PDF (1.261 MB)

The Number of Transects Required to Computea Robust PHABSIM Habitat Index

Thomas R. Payne(1), Steven D. Eggers(1), andDouglas B. Parkinson(2)

Corresponding author: Thomas R. Payne, [email protected]

Abstract. – The value of the number of transects within a segment of an instream flowstudy that will generate a stable relationship between the PHABSIM habitat index anddischarge is examined. Over 600 instream flow studies were reviewed to determine thecommon practice standard, showing a median number of eight transects per four milereach. Since the number of study transects varies by length of reach and purpose ofstudy, transects per mile were computed as a more standardized value, with a medianresult of 2.32 transects per mile for 552 reaches. Studies conducted for water rightsclaims (255) had a median of 1.30 transects per mile and those conducted for projectevaluations (259) had a median of 4.17 transects per mile. Several high transect numberdata sets were also systematically subsampled to determine how many transects are re-quired to produce a robust habitat index function. This analysis showed that approxi-mately 18-20 transects will in most cases yield a result nearly identical to one obtainedusing 40-70 transects per segment. When used for time series analysis and comparisonof flow alternatives, remaining small differences in the functions would be inconsequen-tial. If used for standard setting purposes, however, the differences could affect flow rec-ommendations and more transects are likely to be necessary.

Key words. – Phabsim, IFIM, transect number, habitat modeling, weighted surfacearea, instream flow

INTRODUCTION

Standard one-dimensional hydrau-lic simulation and aquatic habitatmodeling using the Physical HabitatSimulation (PHABSIM) system of theInstream Flow Incremental Method-ology (IFIM) is based on data ac-quired along cross-sectional transects.

Transects are typically placed in ar-eas representative of larger streamsegments, either by selection of rep-resentative reaches or through a pro-cess called habitat mapping (Bovee1997; Bovee et al. 1998). After hy-draulic data is collected on thetransects, hydraulic models are con-structed and calibrated over a rangeof flows, and results are linked with

Hydroécol. Appl. (2004) Tome 14 Vol. 1, pp. 27-53

(1) Thomas R. Payne & Associates, P.O. Box 4678, Arcata, CA 95518(2) Douglas Parkinson & Associates, P.O. Box 131, Bayside, CA 95524

aquatic species habitat suitability cri-teria to produce an index to habitatsuitability in relation to flow, com-monly called weighted usable area orrelative suitability index. This index isthen interpreted to evaluate the im-pact of alternative flow regimes onaquatic resources.

The number of transects believedsufficient to generate the habitat in-dex has never been explicitly identi-fied in the IFIM literature. Numericguidance is lacking because a “suffi-cient” number can vary widely due todifferences in study objectives, diver-sity of channel form, changing pat-terns of hydraulics and aquatichabitat, behavioral use of habitat bytarget species, limitations in availabletime or budget, and the history ofstudy team experience. For example,when all other factors are equal, afairly uniform, channelized streammay be judged well represented byfewer than ten transects, while ahighly variable or braided stream maybe felt to need a much larger numberof transects to account for the higherrange of habitat complexity.

The original perception of riverinehabitat representation using PHABSIMis described in Instream Flow Infor-mation Paper No. 5 (Bovee andMilhous 1978). For any given streamsegment, study sites were selectedon the basis of their “representation”of the overall segment. Part of the se-lection process was intended to sam-ple a stream length approximately 10to 14 times the average channelwidth. This sample length derivedfrom general alluvial river morphol-

ogy precepts on the spacing of suc-cessive riffles at about 5 to 7 times theaverage channel width (Leopold et al.1964), plus the goal of sampling twocomplete riffle and pool sequences(presumably for replication).

Within this study site, placement oftransects was governed by two princi-ples (Bovee and Milhous 1978); first,to sample discrete types of habitat(i.e. pools, riffles, runs), which coversthe purpose for the study, and sec-ond, to sample each hydraulic con-trol, which was required by theprevalent hydraulic model. The place-ment procedure started with manda-tory transects at hydraulic controls,followed by transects in remainingmajor habitat types. If a high level ofdetail is desired, additional transectscan be used to define transitionalhabitat areas (Bovee and Milhous1978). Admonitions about the totalnumber of transects are quite specificin all instances where the issue is ad-dressed:

“At this stage the field crew mustresist the temptation of proliferat-ing the area with cross sections. Asa general rule, never use two crosssections where one would suffice.”(Bovee and Milhous 1978)

“The tendency to dot the land-scape with headstakes becomesvery strong and must be resistedwith an iron will. Remember thatthe greater the detail with which areach is modeled, the more dissim-ilar from adjacent reaches it be-comes.”(Bovee and Milhous 1978)

28 Th. R. Payne et al.

“There is a powerful tendency tocollect as much data as possiblefor any study, expecially if some-one else is paying for it.”(Milhous 1990)

The resulting total number oftransects placed within a representa-tive study site using the original visionwas not provided, but if the logic ofplacement is followed the typicalnumber would be thirteen. This num-ber is derived by assuming therewould be one transect in each of thetwo replicates of three typical majorhabitat types of pool, riffle, and run inthe contiguous site (6), one transectin each transition between the sixtypes (5), and one transect each atthe upper and lower boundaries of thesite (2). Illustrations of transect place-ment in the IFIM literature, whichtypically show only a single (not repli-cate) sequence of habitat units, areconsistent with this number. Whenthe concept of habitat mapping fortransect placement was developed tosupplement the representative reachapproach (Morhardt et al. 1983),there was no discussion of anychange in total transect number.Transects continued to be placed inmajor habitat units in proportion toabundance in ways consistent withprior practice and under standardstudy cost constraints.

Exceptions to what was believed tobe standard practice for transectnumber have surfaced recently in Or-egon and California and prompted thepreparation of this analysis. Insteadof 6, 10, 12 or 14 transects per riversegment, state and federal agencies

have been requiring 30, 45, or evenmore. A new strategy of samplingthree each of every habitat type ex-ceeding a minimum abundance (5-10% by length) has emerged in theseStates, with three transects in eachsampled unit, a procedure which canresult in oversampling infrequenttypes and undersampling abundanttypes. Justification for this intensity ofdata collection has been sparse,mostly centered on the opinions ofstatisticians unfamiliar with PHABSIM,indirect comparisons between oneand two-dimensional modeling sam-pling rates, and an increased senseof pyschological comfort in the qualityof the studies. There has been littleroom for discussion or negotiation,and even direct threats made to denySection 401 water quality certificationif these sampling requirements arenot followed.

LITERATURE REVIEW

There is a relatively high potentialfor low numbers of transects to gener-ate inaccurate habitat index valueson the one hand and high costs asso-ciated with what might be later foundto be an excess number of transectson the other. Given the importance ofbalance between accuracy and costfor PHABSIM studies, there are fewanalyses that have examined thequestion of minimum suitable transectnumber. Morhardt (1986) plotted hab-itat index functions for 9 transectsplaced in different habitat types andconcluded that errors or bias in the

The Number of Transects Required to Compute a Robust PHABSIM Habitat Index 29

weight assigned each transect couldaffect the composite function. Anevaluation of the value of transectsappears in the PHABSIM User’sGuide (Milhous 1990), where habitatindex functions for several specieswere computed from a set of 24transects and visually compared tofunctions for a subsample of4 transects (one per habitat type).With some minor reservations, theGuide concluded that “the shape ofthe relationships are similar...” andthe “number of cross sections can berelatively small...”

Simonson et al. (1994) investi-gated the optimal number andspacing of transects needed to char-acterize the mean values of11 commonly measured stream habi-tat variables (although not thePHABSIM habitat index). Between 35and 40 transects were placed atequal intervals one mean streamwidth (MSW) apart in 86 study siteson 58 Wisconsin streams. The studywas designed to cover at least threeriffle-pool sequences and includemultiple transects within each habitattype. For all streams and habitat vari-ables combined, 20 transects spacedat 2 MSW computed means accuratewithin 5% of the true mean 95% of thetime. With 13 transects, 85% of themeans were within 5% of the truemeans. In general, larger streams re-quired more transects than smallerstreams. Of the variables commonlyused in PHABSIM, the mean of 5 ve-locity measurements per transectwas 5% outside the true mean in 18 of84 sites based on 20 transects. Mean

depth (4 samples per transect) for20 transects was not within 5% of thetrue mean for only 5 of 86 sites. Visualestimates of cover and substratewithin one-half MSW of each transectrequired very few (10 or less)transects to accurately describe thetrue mean.

Williams (1996) attempted to pro-duce confidence intervals around acomposite habitat index functionthrough bootstrap-sampling individ-ual index functions from 15 transectswithin three habitat type strata. Boot-strap sampling with replacement fromsuch a small sample size is a poormethod for this type of analysis be-cause it assumes that transectswithin strata are replicates. In reality,transects within strata may generatesimilar indices but often have an indi-vidual character derived from internalhabitat diversity. The challenge in de-veloping a river model is to includesufficient diversity without over-sampling; bootstrap sampling magni-fies rather than absorbs the extent ofinternal transect diversity. A similarmethod with the same number oftransects and strata was used byBourgeois et al. (1996). Instead of be-ing useful, both studies essentiallyconfirmed the conclusion reached byMorhardt (1986); that differenttransect weights will give differentcomposite index functions, most sig-nificantly in small sample sizes.

A study by Bovee (1997) ontransect number minimized the prob-lematic effect of resampling the sametransects by either randomly or sys-tematically sampling the habitat index


functions of 3 or 5 transects in a popu-lation of 20 placed in a single 61 m(200 ft) habitat unit. The resultsshowed strong similarities betweenhabitat functions of the subsamplesand the total population, with little dif-ference between the two samplesizes or the sampling strategy. Boveeconcluded that pocket water, a com-plex habitat type containing a widevariety of depths and velocities, canbe accurately described with 3 to5 transects. A composite habitat in-dex function from multiple habitatstrata should also reduce the influ-ence of variability found within anygiven strata.

METHODS

Standard Practice for TransectNumber

For the past two decades, ThomasR. Payne & Associates has beenbuilding an archive library of instreamflow studies from various sources, in-cluding contracted work, interlibraryloan, internet searches, and personalcontacts. The library now numbersover 600 references, and is roughlyestimated to contain between one-fifth and one-fourth of all PHABSIMstudies. This meta-data set of studieswas examined to determine whatnumber of transects has been typicalin the application of the IFIM andPHABSIM since the methods weredeveloped. Report sections on studydesign, segment definition, reachstratification, study site selection, and

transect placement were reviewed toextract data on the number oftransects per study site, reach, andsegment, and to compute the numberof transects per unit of representedstream length. Also noted was the re-porting entity and study purpose, yearof study, and geographic region, topossibly identify trends in transectnumber over time, space, or politicalboundaries.

Transect Number Sensitivity Testing

Lack of resistance to the tendencyfor transects to proliferate has resultedin a number of instream flow studiescontaining high sample sizes withinseveral habitat-type strata. Table 1shows eight rivers with large numbersof transects that were subsampledwithin strata to test the effect oftransect sample size on compositehabitat index functions. Rivers areidentified only by stream type andcharacteristics to minimize regulatoryconflicts.

Habitat index functions were cre-ated from these hydraulic files using arange of habitat suitability criteria.Four sets of generic criteria for depthand velocity were created (deep/slow, deep/fast, shallow/slow, shal-low/fast) to include a wide range ofpotential results (Table 2). The chan-nel index variable for substrate and/orcover suitability was set to 1.0, since(with some exceptions) it typicallyonly affects the amplitude of the habi-tat index. Use of actual criteria for realaquatic species was avoided, since



Table 1. – Study Rivers by Stream Type (Large Low (LL) Gradient, Large High (LH) Gradient, SmallHigh (SH) Gradient), Mean Annual Flow in m³/s (Q), Percent Slope (S), Habitat Strata, Number ofTransects in Each Habitat Strata, and Percent of Total Segment Length in Each Habitat Strata forTransect Weighting.

ID Q/S Habitat Strata/Transects by Strata – Percent Habitat by Strata

LL1 33.2 Pool Run/Glide Riffle

0.12 12 74.4 9 19.4 6 6.2

LL2 165 Pool Run/Glide Riffle

0.06 12 77.4 9 11.7 8 10.9

LH1 51.3 Pool Run/Glide Riffle

1.04 27 44.4 24 39.2 20 16.4

LH2 41.8 Pool Rapid Run Riffle

0.79 10 24.6 6 11.0 15 39.0 11 25.4

SH1 1.0 Pool Glide Run Pocket Water (*) Riffle

4.13 13 16.2 8 7.0 5 4.0 12 43.0 9 29.8

SH2 0.5 Pool Rock Garden(**) Run Pocket Water Riffle

5.13 12 22.3 5 8.1 2 1.0 18 56.1 6 12.5

SH3 2.4 Pool Rock Garden Run Riffle

1.40 28 18.5 5 18.5 9 28.6 11 34.4

SH4 1.1 Pool PocketWater Run Riffle

1.53 15 48.0 5 27.5 11 13.4 4 11.1

* small pools mixed with large boulders and high velocity chutes** low gradient run with many emergent large boulders

Table 2. – Generic Habitat Suitability Criteria for Combinations of Depth and Velocity in Deep/Slow,Deep/Fast, Shallow/Slow, and Shallow/Fast.

Slow Fast

Velocity m/s Suitability Velocity m/s Suitability

0.0 1.0 0.0 0.0

0.1 1.0 0.25 1.0

0.5 0.0 0.50 1.0

1.0 0.0

Shallow Deep

Depth m Suitability Depth m Suitability

0.0 0.0 0.0 0.0

0.25 1.0 0.25 0.0

0.5 1.0 0.50 1.0

1.0 0.0 1 1.0

10 0.5

several of the test rivers remain sub-ject to active proceedings.

Computed habitat indices for indi-vidual transects using each of the fourgeneric criteria in each stream weresubsampled by 1) plotting all transecthabitat index functions by habitat typestrata to review similarity and identifyoutliers (Figures 3 and 6), 2) selectingthe index function for a singletransect from each strata, 3) weight-ing each index function according tothe percent of stream length repre-sented by the strata, 4) summing theweighted index functions to create asub-set composite, and 5) graphingand visually comparing the resultagainst the composite function of allavailable transects. For example, inFigure 4 (upper left), the first iterationgave one composite function fromfour transects, compared to the com-posite for all 42 transects. The sec-ond subsample (upper center)consisted of eight transects, the third12, and so on, with each transectweighted by the strata percent di-vided by the number of transects perstrata. This process was completedonce using the sequence of odd-num-bered transects (Subset A), andagain using even-numbered transects(Subset B), for up to 5 transects(20 total) from each strata. In somecases where fewer than 10 transects(5 odd and 5 even) were availablewithin a strata, the larger samplesizes were completed using transectsof the opposing sequence. For riverswhere there were only three habitat

type strata, sampling was repeated toinclude 6 transects of each strata persubset (18 total).

Several mathematical or statisticalapproaches to test differences in thecomposite indices were examinedand discarded as unsatisfactory orunworkable. The absolute (or per-centage) difference between the re-sults for any given flow would notindicate if the indices had similarpeaks and trends. Tests of samplingdifferences about a mean (perSimonson et al. 1994) would not ap-ply since habitat indices are not sam-ples of a population, being firstderived from a composite of manydata points, then adjusted for suitabil-ity and weighted by representation,and finally summed. The bootstraptechnique with replacement used byWilliams (1996) derives broad bound-aries for many samplings, some ofwhich contain many duplicatetransects, and gives no indication ofresults from large sample sizes. Di-vergence testing of indices at specificflows, as done by Bourgeois et al.(1996), does not test single sampleresults or show whether indices arefunctionally identical, as they mayhave the same peak and trend anddiffer only in amplitude, or divergeonly at the outer bounds of extrapola-tion. This analysis therefore relies onvisual comparison of repeat sam-pling by sample size to derive broadconclusions or recommendations,per Milhous (1990) and Bovee(1997).


RESULTS

Standard Practice for TransectNumber

Standard terms for defininginstream flow study areas and guidingplacement of transects include seg-ment, reach, and study site (Bovee etal. 1998, Stalnaker et al. 1995). Asegment is a longer (e.g. hundreds ofchannel widths) section of streamthat is relatively homogeneous interms of hydrology, geomorphology,and pattern. Segments are the funda-mental accounting units for assessingimpacts of flow alteration on total hab-itat and can be characterized by asingle habitat index versus flow rela-tionship. A reach is a comparativelyshort section of stream, one or moreof which can be combined to make upa segment. The actual length of areach is defined by the purpose of thestudy; in some cases a reach is thesite of data collection for a “represen-tative” reach, usually no greater than5-7 times the channel width in length,and in others reaches are portions of

a segment that are more homoge-neous in gradient, hydrology, braid-ing, or other characteristic. A studysite is the physical location wheretransects are placed, whether theyare clustered or broadly distributed.

The research of available instreamflow studies identified 572 studieswhich listed reach or segment length,616 which listed transect number,and 552 which listed both (Table 3).The mean and standard deviation ofthese study parameters are com-puted but are of minimal value sincewidely varying study purposes, reachdefinitions, and study scope controlreach length and transect number.More useful are the median values,which could be considered “typical”:6.44 kilometers for reach length,8.00 for transect number, 0.69 kilo-meters per transect, and 1.44transects per kilometer. The database, although quite large, was inad-equate to derive broader trends intransect number over time, or intransect number differences withinspecific geographic regions.


Table 3. – Sample Size, Mean, Standard Deviation, and Median for Reach Length, Transect Num-ber, Kilometers per Transect, and Transects per Kilometer.

Studies WithReach Length

Studies WithTransect Number

KilometersPer Transect

TransectsPer Kilometer

Sample Size 572 616 552 552

Mean(3) 12.60 km 10.66 1.65 2.88

Std. Deviation 25.10 9.71 3.56 5.23

Median 6.44 km 8.00 0.69 1.44

(3) Mean statistics on kilometers per transect and transects per kilometer are not mathematicalinverses because they are derived from individual study data, not the sums of all data.

Frequency analysis of the 552 stud-ies for which transects per kilometercould be computed is presented inFigure 1. The values for most studiesfell within a range of 0.2 and 4.5transects per kilometer, with somehaving as few as one transect every15 kilometers or more and some witha dozen or more transects over veryshort distances. When the studies arestratified by identified study purpose,the results are: water rights claims(46%), specific project evaluations(47%), basin planning (2%), and re-search (5%). Sample size, mean,standard deviation, and median forreach length, transect number,transects per transect, and transectsper kilometer of water rights claimsand project evaluations are pre-

sented in Table 4. Reach lengths arelonger, number of transects fewer, ki-lometers per transect greater, andtransects per kilometer lower for wa-ter rights claim studies than for pro-ject evaluation studies. A frequencyhistogram of transects per kilometerfor all studies of these two types ispresented in Figure 2.

Transect Number Sensitivity Testing

As an example of the results ob-tained from the sensitivity testing,Figure 3 shows the pool, rapid, run,and riffle habitat index functions bytransect for one of the larger, highergradient sample streams (LH2) usingthe generic deep/fast HSC. StreamLH2 was habitat typed to contain pool


Fig. 1. – Frequency histogram of number of transects per kilometer from 552 PHABSIM studies.

(24.6%), rapid (11.0%), run (39.0%),and riffle (25.4%). Within each ofthese habitat type strata, most of thefunctions show similar patterns of re-

sponse to changing flow, some di-verge from the strata pattern, andsome could fit just about as wellwithin the patterns of other strata.


Fig. 2. – Frequency histogram of number of transects per kilometer for 255 studies conducted forwater rights claims and 259 conducted for project evaluations.

Table 4. – Sample Size, Mean, Standard Deviation, and Median for Reach Length, Transect Num-ber, Kilometers per Transect, and Transects per Kilometer for all Water Rights (WR) and ProjectEvaluation (PE) Studies.

Studies WithReach Length

Studies WithTransect Number

KilometersPer Transect

TransectsPer Kilometer

WR Sample Size 272 280 255 255

PE Sample Size 262 293 259 259

WR Mean 12.53 km 7.69 2.03 1.75

PE Mean 9.29 km 13.88 0.70 4.10

WR Std. Dev. 14.69 6.85 2.19 3.81

PE Std. Dev. 13.04 11.31 1.13 6.14

WR Median 8.05 km 6.00 1.23 0.81

PE Median 4.75 km 10.00 0.39 2.59

These results are typical of individualtransect index responses, and arecaused by hydraulic similarity withinstrata, lack of habitat type uniformitywithin individual transects (e.g. partpool, part run), and transects showingeither unique attributes (e.g. shallowbenches) or transitional characteris-tics (e.g. riffles turning to runs with in-creasing discharge).

The successsive addition oftransects (weighted by strata) to acomposite index function showed pro-gressive convergence towards the fulltransect index function in both SubsetA (e.g. Figure 4) and Subset B (e.g.Figure 5). Eventual convergenceshould be expected, since the se-quence of subsamples would increas-ingly become a greater proportion ofthe complete function; both subsets Aand B showed this pattern. However,the two subsets also showed conver-gence towards each other, eventhough they were (for the most part)composed of indices from differenttransects. Stability in the habitat indexappeared to be reached with approxi-mately 18 to 20 transects, dependingon the number of habitat type strata (3,4, or 5), variation in HSC (Figure 6)and the inclusion or exclusion of par-ticular outliers (Figures 4, 5, 7, and 8).Dividing the transect indices into anupper set and a lower set and makinga similar comparison (results notshown) did not change this conclu-sion.

This pattern of stability was evidentin all sampled streams, whether largeor small, low or high gradient, for allfour generic HSC (Figures 9-16), with

only a few exceptions. The two sub-set 18-transect samples using shal-low/fast HSC in River LL1 (Figure 11,lower left), for example, showed simi-lar trend and magnitude but did notpeak at the same flow. The differenceis enough to affect standard settingdecisions made for flow recommen-dations but is unlikely to have muchinfluence on recommendations de-rived from a time series analysis in-corporating hydrology.

DISCUSSION

Several useful trends in the pro-cess of stable habitat index creationwere revealed through the sub-sampling of larger data sets. First, inless complex streams with low num-ber of habitat type strata, fewertransects seem to be needed to gen-erate a robust index, provided thereare no extreme outliers in the popula-tion. The brook studied by Bourgeoiset al. (1996) had remarkably similarindices for flats, runs, and riffles andcould yield stability with only two orthree transects from each strata. Sec-ond, highly complex streams withmore habitat type strata and very diffi-cult hydraulics still generated a ro-bust index within 20 transects. Thoseindexes still changing as the samplesize approached 20 were streamswith a high percent by length (50-70%) of a single habitat type, so eachsuccessive transect of that type con-tributed significant weight to the com-posite index. As individual transectpercent contribution by weight fell be-



Fig

.3.

–In

divi

dual

tran

sect

habi

tati

ndic

esfo

ra

larg

ehi

ghgr

adie

ntri

ver

(LH

2)fo

rde

ep/f

astH

SC

bypo

ol,r

apid

,run

,and

riff

leha

bita

ttyp

est

rata

.


Fig

.4.

–S

eque

nce

Aof

incr

easi

ngnu

mbe

rof

tran

sect

s(4

,8,1

2,16

,and

20)

for

ala

rge

high

grad

ient

rive

r(L

H2)

with

deep

/fas

tHS

C,i

llust

ra-

ting

the

prog

ress

ion

ofsu

b-sa

mpl

eco

nver

genc

ew

ithth

eco

mpl

ete

data

set,

and

aco

mpa

riso

nw

ithS

eque

nce

Bfo

r20

tran

sect

s.

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

HabitatIndexHabitatIndex




Fig

.5.

–S

eque

nce

Bof

incr

easi

ngnu

mbe

rof

tran

sect

s(4

,8,1

2,16

,and

20)

for

ala

rge

high

grad

ient

rive

r(L

H2)

with

deep

/fas

tHS

C,i

llust

ra-

ting

the

prog

ress

ion

ofsu

b-sa

mpl

eco

nver

genc

ew

ithth

eco

mpl

ete

data

set,

and

aco

mpa

riso

nw

ithS

eque

nce

Afo

r20

tran

sect

s.

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)




Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)


Fig

.6.

–In

divi

dual

tran

sect

habi

tati

ndex

esfo

ra

larg

ehi

ghgr

adie

ntst

ream

(LH

2)fo

rsh

allo

w/s

low

HS

Cby

pool

,rap

id,r

un,a

ndri

ffle

habi

tat

type

stra

ta.


Fig

.7.

–S

eque

nce

Aof

incr

easi

ngnu

mbe

roft

rans

ects

(4,8

,12,

16,a

nd20

)for

ala

rge

high

grad

ient

rive

r(LH

2)w

ithsh

allo

w/s

low

HS

C,i

llus-

trat

ing

the

prog

ress

ion

ofsu

b-sa

mpl

eco

nver

genc

ew

ithth

eco

mpl

ete

data

set,

and

aco

mpa

riso

nw

ithS

eque

nce

Bfo

r20

tran

sect

s.




Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)


Fig

.8.

–S

eque

nce

Bof

incr

easi

ngnu

mbe

roft

rans

ects

(4,8

,12,

16,a

nd20

)for

ala

rge

high

grad

ient

rive

r(LH

2)w

ithsh

allo

w/s

low

HS

C,i

llus-

trat

ing

the

prog

ress

ion

ofsu

b-sa

mpl

eco

nver

genc

ew

ithth

eco

mpl

ete

data

set,

and

aco

mpa

riso

nw

ithS

eque

nce

Afo

r20

tran

sect

s.

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)

Dis

char

ge

(m3 /s

ec)





Fig

.9.

–C

ompa

riso

nof

habi

tati

ndex

esge

nera

ted

usin

gal

ltra

nsec

tsve

rsus

asu

bset

ofth

eto

tal(

sequ

ence

sA

orB

)fo

ra

larg

ehi

ghgr

adie

ntri

ver

(LH

2)w

ithfo

urdi

ffer

entH

SC

.


Fig

.10

.–C

ompa

riso

nof

habi

tati

ndex

esge

nera

ted

usin

gal

ltra

nsec

tsve

rsus

asu

bset

ofth

eto

tal(

sequ

ence

sA

orB

)for

ala

rge

high

grad

ient

rive

r(L

H1)

with

four

diff

eren

tHS

C.


Fig

.11

.–C

ompa

riso

nof

habi

tati

ndex

esge

nera

ted

usin

gal

ltra

nsec

tsve

rsus

asu

bset

ofth

eto

tal(

sequ

ence

sA

orB

)for

ala

rge

low

grad

ient

rive

r(L

L1)

with

four

diff

eren

tHS

C.


Fig

.12

.–C

ompa

riso

nof

habi

tati

ndex

esge

nera

ted

usin

gal

ltra

nsec

tsve

rsus

asu

bset

ofth

eto

tal(

sequ

ence

sA

orB

)for

ala

rge

low

grad

ient

rive

r(L

L2)

with

four

diff

eren

tHS

C.

HS

C:

Dee

p/S

low


Fig

.13

.–C

ompa

riso

nof

habi

tati

ndex

esge

nera

ted

usin

gal

ltra

nsec

tsve

rsus

asu

bset

ofth

eto

tal(

sequ

ence

sA

orB

)for

ala

rge

high

grad

ient

rive

r(S

H1)

with

four

diff

eren

tHS

C.


Fig

.14

.–C

ompa

riso

nof

habi

tati

ndex

esge

nera

ted

usin

gal

ltra

nsec

tsve

rsus

asu

bset

ofth

eto

tal(

sequ

ence

sA

orB

)for

ala

rge

high

grad

ient

rive

r(S

H2)

with

four

diff

eren

tHS

C.


Fig

.15

.–C

ompa

riso

nof

habi

tati

ndex

esge

nera

ted

usin

gal

ltra

nsec

tsve

rsus

asu

bset

ofth

eto

tal(

sequ

ence

sA

orB

)for

ala

rge

high

grad

ient

rive

r(S

H3)

with

four

diff

eren

tHS

C.


Fig

.16

.–C

ompa

riso

nof

habi

tati

ndex

esge

nera

ted

usin

gal

ltra

nsec

tsve

rsus

asu

bset

ofth

eto

tal(

sequ

ence

sA

orB

)for

ala

rge

high

grad

ient

rive

r(S

H4)

with

four

diff

eren

tHS

C.

low about 10%, this effect was damp-ened, even when more-extremeoutliers were included. Third, resultsfor all four generic HSC showed simi-lar patterns for all eight examplerivers. The similarity was evident de-spite the fact that some of the HSCcould be considered less applicableto some of the example rivers (e.g.deep/fast criteria, or large, strongfish, in small high gradient streams),and generated low-magnitude indi-ces due to overall lack of suitable con-ditions.

It could be argued that the habitatindex results produced by the twosubsample sets of 18-20 transects re-main slightly different from each otherand from the total sample, either inamplitude or the flow where the func-tions peak. However, the significanceof these differences depends on theintended purpose and use of theinstream flow study. When used asdesigned by the creators of the IFIM,(incorporation of the habitat indexinto a hydrologic time series for theanalysis of alternatives), there is littlefunctional consequence of the differ-ences and 18-20 transects are suffi-cient. If the study is conducted orinterpreted for rule making or stan-dard setting purposes using eitherpeak of the curve or percentage of thepeak methods, then there could be aconsequence of the differences andmore transects are likely to be neces-sary. As Milhous (1990) noted, “In thefinal analysis, data collected for aninstream flow study is of value only ifthe data influences the decision beingmade.”

The following general guidelinesfor transect number and placementare offered:• The total number of transects

should be proportional to the com-plexity of the habitat hydraulics: 6-10 for simple reaches and 18-20for diverse reaches

• The number of transects needed togenerate a robust index applies toinstream flow decision-makingriver segments, not individual sam-pling sites or reaches

• Transects should be placed in pro-portion to the abundance by lengthof the habitat type strata used, withno individual transect given aweight of more than 5-10%, to min-imize the influence of outliers

• Appropriate habitat type stratifica-tion and proper transect weightingby habitat type are important fea-tures of an adequate PHABSIMstudy

• Habitat indices should always beplotted for individual transects andreviewed for the presence of outli-ers and the reasons for each evalu-ated and understood

• Unless this type of review is con-ducted, unstable composite habi-tat index results can be createdand flawed instream flow decisionscan be made.

REFERENCES

Bourgeois, G., D. Caissie, and N. El-Jabi.1996. Sensitivity analysis of PHABSIMin a small Atlantic salmon stream. Pa-ges B381-B394 in Leclerc, M., H. Ca-


pro, S. Valentin, A. Boudreault, and Y.Cote (eds.). Proceedings of the Se-cond International Symposium on Ha-bitat Hydraulics, Quebec.

Bovee, K.D. 1997. Data collection proce-dures for the Physical Habitat Simula-tion System. U.S. Geological Survey,Biological Resources Division, Ft. Col-lins, CO. 141 pp.

Bovee, K.D., and R.T. Milhous. 1978. Hy-draulic simulation in instream flow stu-dies: theory and techniques. InstreamFlow Information Paper 5. United Sta-tes Fish and Wildlife Service FWS/OBS-78/33. 129pp.

Bovee, K.D., B.L. Lamb, J.M. Bartholow,C.B. Stalnaker, J. Taylor, and J. Hen-riksen. 1998. Stream habitat analysisusing the Instream Flow IncrementalMethodology. U.S.Geological Survey,Biological Resources Division Informa-tion and Technology Report USGS/BRD-1998-0004. viii + 131 pp.

Leopold, L.B., M.G. Wolman, and J.P. Mil-ler. 1964. Fluvial Process in Geomor-phology. W.H. Freeman and Co., SanFrancisco, CA. 522pp.

Milhous, R. T. 1990. User’s Guide to thePhysical Habitat Simulation System –Version II. Instream Flow InformationPaper No. 32. U.S. Fish and WildlifeService Biological Report. 380 p.

Morhardt, J.E., D.F. Hanson, and P.J.Coulston. 1983. Instream flow: impro-ved accuracy through habitat mapping.In Waterpower ’83: International Con-ference on Hydropower (Vol III, pp.1294-1304). September 1983, Knox-ville, Tennessee.

Morhardt, J.E. 1986. Instream flow me-thodologies. Research Project 2194-2,Completion report, Electric Power Re-search Institute, Palo Alto, Californiaby EA Engineering, Science, andTechnology, Inc., Lafayette, California,September 1986. 306pp + apps.

Simonson, T.D., J. Lyons, and P.D. Ka-nehl. 1994. Quantifying fish habitat instreams: transect spacing, samplesize, and a proposed framework. NorthAmerican Journal of Fisheries Mana-gement, 14: 607-615.

Stalnaker, C.B., B.L. Lamb, J. Henriksen,K. Bovee, and J. Bartholow. 1995. TheInstream Flow Incremental Methodolo-gy, a primer for IFIM. National Biologi-cal Service, U.S. Department of theInterior. Biological Report 29, March1995. 45pp.

Williams, J.G. 1996. Lost in space: mini-mum confidence intervals for idealizedPHABSIM studies. Transactions of theAmerican Fisheries Society 125: 458-465.


Documents

PDF (1.261 MB)