Download - Assessing river biotic condition at a continental scale: a European approach using functional metrics and fish assemblages

Journal of Applied Ecology

2006

43

70ndash80

copy 2006 British Ecological Society

Blackwell Publishing Ltd

Assessing river biotic condition at a continental scale a European approach using functional metrics and fish assemblages

D PONTdaggerdagger B HUGUENYdagger U BEIERDagger D GOFFAUXsect A MELCHERpara R NOBLE C ROGERS N ROSET and S SCHMUTZpara

UMR CNRS 5023 Ecologie des Hydrosystegravemes Fluviaux Universiteacute Claude Bernard Lyon 1 43 bd du 11 Novembre 1918 69622 Villeurbanne Cedex France

dagger

Institut de Recherche pour le Deacuteveloppement (IRD) UR 131 Ecologie des Hydrosystegravemes Fluviaux Universiteacute Claude Bernard Lyon 1 43 bd du 11 Novembre 1918 69622 Villeurbanne Cedex France

Dagger

Institute of Freshwater Research National Board of Fisheries 187 93 Drottningholm Sweden

sect

Uniteacute de Recherche en Biologie des Organismes Faculteacutes Universitaires Notre Dame de la Paix Rue de Bruxelles 61 5000 Namur Belgium

para

BOKU-HYDRO Max Emanuelstrasse 17 1180 Wien Austria

Hull International Fisheries Institute University of Hull Hull UK and

daggerdagger

Cemagref Uniteacute HYAX Hydrobiologie 3275 Route de Ceacutezanne Le Tholonet 13612 Aix en Provence France

Summary

1

The need for sensitive biological measures of aquatic ecosystem integrity applicableat large spatial scales has been highlighted by the implementation of the European WaterFramework Directive Using fish communities as indicators of habitat quality in riverswe developed a multi-metric index to test our capacity to (i) correctly model a variety ofmetrics based on assemblage structure and functions and (ii) discriminate between theeffects of natural vs human-induced environmental variability at a continental scale

2

Information was collected for 5252 sites distributed among 1843 European riversData included variables on fish assemblage structure local environmental variablessampling strategy and a river basin classification based on native fish fauna similaritiesaccounting for regional effects on local assemblage structure Fifty-eight metrics reflectingdifferent aspects of fish assemblage structure and function were selected from the availableliterature and tested for their potential to indicate habitat degradation

3

To quantify possible deviation from a lsquoreference conditionrsquo for any given site we firstestablished and validated statistical models describing metric responses to natural envir-onmental variability in the absence of any significant human disturbance We considered thatthe residual distributions of these models described the response range of each metric whateverthe natural environmental variability After testing the sensitivity of these residuals to agradient of human disturbance we finally selected 10 metrics that were combined to obtaina European fish assemblage index We demonstrated that (i) when considering only minimallydisturbed sites the index remains invariant regardless of environmental variability and (ii) theindex shows a significant negative linear response to a gradient of human disturbance

4

Synthesis and applications

In this reference condition modelling approach by includ-ing a more complete description of environmental variability at both local and regionalscales it was possible to develop a novel fish biotic index transferable between catch-ments at the European scale The use of functional metrics based on biological attributesof species instead of metrics based on species themselves reduced the index sensitivity to thevariability of fish fauna across different biogeographical areas

Key-words

biotic river integrity community structure European fish fauna habitatdisturbance modelling Water Framework Directive


(2006)

43

70ndash80doi 101111j1365-2664200501126x

Correspondence D Pont Cemagref Uniteacute HYAX Hydrobiologie 3275 Route de Ceacutezanne Le Tholonet 13612 Aix en ProvenceFrance (fax +33 4 42 66 99 34 e-mail didierpontaixcemagreffr)

71

Biotic integrity assessment of European rivers



43

70ndash80

Introduction

Fish populations and communities are sensitiveindicators of habitat quality in rivers because they reactsignificantly to almost all kinds of anthropogenicdisturbances including eutrophication acidificationchemical pollution flow regulation physical habitatalteration and fragmentation (reviewed by Ormerod2003) This sensitivity to the relative health of theiraquatic environments and the surrounding watershedsis the basis for using biological monitoring of fishes toassess environmental degradation (Fausch

et al

1990)Over the last 30 years a variety of fish-based bioticindices have been widely used to assess river qualityand the use of biologically based multimetric indicesinspired by the index of biotic integrity (IBI) (Karr1981 Karr

et al

1986) has grown rapidly (Simon1999) The main characteristic of these tools is thatthey employ a series of metrics based on assemblagestructure and function that are integrated into anumerical index scaled to reflect the ecological healthof the assemblage Another characteristic is that theyuse the lsquoreference condition approachrsquo (Bailey

et al

1998) comparing an ecosystem exposed to a potentialstress with a reference system unexposed to such astress (Hughes

et al

1998)The accuracy of these biological assessments

depends primarily on the sensitivity of these tools tonatural environmental variation as opposed to human-induced disturbances of river biota To reduce orremove the confounding effects of natural environmentalvariability most authors have validated indices over arestricted range of geographical and environmental sit-uations particular states (Roth

et al

1998 Schleiger2000) ecoregions and drainage areas (McCormick

et al

2001 Smogor amp Angermeier 2001 Emery

et al

2003 Mebane Maret amp Hughes 2003) river sizes(Angermeier amp Schlosser 1987 Simon amp Emery 1995)water thermal regimes (Leonard amp Orth 1986 HughesHowlin amp Kaufmann 2004) and levels of fish diversity(Harris amp Silveira 1999) Most authors also accountfor between-site natural variability by standardizingmetrics in relation to river size Angermeier Smogor ampStauffer (2000) consider that multimetric indices per-form best when coupled with a regional framework sothat the metrics reflect region-specific attributes ofnatural biotic communities Hughes Whittier amp Larsen(1990) called for a necessary compromise between theextremes of uniform nation-wide criteria and uniquecriteria for each waterbody However regardless of theapproach our capacity to distinguish between naturaland human-induced variation of biological conditionsat both local and regional scales remains a crucial point(Hughes

et al

1998)The new European Union (EU) water policy the

Water Framework Directive (WFD) states that allEuropean rivers should be assessed via a reference con-dition approach using bioassessment tools based onfour biotic elements including fish (EU 2000) These

biological assessment tools must also indicate whichfunctional characteristics of the biota are altered in orderto increase the probability of success of ecological riverrehabilitation schemes (Pretty

et al

2003 Giller 2005Palmer

et al

2005)One way to attain this goal is to develop a common

assessment method at the European scale using definedmetrics that remain insensitive to natural environmentalvariability for all unimpaired sites and that are mono-tonically linked to the intensity of human alteration forimpaired sites The objective of this present study wasto develop a fish-based index applicable to all Europeanrivers using a methodology already tested at a nationallevel in France (Oberdorff

et al

2001 2002) We hadtwo main questions (i) Is it possible at the Europeanscale to model correctly a variety of metrics as a func-tion of natural environmental descriptors defined atboth local and regional scales in the absence of anyhuman disturbance (ii) Are we able to quantify forany tested site its deviation from a reference conditionsite having similar natural environmental conditions

Methods

-

We used data from fish surveys of 12 European countriesconducted by several laboratories and governmentalenvironmental agencies (1978ndash2002) These 5252 riverreaches or sites (Fig 1) cover most of the climatic andphysical conditions that occur in Europe

All sites had been sampled using electrofishing tech-niques (DC or PDC (pulsed direct current) waveform)during low flow periods When possible (river depthlt 0middot7 m) river reaches were sampled by wading (64middot9of all sites) For most of these sites the removal methodwas applied (87middot9) and stops nets were not used(88middot7) In large rivers (river depth gt 0middot7 m) samplingwas from boats (35middot1 of all sites) mainly in near-shore areas The size of each sampled site was sufficientto encompass complete sets of characteristic local riverhabitat For 67middot7 of all sites the whole river width wassampled In others cases the whole river section wasonly partially sampled (mainly in near-shore areas) Inorder to standardize the sampling effort we only con-sidered the first passage in all cases Although our dataare subjected to sampling noise this sampling effortwas sufficient to describe the fish assemblage For siteswhere the removal method was applied with threesuccessive passages (2275 sites) the mean percentageof the total number of species caught during the firstpassage was 91middot9 (SD 16middot3) and the mean percentageof total abundance was 63middot2 (SD 13middot1) Samplingeffort was summarized by three variables samplingtechnique (TECH boat or wading) sampling method(METH complete whole river width sampling orpartial) and fished area (FISH) We only retained onefishing occasion per site

72

D Pont

et al



43

70ndash80

For each site the degree of human-induced altera-tions was evaluated based on available data existingknowledge and expert judgement Four disturbancevariables were retained and rated as a function of theirdeviation from a natural state (from 1 no deviation to5 heavily degraded) hydrological disturbances (HYDRclasses 1ndash5 from more than 90 to less than 50 of themean natural water level and from almost no to strongdeviation from natural duration and intensity of flood-ing period) morphological conditions (MORPHfrom negligible morphological alteration to completechannelization with most natural habitats missing)phosphorous nitrogen and total organic carbon(NUTR from conditions within 150 of backgroundlevels to deviation more than 300 from national back-grounds levels) and deviation from critical values of forexample oxygen and pH (TOX) A total disturbanceassessment (DISTURB) was obtained by summing upthe four disturbance variables (range from 4 to 20)

Among the 5252 sites 1608 sites were considered asreference sites (REF) when none of the four dis-turbance variables were rated over 2 This definition

ensured sufficient sample sizes for all countries While notall reference sites were pristine or totally undisturbedthe degree of alteration was null or very low Amongothers sites we distinguished between weakly dis-turbed (WI sites 8 lt DISTURB lt 13) and heavily dis-turbed sites (HI sites DISTURB gt 12)

Each metric-specific model and the final index werevalidated independently by randomly dividing thereference data set into three subsets used for modelcalibration (REF-CAL 1000 sites) model validation(REF-MET 304 sites) and final index validation (REF-IND) We also randomly selected among the weaklydisturbed sites two sets for metric selection (WI-MET958 sites) and final index validation (WI-IND 304sites) and among the heavily disturbed sites one set formetric selection (HI-MET 958 sites) and one set forfinal index validation (HI-IND 304 sites)

Nine abiotic variables were measured in the field orfrom topographical maps or estimated using GIS at

Fig 1 Map of Europe showing 11 river groups and the 5252 sites D Danube E Ebro River MC Mediterranean rivers fromCatalunya MF Mediterranean rivers from France MN Meuse-group rivers NP north Portugal rivers NE northern Europeanplain rivers R Rhocircne River SE south-west Sweden rivers UK United Kingdom rivers WF west France rivers Symbols (circlesand pluses) are only used to distinguish between river groups

73




43

70ndash80

each site altitude (ELE 0ndash1950 m) distance fromsource (DIS 0ndash990 km) basin class (CAT lt 10 km

2

10ndash99 km

2

100ndash999 km

2

1000ndash9999 km

2

gt 10 000 km

2

)reach slope (SLOP 0middot01ndash199 m km

minus

1

) wetted width(WID 0middot5ndash1600 m) mean annual air temperature(TEMP

minus

2ndash+16

deg

C) presenceabsence of a naturallake upstream (LAK) geological type (GEO calcare-ous siliceous) and flow regime (FLOW permanent ortemporary) Four of these explanatory variables (ELESLOP DIS WID) and FISH were log-transformed toreduce the skewness of their distribution

To delineate biologically relevant regional units forEuropean fish we first considered the complete fishfauna lists for each drainage basin unit which representhomogeneous entities with regard to long-term dispersal(Matthews 1998) and explain a significant part of fishcommunity variability (Pont Hugueny amp Oberdorff 2005)However because of the lack of available data forsmall basins we grouped all coastal basins smaller than25 000 km

2

and draining to a given sea coast (ICESFishing Areas httpwwwicesdk) hypothesizing thatthese contiguous basins were in contact during recentHolocene sea level variations For our entire study areawe then compiled data from previous literature to establishlists of native fish species of 19 large basins and 17 groupsof contiguous small basins We examined the similarities(Jaccard Index) between these 36 fauna lists using theunweighted arithmetic average clustering method As acut-off value we chose the similarity level correspondingto the best compromise between a minimal number ofreference sites per cluster (at least 30) and the largestnumber of clusters to increase our description of the spatialvariability of fish faunas at a large scale This procedure(Fig 1) resulted in 11 clusters or river groups (RIVG)

We considered five functional attributes to define thelist of candidate metrics tolerance trophic reproduc-tion habitat and migration (Hughes amp Oberdorff 1999Oberdorff

et al

2001) The 309 fish species caught wereassigned for these attributes based on previous greyor published literature and completed by expertjudgement when necessary (see the web site httpfamebokuacatpublicationshtm 19 Dec 2005) Giventhe scale of the study and the number of species involvednatural history cannot be described in a rigorous quan-titative way for all the species Nevertheless our speciesassignments match well (average percent of match of80) with those realized recently for nine species attributesin Romania (Angermeier amp Davideanu 2004)

Reproduction

Lithophilic species (LITH) require unsilted mineralsubstrate to spawn and their larvae are photophobic

(Balon 1975) They tend to decrease in response tohuman disturbances such as siltation (Berkman ampRabeni 1987) and channelization (Brookes Knight ampShields 1996) Phytophilic species (PHYT) tend tospawn on vegetation and their larvae are not photo-phobic They decrease in response to channelizationbut will commonly increase with aquatic vegetation inrelation with eutrophication

Habitat

The water column (WATE) benthic (BENTH) rhe-ophilic (RHEO) and limnophilic (LIMN) speciesprefer to live and feed in their respective habitat Theabundance of species assigned to these four habitatattributes tends to decrease with increasing habitatalteration (Karr 1981 Oberdorff

et al

2002) RHEOspecies may also increase when river channelizationincreases flow velocity Eurytopic species (EURY) arecharacterized by tolerance of contrasting flow condi-tions and an increase would be indicative of alteration

Trophic guilds

Obligatorily piscivorous species (PISC more than 75fish in the diet Lyons

et al

1995 Goldstein amp Simon1999) and insectivorousinvertivorous species (INSEmore than 75 macro-invertebrates in the diet Lyons

et al

1995) will tend to decrease in response to an alter-ation of their habitat In contrast a metric based onomnivorous species (OMNI more than 25 plantmaterial and more than 25 animal materialSchlosser 1982) will tend to increase in response to dis-turbance as OMNI are able to adapt their trophicregime in response to an alteration of river food webs(Karr 1981)

Tolerance

Tolerant (TOLE) and intolerant (INTO) groups reflectspecies sensitivity to any common impact related toaltered flow regime nutrient regime habitat structureand water chemistry (Karr

et al

1986) Loss of intol-erant species is a response to degradation whereas thenumber of tolerant species will tend to increase withdisturbance

Migration

Potamodromous species (POTA) which migrate withinthe inland waters of a river system (Northcote 1999)and long-migratory (diadromous) species (LONG)which migrate across a transition zone between freshand marine water are expected to decrease in responseto the effects induced by dams and water regulation

The choice of how a metric is expressed is as impor-tant as the selection of the metric itself (Fausch

et al

1990 Karr amp Chu 1999) Each candidate metric wastherefore expressed in four units number of species

74

D Pont

et al



43

70ndash80

(

Ns

) relative number of species (

Ns

number of spe-cies divided by the total species richness) absolute den-sities (

Ni

in number of individuals ha

minus

1

) and relativedensities (

Ni

) Total species richness (RICH) andtotal abundance (DENS) would generally decline withenvironmental degradation (Karr 1981) However anincrease in nutrients (eutrophication) or temperaturecan also lead to an increase with disturbance

We also considered metrics based on non-nativeacclimated species as they can play an important func-tional role in river ecosystems Finally all 309 speciescaught were classified into one or several guilds forwhich we could calculate 58 candidate metrics

For metrics based on abundance data (

n

= 15)stepwise multi-linear regression analysis of each metric[log (

x

+ 1) transformed] on the explanatory variableswas used The squares of each of the five quantitativevariables (ELE TEMP DIS SLOP WID) were alsoincluded to allow for non-linear relationships Formetrics based on the number of species (

n

= 15) weused the same procedure but added an explanatoryvariable FISH as sampling area is well known to influ-ence species richness (Angermeier amp Schlosser 1989)For metrics based on the relative number of species orrelative abundance (

n

= 28) we used stepwise logisticregression analysis All these analyses were performedusing our calibration reference data set (REF-CAL)RIVG TECH METH LAK GEO and FLOW wereentered as dummy variables Variable selection duringthe stepwise procedure was based on the Akaike infor-mation criteria (Hastie amp Pregibon 1993)

Using the independent set of 304 reference sites(REF-MET) we validated each of the resulting metric-specific models expecting that the intercept and theslope of the regression line of the observed vs predictedvalues would not be significantly different fromrespectively 0 and 1 In addition we arbitrarily set aminimal threshold of the variance explained by themodel (determination coefficient) at 0middot30

The residuals of each of the valid metrics (ie thedeviation between the observed and the predicted valueof a metric) measured the range of variation of metricsafter eliminating the effects of environmental variablesand in the absence of any human disturbance Theseresiduals were standardized through subtraction anddivision by respectively the mean and the standarddeviation of the residuals of the reference calibrationdata set (REF-CAL) even when computed on otherdata sets (REF-MET WI-MET HI-MET REF-INDWI-IND HI-IND)

Most of the metrics are expected to be negativelylinked to the intensity of human perturbation Thismeans that the expected value of the residuals for ref-erence sites is zero and less than zero for impacted sitesAssuming that standardized residuals are

N

(01) dis-tributed within reference sites it is possible to compute

the probability of observing a residual value lower thanthe computed one The lower this probability thehigher the probability that a site is impacted For met-rics that are expected to be positively linked to humandisturbance we estimated the probability of observingvalues higher than the computed ones For metrics thatare expected to respond by an increase or a decreasedepending on the type of perturbation we consideredthe probability of observing higher values than thecomputed one for positive residuals or of observinglower values for negative residuals Transforming re-sidual metrics into probabilities as described above isa way of rendering them comparable All probabilitymetrics vary between zero and one and decrease ashuman disturbance increases The expected distr-ibution of these probabilities for reference sites is auniform distribution with mean = 0middot5

Metrics were selected after validation with the REF-MET data set on the grounds of their sensitivity tohuman-induced disturbance and to maximize theindependence among metrics and the diversity ofmetric types First each metric was calculated for thesubsets of weakly (WI-MET) and heavily (HI-MET)altered sites to test the hypothesis that the meanprobability value of REF-MET was higher than WI-MET and that the mean probability value of WI-METwas higher than HI-MET (unilateral

t

-test) Only met-rics that fulfilled these two criteria were retainedhereafter

Then if two metrics were highly correlated (ie Pear-sonrsquos

r

lt 0middot80 or gt

minus

0middot80) we retained the metricbased on a functional species attribute not yet selectedor the metric demonstrating the strongest response to ahigh level of disturbance

For a site given its fish assemblage geographical loca-tion environmental features and the sampling methodused applying the models corresponding to eachof the 10 metrics produces 10 residual values (one permetric) that are subsequently transformed into prob-abilities The final index is obtained by summing upthe 10 probabilities and rescaling the final score from0 to 1 (by dividing it by 10) The index was validatedon three new independent subsets reference (REF-IND) weakly disturbed (WI-IND) and highly dis-turbed (HI-IND) sites We hypothesized that (i) themean value of REF-IND did not differ from 0middot5 and(ii) REF-IND mean value gt WI-IND mean value gt HI-IND mean value (unilateral

t

-test)Two explicit examples (Table 2) that convert actual

site biological and environmental conditions intometric scores and a final index score are given A softwarefreely available on the web (httpfamebokuacat)may be used for this purpose

75




43

70ndash80

Results

Twenty-nine of the 58 metrics were validated (Table 1)Regressions between observed and predicted valueswere highly significant (

R

2

30middot7ndash60middot8) The interceptsand the slopes of the corresponding regression lines didnot significantly differ from zero (Studentrsquos

t

-test

P

-values from 0middot071 to 0middot954) and one (

P

-values from0middot055 to 0middot969) respectively Residual distributionswere checked graphically to verify that they were sym-metrical with only a few outliers

Residuals were calculated and standardized (devi-ation between the observed and the predicted valueTable 2) for each metric and for each of the three datasets REF-CAL WI-MET and HI-MET We thentransformed these residuals into probabilities in agree-ment with our previously defined response hypotheses(Table 1) Among the remaining metrics 17 demon-strated a significant difference between REF-CALand WI-MET mean values (

P

-values from 0middot003 tolt 0middot000001) and between WI-MET and HI-METmean values (

P

-values from 0middot007 to lt 0middot000001) Atthis step three of the five habitat-types metrics (WATE

LIMN EURY) PISC and RICH metrics were ex-cluded As expected the REF-CAL mean values werevery close to 0middot5 (from 0middot499 to 0middot560) for all the re-tained metrics The responses to degradation werein agreement with our previous hypotheses OMNITOLE and PHYT metric types increased while theseven other metric types decreased But metric responsesvaried in intensity with the weakest deviation formetric types demonstrating a positive response tohuman disturbance (OMNI TOLE and PHYT)

Five of the 17 remaining metrics were strongly corre-lated (

Ni

-OMNI and

Ns

-OMNI Pearsonrsquos coefficient

R

= 0middot85

Ni

-LONG and

Ns

-LONG

R

= 0middot89

Ns

-RHEO and

Ns

-LITH

R

= 0middot97

Ni

-RHEO and

Ni

-LITH

R

= 0middot81

Ns

-INSE and

Ns

-INTO

R

= 0middot89)We finally retained 10 metrics (for regression coefficientssee web site httpfamebokuacatpublicationshtm19 Dec 2005) two trophic-based metrics (

Ni

-INSE

Ni

-OMNI) two reproductive guild-based metrics(

Ns

-LITH

Ni

-PHYT) two habitat-based metrics(

Ns

-BENT

Ns

-RHEO) two migration status-basedmetrics (

Ns

-POTA

Ni

-LONG) and two tolerance status-based metrics (

Ns

-INTO

Ns

-TOLE) Three of these

Table 1 List of the 29 metrics retained after the first validation procedure of the multiple linear or logistic models Expectedmetric responses to human disturbances positive response (+) negative response (ndash) positive or negative response (+ndash) R2determination coefficients of the regression of observed vs predicted metric values using the independent reference data set (REF-MET) Mean metrics values (after standardization and transformation into probabilities see text for detailed explanations) forREF-MET and the two data sets of weakly (WI-MET) and highly (HI-MET) disturbed sites P-values of Studentrsquos t-testcomparing REF-MET to WI-MET and WI-MET to HI-MET

MetricsExpected response R2

Mean value(REF-MET)

Mean value(WI-MET)

Mean value(HI-MET)

P-value(REF-WI)

P-value(WI-HI)

Ni-PISC ndash 0middot402 0middot459 0middot363 0middot387 lt 0middot00001 0middot95150Ni-INSE ndash 0middot353 0middot560 0middot228 0middot066 lt 0middot00001 lt 0middot00001Ni-OMNI + 0middot407 0middot512 0middot417 0middot365 lt 0middot00001 000040Ni-EURY + 0middot461 0middot465 0middot431 0middot434 0middot04870 0middot56920Ni-LONG ndash 0middot383 0middot509 0middot293 0middot208 lt 0middot00001 lt 0middot00001Ni-LITH ndash 0middot308 0middot548 0middot258 0middot125 lt 0middot00001 lt 0middot00001Ni-PHYT + 0middot357 0middot530 0middot470 0middot423 0middot00290 0middot00310Ni-TOLE + 0middot446 0middot516 0middot466 0middot479 0middot00630 0middot80870Ns-PISC ndash 0middot466 0middot461 0middot391 0middot406 0middot00020 0middot84140RICH + ndash 0middot608 0middot481 0middot351 0middot339 lt 0middot00001 0middot21240Ns-OMNI + 0middot552 0middot517 0middot462 0middot415 0middot00530 0middot00170Ns-BENT ndash 0middot481 0middot527 0middot338 0middot267 lt 0middot00001 lt 0middot00001Ns-EURY + 0middot555 0middot482 0middot480 0middot458 0middot47120 0middot07470Ns-RHEO ndash 0middot423 0middot512 0middot243 0middot101 lt 0middot00001 lt 0middot00001Ns-WATE ndash 0middot550 0middot500 0middot417 0middot397 0middot00010 0middot10560Ns-LONG ndash 0middot353 0middot504 0middot275 0middot212 lt 0middot00001 lt 0middot00001Ns-POTA ndash 0middot439 0middot499 0middot404 0middot308 lt 0middot00001 lt 0middot00001Ns-LITH ndash 0middot398 0middot512 0middot230 0middot070 lt 0middot00001 lt 0middot00001Ns-TOLE + 0middot493 0middot514 0middot470 0middot464 0middot01980 0middot34700Ni-EURY + 0middot423 0middot473 0middot537 0middot543 0middot99940 0middot64010Ni-RHEO ndash 0middot326 0middot529 0middot426 0middot296 lt 0middot00001 lt 0middot00001Ni-LONG ndash 0middot376 0middot515 0middot517 0middot541 0middot56200 0middot99570Ni-LITH ndash 0middot402 0middot528 0middot386 0middot244 lt 0middot00001 lt 0middot00001Ni-TOLE + 0middot478 0middot519 0middot502 0middot418 0middot20010 lt 0middot00001Ns-INSE ndash 0middot429 0middot510 0middot325 0middot213 lt 0middot00001 lt 0middot00001Ns-EURY + 0middot346 0middot457 0middot607 0middot585 1middot00000 0middot08360Ns-INTO ndash 0middot453 0middot519 0middot314 0middot186 lt 0middot00001 lt 0middot00001Ns-LITH ndash 0middot389 0middot532 0middot260 0middot097 lt 0middot00001 lt 0middot00001Ns-TOLE + 0middot307 0middot538 0middot325 0middot286 lt 0middot00001 0middot00650

76D Pont et al

copy 2006 British Ecological Society Journal of Applied Ecology 43 70ndash80

metrics responded positively to human disturbance(Ni-OMNI Ni-PHYT Ns-TOLE)

The metric values and final index score were computedfor the three independent data sets (REF-IND WI-INDHI-IND) (Fig 2) The mean value of the index (0middot513)in the reference data set did not differ significantly from0middot5 (t = 1middot834 P = 0middot067) The mean index value (0middot343)in the weakly disturbed sites (WI-IND) was signifi-cantly lower than that of the reference sites (t = 16middot546P lt 0middot000001) and significantly higher than that of thehighly disturbed sites (0middot235 t = 10middot36 P lt 0000001)

By examining the percentages of well-classified ref-erence (REF-IND) and disturbed sites (WI-IND andHI-IND) as a function of each index score value wedemonstrated that the best cut-off level for assemblagelsquoimpairmentrsquo was an index value of 0middot423 with 81middot4of the reference and disturbed sites correctly classified

Table 2 Two examples of calculation of the European fish index value from the 10 retained metrics used in the model (site 1undisturbed site with no individual disturbance variable rated over 1 site 2 highly disturbed site with a global disturbance value(DISTURB) of 14) For each site the list of species caught (within parentheses number of individuals caught per species at a givensampling date and the metric set to which the species was assigned) and environmental conditions (see text for acronymsignification) are given Metrics (see text for acronym signification) are expressed in number of species (Ns) relative number ofspecies (Ns number of species divided by the total species richness) absolute densities (Ni in number of individuals haminus1) andrelative densities (Ni) Fish assemblage characteristics are converted into an observed metric Environmental conditions areused to compute a theoretical metric value The observed minus the predicted values are standardized and transformed intoprobabilities In the absence of any disturbance a value of 0middot5 is expected The index is obtained by summing up the 10 metrics

Site 1 River Leven (UK)Sampling date 23 August 2002Fish assemblage

Anguilla anguilla (1 TOLE BENT LONG) Barbatula barbatula (5 BENT RHEO LITH) Cottus gobio (100 INTO BENT RHEO LITH INSE) Lampetra planeri (1 INTO BENT RHEO LITH POTA) Phoxinus phoxinus (55 RHEO LITH) Salmo salar (53 INTO RHEO LITH INSE LONG) Salmo trutta fario (39 INTO RHEO LITH INSE)

Environmentalconditions

RIVG (UnitedKingdom) CAT (lt 100 km2) ELE (45 m) GEO (Siliceous) FLOW (Permanent) LAK (No) TEMP (8middot5 degC) SLOP (2middot75 m kmminus1) DIS (14 km) WID (7middot3 m) METH (Whole) TECH (Wading) FISH (365 m2)

Index value 0middot80

MetricsNi-INSE

Ni-OMNI

Ni-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE

Observed 8middot568 0middot000 0middot000 1middot609 1middot946 1middot099 0middot693 0middot996 0middot570 0middot140Predicted 6middot847 2middot717 0middot788 1middot092 1middot506 0middot706 0middot227 0middot009 0middot004 0middot002Probability 0middot841 0middot896 0middot723 0middot896 0middot889 0middot905 0middot880 0middot727 0middot645 0middot636

Site 2 Seine River (FR)Sampling date 19 September 1996Fish assemblage Abramis brama (4 TOLE BENT OMNI POTA) Anguilla anguilla (22 TOLE BENT LONG) Gobio gobio

(5 INTO BENT RHEO LITH INSE) Leuciscus cephalus (13 RHEO LITH OMNI POTA) Perca fluviatilis (6 TOLE) Rutilus rutilus (159 TOLE OMNI) Sander lucioperca (1) Scardinius erythrophthalmus (6 PHYT OMNI)

Environmental conditions

RIVG (WestFrance) CAT( gt 10000 km2) ELE (8 m) GEO (Calcareous) FLOW (Permanent) LAK (No) TEMP 10middot5 degC SLOP (1middot0 m kmminus1) DIS (615 km) WID (100 m) METH (Partial) TECH (Boat) FISH (1440 m2)


MetricsNi-INSE

Ni-OMNI

NI-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE

Observed values 0middot000 7middot143 3middot761 1middot386 1middot099 0middot693 1middot099 0middot060 0middot000 0middot500Predicted values 5middot481 3middot554 0middot961 1middot929 2middot069 1middot066 0middot894 0middot236 0middot184 0middot261Probability 0middot001 0middot048 0middot018 0middot093 0middot004 0middot107 0middot698 0middot189 0middot159 0middot037

Metrics expressed in ln(x + 1)

Fig 2 Distribution of the index scores for REF-CAL(calibration reference sites) REF-IND (independent referencesites n = 304) WI-IND (weakly disturbed sites n = 304) andHI-IND (highly disturbed sites n = 304)

77Biotic integrity assessment of European rivers


In order to test index independence to natural envi-ronmental variability we performed a stepwise linearregression of the index values on all 10 environmentaldescriptors using the independent reference data set(REF-IND) None of the descriptors was retained andthe part of the index variability explained by thesedescriptors was not significant (R2 = 0middot115 F-test =1middot505 P = 0middot064) When considering each of the 10environmental descriptor separately (Fig 3) multiplecomparison Tukeyrsquos test showed that the index valueswere invariant whatever the value of the descriptortested except for the two highest elevation classes

To evaluate the ability of any impacted site to deviatefrom a reference condition (ie a mean index value of0middot5) we regressed the index values of all independentsites (REF-IND WI-IND HI-IND) on the globalassessment impact variable (IMPACT) The relation-ship (Fig 4) was highly significant (R2 = 0middot4678 n =912 P lt 0middot000001) The standard error was 0middot1236and the residual distribution was normalized (goodness-of-fit KolmogorovndashSmirnov test P = 0middot6013) Residuals

Fig 3 Distribution of the index score for each of the 10 environmental variables (box-plot graphs) for the two independentvalidation data sets (REF-MET and REF-IND n = 608)

Fig 4 Regression of the fish index values on the total human dis-turbance index (from four to 19) Mean index values (and their95 confidence intervals) for each disturbance index class

78D Pont et al


at disturbance level 18 were highly dispersed because ofthe presence of an outlier (index value of 0middot44) andbecause of the low number of sites (seven) at this level

Discussion

This study demonstrates that at a continental-scale(Europe) it was possible to develop a multimetricfish-based index that (i) remained invariant for allunimpaired sites whatever the natural environmentalconditions and (ii) gave a significant negative linearresponse to a gradient of physical and chemical humandisturbances (ie the index can be used to assess thehuman-induced impact on the biotic condition inrivers) To our knowledge this is the first time this kind ofindex has been successfully developed at such a largespatial scale This is particularly noteworthy given thegreat variability of fish composition across differentEuropean biogeographical areas Three features of ourapproach may have contributed to this First usingfunctional metrics instead of taxonomic metricsreduced the sensitivity of the index to the variability infish faunas between biogeographical regions Secondlyincluding the main factors known to affect fish assem-blage structure in the model reduced the influence ofthe geographical and upstreamndashdownstream variabilityof these variables And finally including a biologicallybased regionalized variable added spatial flexibilityto our approach

The advantage of considering functional metrics canbe illustrated by the case of lithophilic species Amongthe 63 lithophilic species included in the LITH metrictype at the European scale only two species are com-mon to all 11 hydrological units (brown trout andrainbow trout) Only 23 species occur in six units while17 species are only present in one hydrological unitDespite this variability in species composition betweenriver units LITH consistently decreased in response tohuman disturbance across Europe as a whole demon-strating that the retained metrics are truly functionalones

As demonstrated by our models functional descrip-tors of fish communities respond to environmental vari-ability in several ways But all the 10 retained metricsresponded significantly to river slope Sampling methodsalso significantly affected eight of the 10 metrics asdemonstrated previously by Reynolds et al (2003)

A tenet of our approach is that the variance of themetrics not accounted for by the environmental vari-ables included in the models should be in large part theresult of human disturbance In fact human distur-bance only explains about 50 of total index variancesuggesting that the models may be improved by addingenvironmental variables not considered in this studyThe unexplained variance in the index may also re-sult from imprecision in fish sampling because of theinescapable differences in fish sampling methods usedbetween different habitats and countries Data ondifferent types of river modifications were not always

comparable between countries so we only retained thefour most reliable and complete disturbance variablesHowever others types of disturbance have to be con-sidered such as fishing and introduced species whichcan affect biotic interactions

Moreover riverine fish assemblages may vary greatlyover time To check this the variance in index scoresassociated with the temporal variability of fish assem-blages andor sampling variability was evaluated bycomputing the standard deviation associated with 12time series (eight to 36 sampling dates) distributedamong four countries (Belgium France LithuaniaSweden) at sites where there were no perceivablechanges in human disturbance intensity during theperiod sampled A mean of the standard deviationsper site of 0middot06 can be compared with the value of0middot169 observed for references sites As the variance ofthe index within reference sites is the result of non-modelled spatial variability as well as to sampling noiseand temporal variability this suggests that sampling noiseaccounts for at most 35 of index variability Hencethe most probable solution to improving the power ofthe index would appear to be by improving the model-ling of its spatial variability (eg by considering newvariables or other modelling approaches)

Another predictive approach based on the model-ling of assemblages within reference sites as a functionof environmental variables RIVPACS has recentlybeen applied to fish assemblages in New Zealand (Joyamp Death 2002) Besides the way assemblages aremodelled (classification vs regression) the main differ-ence with our approach is the use of taxonomic rich-ness instead of several metrics In a sense our metricsRICH may be considered a RIVPACS-type descriptorinserted within a more general multi-metrics approachAs a result we expect our approach to be more powerfuland more flexible Despite the fact that we modelledfunctional metrics instead of taxonomic metrics and thatwe included some important environmental variablesthe spatial variability in metric values has not been fullyaccounted for by our models as exemplified by theinclusion of the regionalized variable lsquoriver grouprsquo ineight of the 10 retained models The two metrics typesthat are insensitive to regional classification of fishfauna are the omnivorous species (Ni-OMNI) and tol-erant species (Ns-TOLE) ie metrics comprised ofgeneralist species able to colonize a wide spectrum ofriver environments Among the 29 characteristicallyomnivorous species 21 of them are common to at leastsix of the 11 river groups

In previous works most authors consider thatassessment criteria must be region-specific (Anger-meier Smogor amp Stauffer 2000) Ecologically definedregions are thereby considered to be relevant entitieseven if there is still considerable debate regarding howthey should be defined (Omernik amp Bailey 1997 VanSickle amp Hughes 2000) In our approach we explicitlyconsidered this question by including in our models anenvironmental variable acting at regional scales (river



groups) in accordance with current views emphasiz-ing regional influences on biodiversity (Ricklefs ampSchluter 1993) Our regional classification is generallyin agreement with the classical biogeographical historyof Europe (Banarescu 1992) Fish faunas from theNetherlands northern Germany Poland Lithuaniaand northern Sweden (European North Plain rivergroup) appear as similar related to their common re-colonization after the glacial periods The Baltic Sea wasoligohaline and did not represent an ecological barrierto dispersal The lsquoriver grouprsquo variable participatesadditively in our models meaning that the models arequalitatively consistent over Europe For instance ametric that is positively correlated to river slope in agiven region will also be positively correlated with riverslope in other European regions Hence the models arepartly transferable between regions but some regionaladjustments are needed Interregional variations maybe linked to variation in taxonomy and phylogenetichistory that in turn affect metric distribution withinfaunas and also to spatial variation in environmentalconstraints not included into models but which canalso affect functional characteristics of fish assem-blages (Smogor amp Angermeier 2001) Clearly furtherwork is needed to identify the factors underlying theregional component of our models

This present approach is based on modelling fishassemblage structure in reference sites Thus the defi-nition and quality of the reference data set are keyissues We collected information from a very largenumber of sites distributed among 1843 Europeanrivers compiling a database unique in Europe Althoughsites are not evenly distributed they cover virtually allthe environmental situations a European fish speciescan encounter within its area of distribution Howevernatural large flood plain rivers were lacking in ourreference data set because of their rarity in westernEurope and the efficiency of our index to assess suchenvironments needs to be improved in the future Wedid not restrict our selection of reference sites to onlyundisturbed sites but also considered sites slightlyimpacted However the mean index value for undis-turbed sites is slightly but significantly higher than forslightly impacted sites (5 le DISTURB le 8) (0middot518against 0middot502 t-test value = 2middot349 P = 0middot019) suggest-ing that distinguishing between these two groups in thefuture would increase the power of the index

In conclusion the need to define sensitive biologicalmeasures of aquatic ecosystem integrity transferable toother catchments or regions at continental scales isnow clear especially in Europe with the implementationof the WFD The solution we have used to meet thisgoal is to include in our reference condition model-ling approach a more complete description of abioticand biotic environmental variability at both local andregional scales This sort of tool has never been devel-oped before at a continental scale Using this approachour models and the final index are transferable butonly for sites and rivers belonging to the area consid-

ered in our previous calibration data set (REF-CAL)A generalization of our method to an even larger areais possible but only by collecting new data coveringthese new areas and by recalibrating our models Thismethodology could also be improved by including bet-ter biological knowledge in the definition of the metrictypes improving disturbance assessment and by usingnew statistical techniques Lastly the principles of ourmethodology could be applied to a wide variety of bio-logical groups

Acknowledgements

This work was funded by the European Commissionunder the Fifth Framework Programme (FAMEproject contract number EVK1-CT-2001-00094) Sitedata and many insights for this study were providedby numerous biologists W Andrzejczak J BackxR Barbieri C Belpaire R Berg R Berrebi J Boche-chas J Bocian J Boumlhmer R Borrough J BreineT Buijse N Caiola F Casals J de Leeuw E DegermanT Demol U Dussling A Economou T Ferreira CFrangez IG Cowx PD Gerard S Giakoumi GGrenouillet R Haberbosch R Haunschmid A JagschK Karras V Kesminas P Kestemont M LapinskaJ Molis A Muumlhlberg J Oliveira JM Olivier JPettersson Paul Quataert Y Reyols H SchmidI Simoens A Sostoa A Starkie M StoumboudiG Verhaegen T Virbickas E Winter H Wirloumlf MZalewski S Zogaris We are particularly grateful toGertrud Haidvogl for the management of the FAMEproject Two anonymous referees supplied helpfulcomments on the paper

References

Angermeier PL amp Davideanu G (2004) Using fish commu-nities to assess streams in Romania initial development ofan index of biotic integrity Hydrobiologia 511 65ndash78

Angermeier PL amp Schlosser IJ (1987) Assessing bioticintegrity of the fish community in a small Illinois streamNorth American Journal of Fisheries Management 7 331ndash338

Angermeier PL amp Schlosser IJ (1989) Speciesndasharea rela-tionships for stream fishes Ecology 70 1450ndash1462

Angermeier PL Smogor RA amp Stauffer JR (2000)Regional frameworks and candidate metrics for assessingbiotic integrity in mid-atlantic highland streams Trans-actions of the American Fisheries Society 129 962ndash981

Bailey RC Kennedy MG Dervish MZ amp Taylor RM(1998) Biological assessment of freshwater ecosystemsusing a reference approach comparing predicted andactual benthic invertebrate communities in Yukon streamsFreshwater Biology 39 765ndash774

Balon EK (1975) Reproductive guilds of fishes a proposaland definition Canadian Journal of Fisheries and AquaticSciences 32 821ndash864

Banarescu P (1992) Zoogeography of Fresh Waters II Distri-bution and Dispersal of Freshwater Animals in North Americaand Eurasia AULA-Verlag Wiesbaden Germany

Berkman HE amp Rabeni CF (1987) Effect of siltation on fishcommunities Environmental Biology of Fishes 18 285ndash294

Brookes A Knight SS amp Shields D Jr (1996) Habitat

80D Pont et al


enhancement River Channel Restoration (eds A Brookes ampFD Shields Jr) pp 103ndash126 John Wiley amp Sons LtdChichester UK

Emery EB Simon TP McCormick FH AngermeierPL Deshon JE Yoder CO Sanders RE PearsonWD Hickman GD Reash RJ amp Thomas JA (2003)Development of a multimetric index for assessing thebiological condition of the Ohio River Transactions of theAmerican Fisheries Society 132 791ndash808

Fausch KD Lyons J Karr JR amp Angermeier PL (1990)Fish communities as indicators of environmental degrada-tion American Fisheries Society Symposium 8 123ndash144

Giller PS (2005) River restoration seeking the ecologicalstandards Journal of Applied Ecology 42 201ndash207

Goldstein RM amp Simon TP (1999) Toward a united defini-tion of guild structure for feeding ecology of North Americanfreshwater fishes Assessing the Sustainability and BiologicalIntegrity of Water Resources using Fish Communities (edTP Simon) pp 123ndash220 Lewis Press Boca Raton FL

Harris JH amp Silveira R (1999) Large-scale assessments ofriver health using an index of biotic integrity with low-diversity fish communities Freshwater Biology 41 235ndash252

Hastie TJ amp Pregibon D (1993) Generalized linear modelsStatistical Models in S (eds JM Chambers amp TJ Hastie)pp 195ndash247 Chapman amp Hall London UK

Hughes RM amp Oberdorff T (1999) Applications of IBIconcepts and metrics to waters outside the United Statesand Canada Assessing the Sustainability and BiologicalIntegrity of Water Resources Using Fish Communities (edTP Simon) pp 79ndash83 Lewis Press Boca Raton FL

Hughes RM Howlin S amp Kaufmann PR (2004) Abiointegrity index (IBI) for coldwater streams of westernOregon and Washington Transactions of the AmericanFisheries Society 133 1497ndash1515

Hughes RM Kaufmann PR Herlihy AT KincaidTM Reynolds L amp Larsen DP (1998) A process fordeveloping and evaluating indices of fish assemblage inte-grity Canadian Journal of Fisheries and Aquatic Sciences 551618ndash1631

Hughes RM Whittier CMR amp Larsen DP (1990) Aregional framework for establishing recovery criteria Environ-mental Management 14 673ndash683

Joy MK amp Death RG (2002) Predictive modelling offreshwater fish as a biomonitoring tool in New ZealandFreshwater Biology 47 2261ndash2275

Karr JR (1981) Assessment of biotic integrity using fishcommunities Fisheries 6 21ndash27

Karr JR amp Chu EW (1999) Restoring Life in RunningWaters Better Biological Monitoring Island Press Wash-ington DC

Karr JR Fausch KD Angermeier PL Yant PR ampSchlosser IJ (1986) Assessing Biological Integrity in RunningWaters A Method and its Rationale Special PublicationNo 5 Illinois Natural History Survey Champaign IL

Leonard PM amp Orth DJ (1986) Application and testingof an index of biotic integrity in small coolwater streamsTransactions of the American Fisheries Society 115 401ndash414

Lyons J Navarro-Perez S Cochran PA Santana EC ampGuzman-Arroyo M (1995) Index of biotic integrity basedon fish assemblages for the conservation of streams andrivers in West-Central Mexico Conservation Biology 9569ndash584

McCormick FH Hughes RM Kaufmann PR PeckDV Stoddard JL amp Herlihy AT (2001) Development ofan index of biotic integrity for the mid-Atlantic highlandsregion Transactions of the American Fisheries Society 130857ndash877

Matthews WJ (1998) Patterns in Freshwater Fish EcologyChapman amp Hall New York NY

Mebane CA Maret TR amp Hughes RM (2003) An index

of biological integrity (IBI) for Pacific Northwest riversTransactions of the American Fisheries Society 132 239ndash261

Northcote TG (1999) Migratory behaviour of fish andits significance to movement through riverine fish passagefacilities Fish Migration and Fish Bypasses (eds S Jung-wirth S Schmutz amp S Weiss) pp 3ndash18 Fishing NewsBooks Blackwell Science Oxford UK

Oberdorff T Pont D Hugueny B amp Chessel D (2001) Aprobabilistic model characterizing fish assemblages ofFrench rivers a framework for environmental assessmentFreshwater Biology 46 399ndash415

Oberdorff T Pont D Hugueny B amp Porcher JP (2002)Development and validation of a fish-based index (FBI) forthe assessment of lsquoriver healthrsquo in France Freshwater Bio-logy 47 1720ndash1734

Omernik JM amp Bailey RG (1997) Distinguishing betweenwatersheds and ecoregions American Water ResourceAssociation 33 935ndash949

Ormerod SJ (2003) Current issues with fish and fisherieseditorrsquos overview and introduction Journal of Applied Ecol-ogy 40 204ndash213

Palmer MA Bernhardt ES Allan JD Lake PSAlexander G Brooks S Carr J Clayton S DahmCN Follstad Shah J Galat DL Loss SG GoodwinP Hart B Hassett DD Jenkinson R Kondolf GMLave R Meyer JL OrsquoDonnell TK Pagano L ampSudduth E (2005) Standards for ecologically successfulriver restoration Journal of Applied Ecology 42 208ndash217

Pont D Hugueny B amp Oberdorff T (2005) Modelling hab-itat requirement of European fishes do species have similarresponses to local and regional environmental constraintsCanadian Journal of Fisheries and Aquatic Sciences 62163ndash173

Pretty JL Harrison SSC Shepherd DJ Smith AG ampHey RD (2003) River rehabilitation and fish populationsassessing the benefit of instream structures Journal ofApplied Ecology 40 251ndash265

Reynolds L Herlihy AT Kaufmann PR Gregory SVamp Hughes RM (2003) Electrofishing effort requirementsfor assessing species richness and biotic integrity in westernOregon streams North American Journal of FisheriesManagement 23 450ndash461

Ricklefs RE amp Schluter D (1993) Species Diversity in Eco-logical Communities University of Chicago Press Chicago IL

Roth NE Southerland M Chaillou J Klauda RKazyak P Stranko S Weisberg S Hall JR amp MorganR (1998) Maryland biological stream survey developmentof a fish index of biotic integrity Environmental Monitoringand Assessment 51 89ndash106

Schleiger SL (2000) Use of an index of biotic integrity todetect effects of land uses on stream fish communitiesin westernndashcentral Georgia Transactions of the AmericanFisheries Society 129 1118ndash1133

Schlosser IJ (1982) Fish community structure and functionalong two habitat gradients in a headwater stream Eco-logical Monographs 52 395ndash414

Simon TP (1999) Assessing the Sustainability and BiologicalIntegrity of Water Resources Using Fish CommunitiesLewis Publishers Boca Raton FL

Simon TP amp Emery EB (1995) Modification and assess-ment of an index of biotic integrity to quantify waterresource quality in great rivers Regulated Rivers Researchand Management 11 283ndash298

Smogor RA amp Angermeier PL (2001) Determining aregional framework for assessing biotic integrity of VirginiaStreams Transactions of the American Fisheries Society130 18ndash35

Received 3 January 2005 final copy received 8 September 2005Editor Paul Giller

71




43

70ndash80

Introduction

Fish populations and communities are sensitiveindicators of habitat quality in rivers because they reactsignificantly to almost all kinds of anthropogenicdisturbances including eutrophication acidificationchemical pollution flow regulation physical habitatalteration and fragmentation (reviewed by Ormerod2003) This sensitivity to the relative health of theiraquatic environments and the surrounding watershedsis the basis for using biological monitoring of fishes toassess environmental degradation (Fausch

et al

1990)Over the last 30 years a variety of fish-based bioticindices have been widely used to assess river qualityand the use of biologically based multimetric indicesinspired by the index of biotic integrity (IBI) (Karr1981 Karr

et al

1986) has grown rapidly (Simon1999) The main characteristic of these tools is thatthey employ a series of metrics based on assemblagestructure and function that are integrated into anumerical index scaled to reflect the ecological healthof the assemblage Another characteristic is that theyuse the lsquoreference condition approachrsquo (Bailey

et al

1998) comparing an ecosystem exposed to a potentialstress with a reference system unexposed to such astress (Hughes

et al

1998)The accuracy of these biological assessments

depends primarily on the sensitivity of these tools tonatural environmental variation as opposed to human-induced disturbances of river biota To reduce orremove the confounding effects of natural environmentalvariability most authors have validated indices over arestricted range of geographical and environmental sit-uations particular states (Roth

et al

1998 Schleiger2000) ecoregions and drainage areas (McCormick

et al

2001 Smogor amp Angermeier 2001 Emery

et al

2003 Mebane Maret amp Hughes 2003) river sizes(Angermeier amp Schlosser 1987 Simon amp Emery 1995)water thermal regimes (Leonard amp Orth 1986 HughesHowlin amp Kaufmann 2004) and levels of fish diversity(Harris amp Silveira 1999) Most authors also accountfor between-site natural variability by standardizingmetrics in relation to river size Angermeier Smogor ampStauffer (2000) consider that multimetric indices per-form best when coupled with a regional framework sothat the metrics reflect region-specific attributes ofnatural biotic communities Hughes Whittier amp Larsen(1990) called for a necessary compromise between theextremes of uniform nation-wide criteria and uniquecriteria for each waterbody However regardless of theapproach our capacity to distinguish between naturaland human-induced variation of biological conditionsat both local and regional scales remains a crucial point(Hughes

et al

1998)The new European Union (EU) water policy the

Water Framework Directive (WFD) states that allEuropean rivers should be assessed via a reference con-dition approach using bioassessment tools based onfour biotic elements including fish (EU 2000) These

biological assessment tools must also indicate whichfunctional characteristics of the biota are altered in orderto increase the probability of success of ecological riverrehabilitation schemes (Pretty

et al

2003 Giller 2005Palmer

et al

2005)One way to attain this goal is to develop a common

assessment method at the European scale using definedmetrics that remain insensitive to natural environmentalvariability for all unimpaired sites and that are mono-tonically linked to the intensity of human alteration forimpaired sites The objective of this present study wasto develop a fish-based index applicable to all Europeanrivers using a methodology already tested at a nationallevel in France (Oberdorff

et al

2001 2002) We hadtwo main questions (i) Is it possible at the Europeanscale to model correctly a variety of metrics as a func-tion of natural environmental descriptors defined atboth local and regional scales in the absence of anyhuman disturbance (ii) Are we able to quantify forany tested site its deviation from a reference conditionsite having similar natural environmental conditions

Methods

-

We used data from fish surveys of 12 European countriesconducted by several laboratories and governmentalenvironmental agencies (1978ndash2002) These 5252 riverreaches or sites (Fig 1) cover most of the climatic andphysical conditions that occur in Europe

All sites had been sampled using electrofishing tech-niques (DC or PDC (pulsed direct current) waveform)during low flow periods When possible (river depthlt 0middot7 m) river reaches were sampled by wading (64middot9of all sites) For most of these sites the removal methodwas applied (87middot9) and stops nets were not used(88middot7) In large rivers (river depth gt 0middot7 m) samplingwas from boats (35middot1 of all sites) mainly in near-shore areas The size of each sampled site was sufficientto encompass complete sets of characteristic local riverhabitat For 67middot7 of all sites the whole river width wassampled In others cases the whole river section wasonly partially sampled (mainly in near-shore areas) Inorder to standardize the sampling effort we only con-sidered the first passage in all cases Although our dataare subjected to sampling noise this sampling effortwas sufficient to describe the fish assemblage For siteswhere the removal method was applied with threesuccessive passages (2275 sites) the mean percentageof the total number of species caught during the firstpassage was 91middot9 (SD 16middot3) and the mean percentageof total abundance was 63middot2 (SD 13middot1) Samplingeffort was summarized by three variables samplingtechnique (TECH boat or wading) sampling method(METH complete whole river width sampling orpartial) and fished area (FISH) We only retained onefishing occasion per site

72

D Pont

et al



43

70ndash80







73




43

70ndash80


2

10ndash99 km

2

100ndash999 km

2

1000ndash9999 km

2

gt 10 000 km

2


minus

1


minus

2ndash+16

deg



2



et al


Reproduction



Habitat


et al


Trophic guilds


et al


et al


Tolerance


et al


Migration



et al


74

D Pont

et al



43

70ndash80

(

Ns


Ns


Ni


minus

1


Ni




n


x


n


n





N




t



r

lt 0middot80 or gt

minus



t



75




43

70ndash80

Results


R

2


t

-test

P


P



P


P




Ni

-OMNI and

Ns


R

= 0middot85

Ni

-LONG and

Ns

-LONG

R

= 0middot89

Ns

-RHEO and

Ns

-LITH

R

= 0middot97

Ni

-RHEO and

Ni

-LITH

R

= 0middot81

Ns

-INSE and

Ns

-INTO

R


Ni

-INSE

Ni


Ns

-LITH

Ni


Ns

-BENT

Ns


Ns

-POTA

Ni


Ns

-INTO

Ns




Mean value(REF-MET)

Mean value(WI-MET)

Mean value(HI-MET)

P-value(REF-WI)

P-value(WI-HI)


76D Pont et al











MetricsNi-INSE

Ni-OMNI

Ni-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE







MetricsNi-INSE

Ni-OMNI

NI-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE










78D Pont et al



Discussion















Acknowledgements


References










80D Pont et al








































72

D Pont

et al



43

70ndash80







73




43

70ndash80


2

10ndash99 km

2

100ndash999 km

2

1000ndash9999 km

2

gt 10 000 km

2


minus

1


minus

2ndash+16

deg



2



et al


Reproduction



Habitat


et al


Trophic guilds


et al


et al


Tolerance


et al


Migration



et al


74

D Pont

et al



43

70ndash80

(

Ns


Ns


Ni


minus

1


Ni




n


x


n


n





N




t



r

lt 0middot80 or gt

minus



t



75




43

70ndash80

Results


R

2


t

-test

P


P



P


P




Ni

-OMNI and

Ns


R

= 0middot85

Ni

-LONG and

Ns

-LONG

R

= 0middot89

Ns

-RHEO and

Ns

-LITH

R

= 0middot97

Ni

-RHEO and

Ni

-LITH

R

= 0middot81

Ns

-INSE and

Ns

-INTO

R


Ni

-INSE

Ni


Ns

-LITH

Ni


Ns

-BENT

Ns


Ns

-POTA

Ni


Ns

-INTO

Ns




Mean value(REF-MET)

Mean value(WI-MET)

Mean value(HI-MET)

P-value(REF-WI)

P-value(WI-HI)


76D Pont et al











MetricsNi-INSE

Ni-OMNI

Ni-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE







MetricsNi-INSE

Ni-OMNI

NI-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE










78D Pont et al



Discussion















Acknowledgements


References










80D Pont et al








































73




43

70ndash80


2

10ndash99 km

2

100ndash999 km

2

1000ndash9999 km

2

gt 10 000 km

2


minus

1


minus

2ndash+16

deg



2



et al


Reproduction



Habitat


et al


Trophic guilds


et al


et al


Tolerance


et al


Migration



et al


74

D Pont

et al



43

70ndash80

(

Ns


Ns


Ni


minus

1


Ni




n


x


n


n





N




t



r

lt 0middot80 or gt

minus



t



75




43

70ndash80

Results


R

2


t

-test

P


P



P


P




Ni

-OMNI and

Ns


R

= 0middot85

Ni

-LONG and

Ns

-LONG

R

= 0middot89

Ns

-RHEO and

Ns

-LITH

R

= 0middot97

Ni

-RHEO and

Ni

-LITH

R

= 0middot81

Ns

-INSE and

Ns

-INTO

R


Ni

-INSE

Ni


Ns

-LITH

Ni


Ns

-BENT

Ns


Ns

-POTA

Ni


Ns

-INTO

Ns




Mean value(REF-MET)

Mean value(WI-MET)

Mean value(HI-MET)

P-value(REF-WI)

P-value(WI-HI)


76D Pont et al











MetricsNi-INSE

Ni-OMNI

Ni-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE







MetricsNi-INSE

Ni-OMNI

NI-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE










78D Pont et al



Discussion















Acknowledgements


References










80D Pont et al








































74

D Pont

et al



43

70ndash80

(

Ns


Ns


Ni


minus

1


Ni




n


x


n


n





N




t



r

lt 0middot80 or gt

minus



t



75




43

70ndash80

Results


R

2


t

-test

P


P



P


P




Ni

-OMNI and

Ns


R

= 0middot85

Ni

-LONG and

Ns

-LONG

R

= 0middot89

Ns

-RHEO and

Ns

-LITH

R

= 0middot97

Ni

-RHEO and

Ni

-LITH

R

= 0middot81

Ns

-INSE and

Ns

-INTO

R


Ni

-INSE

Ni


Ns

-LITH

Ni


Ns

-BENT

Ns


Ns

-POTA

Ni


Ns

-INTO

Ns




Mean value(REF-MET)

Mean value(WI-MET)

Mean value(HI-MET)

P-value(REF-WI)

P-value(WI-HI)


76D Pont et al











MetricsNi-INSE

Ni-OMNI

Ni-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE







MetricsNi-INSE

Ni-OMNI

NI-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE










78D Pont et al



Discussion















Acknowledgements


References










80D Pont et al








































75




43

70ndash80

Results


R

2


t

-test

P


P



P


P




Ni

-OMNI and

Ns


R

= 0middot85

Ni

-LONG and

Ns

-LONG

R

= 0middot89

Ns

-RHEO and

Ns

-LITH

R

= 0middot97

Ni

-RHEO and

Ni

-LITH

R

= 0middot81

Ns

-INSE and

Ns

-INTO

R


Ni

-INSE

Ni


Ns

-LITH

Ni


Ns

-BENT

Ns


Ns

-POTA

Ni


Ns

-INTO

Ns




Mean value(REF-MET)

Mean value(WI-MET)

Mean value(HI-MET)

P-value(REF-WI)

P-value(WI-HI)


76D Pont et al











MetricsNi-INSE

Ni-OMNI

Ni-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE







MetricsNi-INSE

Ni-OMNI

NI-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE










78D Pont et al



Discussion















Acknowledgements


References










80D Pont et al








































76D Pont et al











MetricsNi-INSE

Ni-OMNI

Ni-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE







MetricsNi-INSE

Ni-OMNI

NI-PHYT

Ns-BENT

Ns-RHEO

Ns-LONG

Ns-POTA

Ni-LITH

Ns-INTO

Ns-TOLE










78D Pont et al



Discussion















Acknowledgements


References










80D Pont et al














































78D Pont et al



Discussion















Acknowledgements


References










80D Pont et al














































Acknowledgements


References










80D Pont et al








































80D Pont et al






































