the floodplain(vegetated/nonvegetated). The lower bank indicators

CSRS PCSEP RBP

• Primary bed material

Upp

er

• Landform slope • Epifaunal substrate/available cover• Bed/bank protection • Mass wasting • Embeddedness• Degree of incision • Debris jam potential • Velocity/depth regime• Degree of constriction • Vegetative bank protection • Sediment deposition• Stream bank erosion • Channel flow status• Stream bank instability

Lower

• Channel capacity/enlargement • Channel alteration• Established riparian woody vegetation cover • Bank rock content • Frequency of riffles (or bends)• Occurrence of bank accretion • Obstructions to flow • Bank stability• Stage of evolution • Cutting • Vegetative protection

• Deposition • Riparian vegetative zone width

Boom

• Rock angularity• Brightness• Consolidation• Size distribution• Scouring/deposition• Aquatic vegetation

FIGURE 2. Criteria Assessed by Each Method. CSRS is the Channel Stability Ranking Scheme (Simon and Downs, 1995).PCSEP is the Pfankuch Channel Stability Evaluation Procedure (Pfankuch, 1975), and RBP is the Rapid Bioassessment Protocol

(Barbour et al., 1999). The PCSEP categorizes its criteria into three groups: upper banks, lower banks, and stream bottom.



focus on the effects of bankfull flow and are good pre-dictors of potential instability. The bottom channelindicators assist in determining if the stream isaggrading or degrading by ranking deposition, scour,and bed material size.

The PCSEP consists of 15 criteria, with maximumvalues ranging from 4 to 24, for a total possible scoreof 152. Scores for the PCSEP are categorized intoranges: excellent, <38; good, 39-70; fair, 77-114; poor,>115. The purpose of these categories is to providemanagers with information on the potential forstreams to handle significant changes in discharge,based on the level of stability. By ranking individualreaches, problematic or unstable reaches can be iden-tified, and management measures can be taken tominimize further destabilization of those reaches.

Analyses

To facilitate interpretation between survey tech-niques, all scores were normalized to the same scale.Thus, all individual reach scores were converted tothe proportion of their maximum possible score, andthe CSRS and PCSEP proportional scores then weresubtracted from 1 so that higher scores correspond tomore stable streams. Geomorphic and habitat pat-terns were compared by calculating partial correla-tions of each pair of assessment techniques across all12 stream reaches (i.e., three correlations with n = 12for each). Partial correlations between techniquetypes control for the third variable, which of thethree streams the score came from, by using thestreams as dummy variables. One-tailed t-tests wereperformed on each of the partial correlation coeffi-cients (q) corresponding to each pairing of the threetechniques and these were assessed at an alpha levelof 0.05 to determine statistical significance.

Differences between the three streams and threeassessment techniques were further examined usingan analysis of variance (ANOVA). Following themethods of Dixon (2003), likelihood ratios for compet-ing models corresponding to the hypotheses describedabove were calculated. The likelihood ratio is:

k ¼ Model 1 unexplained variation

Model 2 unexplained variation

� �n2 ð1Þ

where n is the total number of observations. The unex-plained variation in each ANOVA model is the sum ofsquares not attributable to the components of thatmodel. Likelihood ratios express how many times morelikely the data are given the best fit of one model com-pared to the best fit of the other model (Dixon, 2003).As k deviates from 1, evidence grows that the model

with less unexplained variation is superior to the othermodel. Moderate to strong evidence that this model issuperior to the other is indicated by a k value of around10 (or 0.1 if the superior model is in the numerator),while ratios that substantially exceed 10 indicate clearevidence that this model is superior to the other (Good-man and Royall, 1988; Dixon and O’Reilly, 1999). Thisgraded evidence approach is more conducive to under-standing the implications of stream assessment meth-ods than dichotomous statistical analyses that offersimplified conclusions about which factors explain dif-ferences among streams. How the likelihood ratios areinterpreted is based on the particular models in ques-tion and must be tied to clear theoretical understand-ings of stream morphology and ecology.

Statistical models representing each of the two sin-gle factors, assessment technique and stream, and theinteraction between them (stream 9 technique), werecompared. This consisted of pooling scores from allstreams for the single-factor model representing tech-nique differences, and pooling scores from all tech-niques for the single-factor model representing streamdifferences. These models were also compared to theadditional model that is based on the dynamic equilib-rium concept and hypothesizes that streams consist ofsome stable and some unstable reaches regardless ofecological integrity. This model predicts that RGAs arenot able to differentiate streams with different levelsof disturbance, whereas the RBP is able to do so. Whiledisturbed streams are likely to exhibit consistentinstability, undisturbed streams may also have someunstable reaches due to equilibrium dynamics, thuslowering the average RGA score of the stream. Themodel representing this hypothesis is a specific inter-action between the stream and technique factors andcan be built using an ANOVA contrast that pits RBPscores for the relatively undisturbed streams, Clearand McIntosh Creeks, against all of the other scores.This proposes that score variation is best attributableto differences between these two groups of conditions,capturing the idea that RGA scores will remain similarfor all streams and will match RBP scores at only themost disturbed stream. The sum of squares for thiscontrast is calculated using:

SScontract ¼ nðP ci�xiÞ2Pc2i

ð2Þ

where �xi is the mean for the ith condition, ci is the con-trast coefficient for that condition, and n is the numberof observations in each condition (Dixon, 2003). Thecontrast coefficients are defined to sum to zero, result-ing here in ci = 1/2 for each of the two conditions in thefirst group of the contrast (RBP at Clear and McIn-tosh) and ci = �1/7 for each of the remaining sevenconditions in the other group. This sum of squares



then is subtracted from the total sum of squares toachieve the unexplained variation for the contrastmodel, which can be compared to the unexplained vari-ation in the other models in likelihood ratio form.

Stream and technique differences were furtherinvestigated by using the data for each stream sepa-rately to compare scores among assessment tech-niques at a particular stream. This tested the relateddynamic equilibrium hypothesis that only undis-turbed streams will show differences between assess-ment techniques. All analyses were conducted usingIBM� SPSS� Statistics 19.0 software.

RESULTS

Correlation of Assessment Techniques

Normalized scores for the CSRS, PCSEP, and RBPmethods are provided in Table 2 for each of the 12stream reaches. All three assessment methods rankedthe three streams in the same order of quality fromlowest to highest, Elton, Clear, and McIntosh Creeks.The RBP, however, exhibited more variation amongstreams than the other two, whereas the two geomor-phic assessment techniques exhibited more variationwithin each stream than the RBP (Table 3). ThePCSEP, in particular, had the highest standard devi-ation in all streams compared to the other two meth-ods. Results for the two geomorphic assessments,CSRS and PCSEP, are strongly positively correlated(q = 0.844, t[df=8] = 4.456, p = 0.001). Conversely, theRBP method is not significantly correlated with either the CSRS method (q = 0.276, t[df=8] = 0.812,

p = 0.221) or the PCSEP method (q = 0.264,t[df=8] = 0.774, p = 0.230).

Comparison of Streams

Likelihood ratios comparing single-factor models toa null model (no variance explained) showed strongsupport for the technique factor (k = 6,900,000) butlittle support for the stream factor (k = 4.3; Table 4).The model of the interaction effect (stream 9 tech-nique) also showed little support over the null model(k = 2.3). These likelihood ratios for the single factorsindicated clear differences between the techniquesbut unclear differences between the streams. Basedon the mean values for each stream (Table 3), each ofthe three assessment methods ranked the streams inthe same order of quality. The difference betweenstreams, however, was larger for the RBP methodthan for the other methods (Figure 3). To determineif score variation is best explained by stream

TABLE 2. Results for the Three Assessment Techniques at FourReaches Within Each of the Three Streams. Scores were normal-ized by converting the raw score to the proportion of the maximumpossible score for that technique. The CSRS and PCSEP propor-tional scores were then subtracted from 1 so that high scores repre-sent more stable reaches.

Stream ReachLength(m)

Assessment Scores

CSRS PCSEP RBP

Elton (heavilydisturbed)

E1 180 0.556 0.382 0.560E2 127 0.611 0.566 0.715E3 306 0.500 0.257 0.655E4 370 0.778 0.572 0.670

Clear (marginallydisturbed)

C1 187 0.583 0.454 0.868C2 234 0.556 0.349 0.800C3 147 0.694 0.625 0.730C4 151 0.653 0.520 0.865

McIntosh(undisturbed)

M1 272 0.764 0.651 0.920M2 277 0.667 0.572 0.870M3 226 0.556 0.401 0.810M4 309 0.653 0.382 0.870

TABLE 3. Mean (�x) and Standard Deviation (r) Scoresfor the Three Assessment Techniques at Each Stream.

Assessment Technique (�x � r)

Stream CSRS PCSEP RBP

Elton 0.611 � 0.120 0.444 � 0.153 0.650 � 0.065Clear 0.622 � 0.064 0.487 � 0.116 0.816 � 0.065McIntosh 0.660 � 0.085 0.502 � 0.132 0.868 � 0.045

TABLE 4. ANOVA Model Results for the Full Dataset. Lambda (k)values are the likelihood ratios. The single-factor and interactionmodels are compared to a null model and the contrast model (seetext) is compared to the stream factor model. The contrast modelrepresents the dynamic equilibrium-based hypothesis that predictsthat all streams, regardless of disturbance level, will have a mix ofgeomorphically stable and unstable reaches. This is compared tothe stream factor model to determine if score variation is bestexplained by stream differences or by how the assessment methodsdifferentially capture dynamic equilibrium conditions.

Source of VariationSums ofSquares

UnexplainedVariation

k (comparedto null)

Single factor-Stream 0.073 0.855 4.3Single factor-Technique 0.541 0.387 6,900,000Interaction-Stream9 Technique

0.043 0.885 2.3

Error 0.271 - -Total (null) 0.928 0.928 -

k (comparedto streamfactor)

Contrast (see text) 0.466 0.461 66,000



differences or by how the assessment methods differ-entially capture dynamic equilibrium conditions, thecontrast model based on the equilibrium concept wascompared to the model consisting of just the streamfactor. This comparison yielded a likelihood ratio ofk = 66,000, indicating substantial support for thecontrast model over the model only incorporatingstream differences (Table 4).

The last dynamic equilibrium-based hypothesis isthat only relatively undisturbed streams will showdifferences between the assessment techniques. Thishypothesis was tested by calculating three separatelikelihood ratios (one for each stream) that compare amodel featuring the technique factor with a nullmodel (i.e., no difference between techniques). Thetwo relatively undisturbed streams, Clear and

McIntosh Creeks, indicate substantial support thatthe techniques assess these streams differently(k = 6,800 and 7,000, respectively; Table 5). EltonCreek indicates strong, yet considerably less, supportthat the three techniques assess this highly disturbedstream differently (k = 30).

DISCUSSION

All three assessment methods ranked the streamsin the same order of quality (Elton, Clear, and Mc-Intosh Creeks), corroborating the a priori designationof disturbance levels. As expected, the two RGAmethods are strongly correlated, while neither iscorrelated with the RBP. Despite the correlationbetween RGAs, some minor differences in ranking doexist at the most and least disturbed sites. For exam-ple, when the assessments were conducted in April2009, reach 3 in Elton Creek was ranked as poor bythe PCSEP method, whereas none of the reaches wereidentified as unstable by CSRS (i.e., a score above 20).In the spring of 2010, reach 3 did have a significantbank failure, resulting in emergency bank stabiliza-tion and an instream restoration plan. A difference inhow stability is scored by these two methods doesexist, depending on where the instability occurs in thechannel (upper banks vs. lower banks). More analysiswould be necessary to determine if the PCSEP methodmore accurately predicts instability by comparingother sites over time because the remaining threereaches in Elton Creek have not significantly changedsince the original assessment in 2009.

The variation in scores among streams was greaterfor the RBP method than for the RGA methods (Fig-ure 3). When reach scores are aggregated to thestream level, the geomorphic assessment scoresremain relatively constant across all streams despitevarying disturbance levels, whereas the RBP scores

FIGURE 3. Box Plots of Normalized Assessment Scores for EachMethod at Each Stream. Bars indicate maximum and minimumscores. The contrast model representing the dynamic equilibrium-based hypothesis (see text) pits the RBP scores for Clear and Mc-Intosh Creeks (Group A) against all other scores (Group B).

TABLE 5. Separate ANOVA Model Results for Each of the Three Streams. Lambda (k) valuesare the likelihood ratios and compare the technique factor models to the null model.

Stream Source of VariationSums ofSquares

UnexplainedVariation

k (comparedto null)

Elton Single factor-technique 0.096 0.126 30Error 0.126 -Total (null) 0.222 0.222

Clear Single factor-technique 0.219 0.065 6,800Error 0.065 -Total (null) 0.284 0.284

McIntosh Single factor-technique 0.269 0.080 7,000Error 0.080 -Total (null) 0.349 0.349



varied with disturbance level. Previous studies com-paring RBP and RGA methods have shown significantpositive correlations between the two (e.g., Sullivanet al., 2004; Clark et al., 2008), suggesting that geo-morphically stable streams provide better habitat thanunstable streams. This stability, however, is defined inthe context of a dynamic equilibrium allowing foradjustment within a stream system. Reach 4 in EltonCreek is an example of how stability alone does notnecessarily indicate good habitat: a previouslystraightened portion of the stream remained stablesince the late 1970s, thereby scoring high on the RGAsbut low on the RBP. While in many cases RGA andRBP methods can be linked, other habitat parametersfocus heavily on microscale features that are not ade-quately captured by reach-level stability scores. More-over, the RBP emphasizes diversity of bed featureswithin the reach (e.g., four different velocity/depthregimes and frequency of riffles), which is notaddressed by the RGA methods explicitly. While someadditional biological indicators are captured in thePfankuch method, the Simon-Downs scheme accentu-ates a single state, i.e., stable or unstable.

It is apparent from the model results that RGAswere not able to predict ecological conditions of thestreams accurately. This is likely because all of thestreams assessed consist of localized stable andunstable reaches, consistent with the concept ofdynamic equilibrium. The contrast model based onthis concept performed the best because it accountsfor the capacity of less disturbed streams to maintainhigh ecological integrity despite some inherent insta-bility due to stream dynamics. It has been shownhere and elsewhere (e.g., Ward et al., 2003) that theinformation provided by geomorphic and biologicalassessments is not always consistent among streams,meaning that implementation must consider howthese techniques are employed and interpreted. Fur-thermore, how these methods correlate depends onthe type of stream (Sullivan, 2012) or its level of dis-turbance (Figure 3). As seen here, geomorphic indicescan work as proxies for habitat indices in highly dis-turbed systems such as Elton Creek, but they maynot perform that function in less disturbed systemssuch as Clear Creek or McIntosh Creek. Using RGAsin undisturbed watersheds in states of dynamic equi-librium might inaccurately identify reaches of lowbiological potential. The concept of dynamic equilib-rium does allow for a mix of stable and unstablereaches to occur within a stream system. Similarly,each reach may have quality habitat distributedamong poor habitat. Clark et al. (2008) raised theissue of mismatched scales between rapid streamassessments and the habitat characteristics they areattempting to capture. It was suggested that habitatshould be assessed at multiple scales and that the

spatial distribution of habitat should be considered.Spatial distribution of habitat is an important streamfeature (Tetzlaff et al., 2007) that is not captured byreach-scale assessments.

To help resolve these problems in stream assess-ments, researchers and practitioners must employtheir assessment methods carefully and critically.Comparisons of assessment scores between streamsand among techniques should be framed using ecolog-ically based hypotheses about how those streams arefunctioning in response to the driving forces of theirwatersheds. The results presented here support themodel based on dynamic equilibrium concepts forstreams in our study area, demonstrating how streamanalyses might benefit from ecological hypothesis-based inquiry. The authors advocate that studies gobeyond examining correlations between assessmenttechniques and investigate responses of techniques tovarying ecological settings (e.g., Sullivan, 2012).While this study does not resolve the issues encom-passing the stability-habitat duality (other studiesshow stronger parallels between channel stabilityand ecological metrics, e.g., Sullivan et al., 2004;Clark et al., 2008), it does present an example of howanalytical exploration of these relationships usingmultiple models can provide a deeper understandingof assessment techniques. This supports the recom-mendation of Sullivan and Watzin (2008), who com-bined the use of geomorphic and habitat assessmentsto effectively relate stream physical habitat to streamecological productivity.

Rapid stream assessment techniques that focusprimarily on channel stability may be inappropriatefor stream restoration activities. When planning forrestoration, focusing simply on areas deemed “unsta-ble” will not necessarily correspond to those areas inneed of ecological restoration, as the RGA methodsmay not be sensitive enough to detect streams indynamic equilibrium, perhaps experiencing naturaldisturbances in the system (e.g., fallen trees and asubsequent period of channel adjustment). UsingRGAs as the primary method to prioritize restorationlocations may perpetuate the creation of stable habi-tat features, which may reduce habitat heterogeneityand ecological functions (Florsheim et al., 2008).Stream restoration practitioners should recognize theinherent autogenic aspect of geomorphic instabilitywithin dynamic systems, rather than imposing habi-tat heterogeneity within a context of artificially stablestreams. The process-based approach to stream resto-ration advocated by Bernhardt and Palmer (2011), forexample, calls for an abandonment of geomorphicstructure-focused restoration, with greater emphasisplaced on informing restoration based on previousresults and ecological theory. The dynamic equilib-rium concept discussed here is one component of



stream ecological theory that can help progress resto-ration beyond the narrow structure-based paradigm.The results here indicate that correlations betweenassessment techniques can change as a function ofchannel disturbance. Landscape context is an impor-tant driver of both geomorphic and habitat condi-tions; terrestrial source materials, such as largewoody debris, can dictate the frequency and magni-tude of geomorphically effective events and the devel-opment of heterogeneous habitat (e.g., Montgomeryet al., 1995; Cordova et al., 2007; Arkle et al., 2010).Considering the overall biogeomorphic context ofstream reaches can greatly enhance management andrestoration activities (Montgomery and MacDonald,2002; Bledsoe et al., 2012).

The temporal component of dynamic equilibriumalso may play a role in restoration success. Structur-ally focused restorations often do not speed the rate ofrecovery for dynamic streams (Miller and Kochel,2010). Reach-scale geomorphic adjustments respond tobankfull flow events on time intervals of multipleyears, while microhabitat or pool-riffle habitat canrespond to changes over weeks to months (Allan,1995). While RGAs target recent channel adjustmentsand are expected to capture natural dynamics, theirability to identify adjustment thresholds that causechanges in biotic communities is unknown (Sullivanet al., 2004), and substantial time lags can limit eco-logical response after restoration (Hamilton, 2012).Doyle et al. (2005) help address these issues by pre-senting an analytical framework based around ecologi-cally effective discharge for determining howdischarge drives both geomorphic and ecological pro-cesses simultaneously. For many systems, the spatialand temporal dynamics of stability and habitat willdictate the results of restoration, and as such, restora-tion must be viewed from this dynamic equilibriumperspective.

CONCLUSIONS

Rapid assessment techniques are essential tools forwatershed planners and restoration practitioners toquickly determine impaired reaches deserving inter-vention. Translating these assessments into restora-tion priorities may be problematic in some cases ifthe results are reported and interpreted only at thereach level. Those techniques focusing only on thegeomorphic features of a stream reach (RGAs) mayresult in incorrect characterization of an impairedecosystem, when in fact the reach may simply be in astate of transition with no significant long-term effecton the biota. While RGA methods can identify stream

reaches that are highly degraded and can predictpotential for future instability, these methods maynot recognize more subtle natural variability in astream at the time of the assessment. Furthermore,by excluding categories that address instream habitatfeatures such as those in the RBP, the stability rank-ing schemes have limited ability to accurately iden-tify sites in need of habitat restoration. Some streamchannel instability may be necessary within the con-text of dynamic equilibrium. The RGA methods atpresent do not account for this type of instability, asthey were designed to assess channel instability, nat-ural or man-made, that could potentially impactinfrastructure. By expanding this study to include aneven wider range of stream conditions, a greaterunderstanding of the utility of these assessment tech-niques could be obtained. The concept of dynamicequilibrium should be investigated further as itimpacts stability-based rapid assessment techniquesso that these tools can be improved, thereby enhanc-ing stream restoration activities.

ACKNOWLEDGMENTS

This work was supported by NSF award #0654305, the Ecosys-tem Restoration through Interdisciplinary Exchange (ERIE) IGERTprogram. We thank ERIE affiliates, including David Blersch andAlan Rabideau for their assistance and support. We also thankScott Cornett of New York State Department of EnvironmentalConservation for site support, Jared Aldstadt for help with analy-ses, and two anonymous reviewers who provided many helpfulcomments on ways to improve the manuscript.

LITERATURE CITED

Allan, J.D., 1995. Stream Ecology: Structure and Function of Run-ning Waters. Chapman and Hall, London, United Kingdom,ISBN-13: 978-0412355301.

Allan, J.D. and M.M. Castillo, 2007. Stream Ecology: Structure andFunction of Running Waters (Second Edition). Springer, Dordr-echt, The Netherlands, ISBN-13: 978-1402055829.

Arkle, R.S., D.S. Pilliod, and K. Strickler, 2010. Fire, Flow, andDynamic Equilibrium in Stream Macroinvertebrate Commu-nities. Freshwater Biology 55(2):299-314. doi:10.1111/j.1365-2427.2009.02275.x.

Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling, 1999.Rapid Bioassessment Protocols for Use in Streams and Wade-able Rivers: Periphyton, Benthic Macroinvertebrates, and Fish(Second Edition). EPA 841-B-99-002, U.S. EnvironmentalProtection Agency, Washington, D.C.

Bernhardt, E.S. and M.A. Palmer, 2011. River Restoration: TheFuzzy Logic of Repairing Reaches to Reverse Catchment ScaleDegradation. Ecological Applications 21(6):1926-1931.

Bernhardt, E.S., M.A. Palmer, D.J. Allan, G.J. Alexander, K. Barnas,S. Brooks, J. Carr, S. Clayton, C.N. Dahm, J. Follstad-Shah, D. Ga-lat, S. Gloss, P. Goodwin, D. Hart, B. Hassett, R. Jenkinson, S.Katz, G.M. Kondolf, P.S. Lake, R. Lave, J.L. Meyer, and T.K.O’Donnell, 2005. Synthesizing U.S. River Restoration Efforts.Science 308(5722):636-637, doi: 10.1126/science.1109769.



Bledsoe, B.P., E.D. Stein, R.J. Hawley, and D. Booth, 2012. Frame-work and Tool for Rapid Assessment of Stream Susceptibility toHydromodification. Journal of the American Water ResourcesAssociation 48(4):788-808, doi: 10.1111/j.1752-1688.2012.00653.x.

Clark, J.S., D.M. Rizzo, M.C. Watzin, and W.C. Hession, 2008. Spa-tial Distribution and Geomorphic Condition of Fish Habitat inStreams: An Analysis Using Hydraulic Modelling and Geostatis-tics. River Research and Applications 24(7):885-899, doi: 10.1002/rra.1085.

Cordova, J.M., E.J. Rosi-Marshall, M. Yamamuro, and G.A. Lam-berti, 2007. Quantity, Controls and Functions of Large WoodyDebris in Midwestern USA Streams. River Research and Appli-cations 23(1):21-33, doi: 10.1002/rra.963.

Darby, S. and D. Sear, 2008. River Restoration: Managing theUncertainty in Restoring Physical Habitat. Wiley and Sons,West Sussex, United Kingdom, ISBN-13: 978-0470867068.

Dixon, P., 2003. The P-value Fallacy and How to Avoid It. Cana-dian Journal of Experimental Psychology 57(3):189-202, doi: 10.1037/h0087425.

Dixon, P. and T. O’Reilly, 1999. Scientific Versus Statistical Infer-ence. Canadian Journal of Experimental Psychology 53(2):133-149, doi: 10.1037/h0087305.

Doyle, M.W., E.H. Stanley, D.L. Strayer, R.B. Jacobson, and J.C.Schmidt, 2005. Effective Discharge Analysis of Ecological Pro-cesses in Streams. Water Resources Research 41(11):W11411,doi: 10.1029/2005WR004222.

Florsheim, J.L., J.F. Mount, and A. Chin, 2008. Bank Erosion as aDesirable Attribute of Rivers. BioScience 58(6):519-529, doi: 10.1641/B580608.

Frissell, C.A., W.J. Liss, C.E. Warren, and M.D. Hurley, 1986. AHierarchical Framework for Stream Habitat Classification:Viewing Streams in a Watershed Context. Environmental Man-agement 10(2):199-214, doi: 10.1007/BF01867358.

Goodman, S.N. and R. Royall, 1988. Evidence and ScientificResearch. American Journal of Public Health 78(12):1568-1574.

Hamilton, S.K., 2012. Biogeochemical Time Lags May DelayResponses of Streams to Ecological Restoration. FreshwaterBiology 57(Suppl. 1):43-57, doi: 10.1111/j.1365-2427.2011.02685.x.

Harrelson, C.C., C.L. Rawlins, and J.P. Potyondy, 1994. StreamChannel Reference Sites: An Illustrated Guide to Field Tech-nique. General Technical Report RM-245. USDA Forest Service,Rocky Mountain Forest and Range Experiment Station, FortCollins, Colorado, 61 pp.

Hassett, B., M.A. Palmer, E.S. Bernhardt, S. Smith, J. Carr, andD. Hart, 2005. Restoring Watersheds Project by Project: Trendsin Chesapeake Bay Tributary Restoration. Frontiers in Ecologyand the Environment 3(5):259-267, doi: 10.1890/1540-9295(2005)003[0259:RWPBPT]2.0.CO;2.

Hey, R.D., 1988. Mathematical Models of Channel Morphology. In:Modeling Geomorphological Systems, M.G. Anderson (Editor).Wiley and Sons, New York City, New York, pp. 99-126, ISBN-13: 978-0471918004.

Hughes, R.M., A.T. Herlihy, and P.R. Kaufmann, 2010. An Evalua-tion of Qualitative Indexes of Physical Habitat Applied to Agri-cultural Streams in Ten U.S. States. Journal of the AmericanWater Resources Association 46(4):792-806, doi: 10.1111/j.1752-1688.2010.00455.x.

Julien, P.Y., 2002. River Mechanics. Cambridge University Press,New York City, New York, ISBN-13: 978-0521529709.

Karr, J.R. and E.W. Chu, 2000. Sustaining Living Rivers. Hydrobi-ologia 422(423):1-14, doi: 10.1023/A:1017097611303.

Kondolf, G.M., S.D. Anderson, R. Lave, L. Pagano, A.M. Meren-lender, and E.S. Bernhardt, 2007. Two Decades of River Resto-ration in California: What Can We Learn? Restoration Ecology15(3):516-523, doi: 10.1111/j.1526-100X.2007.00247.x.

Lane, E.W., 1955. The Importance of Fluvial Morphology inHydraulic Engineering. Proceedings of the American Society ofCivil Engineers 81:1-17.

Miller, J.R. and C. Kochel, 2010. Assessment of Channel Dynamics,In-Stream Structures and Post-Project Channel Adjustments inNorth Carolina and Its Implications to Effective Stream Resto-ration. Environmental Earth Sciences 59(8):1681-1692, doi: 10.1007/s12665-009-0150-1.

Montgomery, D.R., J.M. Buffington, R.D. Smith, K.M. Schmidt,and G. Pess, 1995. Pool Spacing in Forest Channels. WaterResources Research 31(4):1097-1105, doi: 10.1029/94WR03285.

Montgomery, D.R. and L.H. MacDonald, 2002. Diagnostic Approachto Stream Channel Assessment and Monitoring. Journal of theAmerican Water Resources Association 38(1):1-16, doi: 10.1111/j.1752-1688.2002.tb01530.x.

Odum, H.T., 1994. Ecological and General Systems: An Introduc-tion to Systems Ecology, Revised Edition. University Press ofColorado, Niwot, Colorado, ISBN-13: 978-0870813207.

Palmer, M.A., H.L. Menninger, and E.S. Bernhardt, 2010. RiverRestoration, Habitat Heterogeneity and Biodiversity: A Failureof Theory or Practice? Freshwater Biology 55(Suppl. 1):205-222,doi: 10.1111/j.1365-2427.2009.02372.x.

Pfankuch, D.J., 1975. Stream Reach Inventory and Channel Stabil-ity Evaluation, USDA Forest Service, R1-75-002. GovernmentPrinting Office, #696-260/200, Washington, D.C., 26 pp.

Rosgen, D.L., 1996. Applied River Morphology. Wildland Hydrol-ogy, Pagosa Springs, Colorado, ISBN-13: 978-0965328906.

Rosgen, D.L., 2001. A Stream Channel Stability Assessment Meth-odology. Seventh Federal Interagency Sedimentation Confer-ence, Reno, Nevada. http://pubs.usgs.gov/misc_reports/FISC_1947-2006/, accessed December 2012.

Shields, D.F., R.R. Copeland, P.C. Klingeman, M.W. Doyle, and A.Simon, 2003. Design for Stream Restoration. Journal of Hydrau-lic Engineering 129(8):575-584, doi: 10.1061/(ASCE)0733-9429(2003)129:8(575).

Simon, A. and P.W. Downs, 1995. An Interdisciplinary Approachto Evaluation of Potential Instability in Alluvial Channels.Geomorphology 12(3):215-232, doi: 10.1016/0169-55X(95)00005-P.

Simon, A., M. Doyle, M. Kondolf, F.D. Shields, Jr., B. Rhoads, andM. McPhillips, 2007. Critical Evaluation of How the RosgenClassification and Associated “Natural Channel Design” Meth-ods Fail to Integrate and Quantify Fluvial Processes and Chan-nel Response. Journal of the American Water ResourcesAssociation 43(5):1117-1131, doi: 10.1111/j.1752-1688.2007.00091.x.

Simon, A. and C.R. Hupp, 1986. Channel Widening Characteristicsand Bank Slope Development Along a Reach of Cane Creek,West Tennessee. In: Selected Papers in Hydrologic Sciences, S.Subitzsky (Editor). U.S. Geological Survey Water-Supply Paper2290, pp. 113-126.

Simon, A. and L. Klimetz, 2008. Magnitude, Frequency, and Dura-tion Relations for Suspended Sediment in Stable (“Reference”)Southeastern Streams. Journal of the American WaterResources Association 44(5):1270-1283, doi: 10.1111⁄j.1752-1688.2008.00222.x.

Sullivan, S.M.P., 2012. Geomorphic-Ecological Relationships HighlyVariable Between Headwater and Network Mountain Streamsof Northern Idaho, United States. Journal of the AmericanWater Resources Association 48(6):1221-1232, doi: 10.1111/j.1752-1688.2012.00682.x.

Sullivan, S.M.P. and M.C. Watzin, 2008. Relating Stream PhysicalHabitat Condition and Concordance of Biotic ProductivityAcross Multiple Taxa. Canadian Journal of Fisheries and Aqua-tic Sciences 65(12):2667-2677, doi: 1139/F08-165.

Sullivan, S.M.P., M.C. Watzin, and W.C. Hession, 2004. Under-standing Stream Geomorphic State in Relation to Ecological



Integrity: Evidence Using Habitat Assessments and Macroinver-tebrates. Environmental Management 34(5):669-683, doi: 10.1007/s00267-004-4032-8.

Tetzlaff, D., C. Soulsby, P.J. Bacon, A.F. Youngson, C. Gibbins,and I.A. Malcolm, 2007. Connectivity Between Landscapes andRiverscapes: A Unifying Theme in Integrating Hydrologyand Ecology in Catchment Science? Hydrological Processes21(10):1385-1389, doi: 10.1002/hyp.6701.

USEPA (U.S. Environmental Protection Agency), 1998. NationalWater Quality Inventory Report to Congress, 305b Report.Washington, D.C.

Ward, T.A., K.W. Tate, E.R. Atwill, D.F. Lile, D.L. Lancaster, N.McDougald, S. Barry, R.S. Ingram, H.A. George, W. Jensen,W.E. Frost, R. Phillips, G.G. Markegard, and S. Larson, 2003. AComparison of Three Visual Assessments for Riparian andStream Health. Journal of Soil and Water Conservation58(2):83-88.



Rapid geomorphic and habitat stream assessment techniques inform restoration differently based on levels of stream disturbance

Documents