HAL Id: hal-00304772https://hal.archives-ouvertes.fr/hal-00304772

Submitted on 1 Jan 2003

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Investigating the influence of heavy metals onmacro-invertebrate assemblages using Partial Cononical

Correspondence Analysis (pCCA)G. Beasley, P. E. Kneale

To cite this version:G. Beasley, P. E. Kneale. Investigating the influence of heavy metals on macro-invertebrate as-semblages using Partial Cononical Correspondence Analysis (pCCA). Hydrology and Earth SystemSciences Discussions, European Geosciences Union, 2003, 7 (2), pp.221-233. �hal-00304772�

Investigating the influence of heavy metals on macroinvertebrate assemblages using Partial Canonical Correspondence Analysis

221

Hydrology and Earth System Sciences, 7(2), 221–233 (2003) © EGU

������������������ ������� ������� ������������������

����� ������������ ������ ���������������� ����

�������

Gary Beasley and Pauline E. Kneale

School of Geography, University of Leeds, Leeds LS2 9JT, UK

Email for corresponding author: P.E.Kneale@geog.leeds.ac.uk

�������

This paper defines the spectrum of impairment to stream macroinvertebrates arising from urban runoff. Field sampling of stream sediments at62 sites across Yorkshire, UK was used to investigate the influence of heavy metals and habitat on macroinvertebrate family distributionusing partial Canonical Correspondence Analysis (pCCA). Increasing urbanization and trafficking was associated with increasing levels ofmetal pollution but, even when traffic is light, family numbers can be reduced by 50%. Industrial areas and motorway runoff depressmacroinvertebrate numbers but drainage from streets with no off-road parking in residential areas can have similar impacts. The heavy metalsin the sediment accounted for approximately 24% of the variation in macroinvertebrate community composition while the physical habitatvariables used in RIVPACS (River InVertebrate Prediction And Classification System) (Wright, 2000) accounted for an additional 30%. Zincand nickel were the main metal influences regardless of the time of sampling; at these sites copper is less than critical. Results agree withthose reported in other studies in which families mainly from the orders Ephemeroptera (mayfly), Plecoptera (stonefly) and Tricoptera(caddisfly) displayed metal sensitivity in that they were absent from metal polluted streams. However, within each of these orders, a continuumof sensitivity is evident: this highlights the risks of generalising on orders rather than using family or indeed species data.

Keywords: macroinvertebrates, heavy metals, urban streams, tolerance, sensitivity

������������

Anthropogenic activities that change any catchment can leadto adverse impacts on receiving waters (Beasley and Kneale,2002). Metal pollution of sediments resulting from urbanrunoff exerts a deleterious impact on freshwatermacroinvertebrates particularly the loss of metal sensitiveorders such as Ephemeroptera (mayfly), Plecoptera(stonefly) and Tricoptera, (caddis) and acutehystopathological impairment and chronic fatality of fishspecies (Clements et al., 2000; Farag et al., 1999; Hickeyand Clements, 1998; Karouna-Renier and Sparling, 2001;Ruse and Hermann, 2000; Sriyaraj and Shutes, 2001).

Generally, increasing urbanization and road constructionmeans that heavy metals derived from non-point sourcesare likely to cause further impairment of stream ecologybut current knowledge of metal contamination is relatedprimarily to point and downstream measurements fromknown sources (Garcia-Criado et al., 1999; Gower et al.,1994, 1995; Griffith et al., 2001; Nelson and Roline, 1999),from sampling at sites where problems were anticipated(Perdikaki and Mason, 1999) and toxicity assays (Tuckerand Burton, 1999). While control measures for point sourcedischarges have improved water quality, streambedsediments remain the major repositories of urbancontaminants, the spatial ranges and degrees of which arelargely unknown.

Non-point sources of heavy metals in urban and industrialareas arise from a variety of sources. Sansalone andBuchberger (1997) identify vehicle related pollutants fromoil and tar products, wear and tear on tyres and brakes,

* This is an extended version of a paper presented at the BritishHydrological Society’s 8th National Hydrology Symposium, Birmingham,2002, and published in the Proceedings as: Beasley, G. and Kneale, P.E.,Contamination risk and in-stream ecological stress: metals, PAHs andmacroinvertebrates.

Gary Beasley and Pauline E. Kneale

222

dioxins, oxygenated compounds, halogenated phenols,metals, hydrocarbons, de-icing salts and asbestos. The decayof metal and road surfaces adds to the pollution load. Andoh(1994) and Marsalek et al. (1999) suggest that road runoffis the principal source of pollution and Pitt and Baron (1989)showed just how variable pollutants in urban surface runoffcan be (Table 1). Tyres and brakes are associated with copperand zinc corrosion, lead comes from petroleum additivesand emissions, copper and nickel from moving parts inengines, cadmium from galvanized metals, and arsenic,cadmium and copper from weed killers, fertilizers andpesticides. De-icing systems, building materials and metalobjects washed by the rain are potential sources for metalsin surface runoff.

Management of urban stream chemistry relies onidentifying, through laboratory assays, the toxicconcentrations that exert acute or chronic impairment and,using field studies, determining urban land uses that generateconcentrations exceeding these levels. Research has beenfocused on the association between heavy metals andvehicles (Andoh, 1994; Marsalek et al., 1999),predominantly on heavily trafficked roads, such asmotorways (Shutes, 1984; Maltby et al., 1995a, 1995b).Limitations of both the laboratory and field basedinvestigations have been well documented (LaPoint et al.,1984; Gower et al., 1994, 1995). To overcome deficienciesof these techniques, investigations that model macro-invertebrate assemblages from environmental variables thatinclude ‘natural’ stress parameters and contaminants havebeen used (Gower et al., 1994, 1995; Nelson and Roline,1999; Reinhold-Dudok van Heel and den Besten, 1999;Brown and May, 2000). There is a paucity of researchrelating macroinvertebrate and environmental contaminants,especially bioavailable sediment metal concentrations. Thispaper investigates which heavy metals need to be controlledbecause of their influence on macroinvertebrate communitycompositions and, by examining stream sediment metal

chemistry at 62 sites with various land uses in westYorkshire, identifies where control measures in respect ofland use should be implemented.

The study used partial Canonical CorrespondenceAnalysis (pCCA) to determine: (1) the importance ofstreambed bioavailable heavy metal concentrations indetermining macroinvertebrate assemblages in comparisonwith natural habitat characteristic influences; (2) the relativeimportance of different heavy metals on macroinvertebratecommunity compositions; and (3) the macroinvertebratefamilies that are tolerant of and those that are sensitive tometal loading.

������� ���

This study used sediment samples from 62 first-orderstreams in Yorkshire, UK (Fig. 1). Sites were selected 25 mabove and below surface storm water inflows on rural,residential, industrial and motorway land uses, not just atspecific ‘probable worst case’ sites. Each site was selectedafter careful inspection of the surface sewer to ensure thatno other discharges were present. Some sites receive runofffrom multiple land-use types, for example, motorways androad junctions (Table 2). Sites were ordered subjectivelybased on hypothesised metal contamination taking intoaccount vehicle and infrastructure density in conjunctionwith stream dimensions. The sites were sampled forsediment metal concentrations and physical environmentalcharacteristics in accordance to RIVPACS protocol (Furse,2000; Wright, 2000) in May and September 1999 and somewere resampled the following year as a spot check forconsistency of the 1999 measurements. To minimise theinfluence of recent surface runoff and to maintainconsistency between stations, samples were not taken in thedays immediately following a storm. The environmentalvariables are site altitude, distance from source, stream slope,stream width, stream depth, discharge class, percentage

Table 1. Concentrations of contaminants in runoff from urban areas (Pitt and Barron, 1989, in Novotny, 1995).

HEAVY METAL SOURCE AREAS

(µg l-1) Parking Roofs Storage Street Vehicle Landscapedareas surfaces maintenance areas parks and gardens

Cadmium 0.7 – 70 0.8 – 30 2.4 – 10 0.7 - 220 8 - 30 0.04 – 1Chromium 18 – 310 7 – 510 60 – 340 3.3 - 30 19 - 320 100 – 250Copper 20 – 770 17 – 900 30 – 300 15 – 1250 8.3 - 580 80 – 300Lead 30 – 130 13 – 170 30 – 330 30 - 150 75 - 110 9.4 – 70Nickel 40 – 130 5 –70 30 – 90 3 – 70 35 - 70 30 – 130Zinc 30 – 150 100 – 1580 66 – 290 58 - 130 67 - 130 32 – 1160

Investigating the influence of heavy metals on macroinvertebrate assemblages using Partial Canonical Correspondence Analysis

223

Table 2. Land use and site data * = Denotes a Major ‘A class’ Road contributing runoff to the stream

Land use Site Numbers

Rural, no urban or road influences A 1, 2‘Clean’ upstream land use cover A 3, 5, 7, 9, 13, 17, 20, 31, 42, 44, 48, 50, 55, 57, 61Residential � 4, 6, 8, 10, 11, 12, 14, 16*, 18, 19, 21*, 22, 23, 24, 25, 26, 27, 28, 29, 30, 32, 33,

34, 35, 36, 37, 38, 43, 46, 47, 49, 51Retail Park with extensive car park ∆ 39Industrial � 15, 40, 41, 45*, 53, 54Motorway � 52, 54, 56, 58, 59, 60, 62

A Pristine, � Residential � Motorway ∆ Retail, � Industrial

Fig. 1. Location of sampling sites

Gary Beasley and Pauline E. Kneale

224

particle cover from silt to boulders, dissolved oxygen,electrical conductivity and pH. The streambed sedimentchemistry and macroinvertebrate numbers did not varysignificantly or consistently between sampling dates.

����������������� ����� �

Random triplicate sediment samples were collected fromthree distinct microhabitats within each sampling reach andintegrated to a composite sample of approximately 1 kg.The bed material was inspected over a distance of 7 metresand samples were taken from different points to get arepresentative mix of material sizes, from both protectedand more open stream sites. This sampling strategyovercame the heterogeneity of heavy metal concentrationswhile simultaneously reducing sampling variance (Argyrakiet al., 1995). Heterogeneity of the sample was of moreconcern in the horizontal plane as pilot studies showed that,in these headwater streams, the macroinvertebrates are rarelyfound below a depth of 5 cm; indeed in some sections thebedrock is even closer to the surface. Sediment was collectedfrom the near-surface, (0–5 cm) using a plastic trowel thatproduced a sample with minimal streambed disturbance, lowrisk of contamination and minimal loss of the finest particlesthat generally possess the highest metal concentrations.Rinsing the trowel within the flow downstream of eachsampling point ensured quality control and the preventionof cross contamination. The sediment was placed in anairtight zip sealed polyethylene bag and double bagged tosafeguard against cross contamination.

���� ������������ ����� �

Three one-minute kick samples were collected fromapproximately 3 m2 of streambed to a depth of 5 cm, withthe net (25 cm × 20 cm × 30 cm with 1 mm mesh) heldvertically and the frame at right angles to the current. Toavoid cross contamination between sampling stations inclose proximity, sampling proceeded from downstream toupstream sites. Sampling three microhabitats maximized thechance of collecting a more complete assemblage. Sampleswere decanted into 1100 ml polypropylene bottles with justenough water to keep the sample damp so as to reducedamage and retard the activities of carnivores duringtransportation.

��� ��� ���������

Heavy metalsThe sediment was oven dried at 30°C for five days and thenlaid out in a dust free laboratory for a further seven daysbefore representative sub-samples were taken using coning

and quartering. In preparation for sieving, each sub-samplewas ground lightly using an agate mortar and pestle. Eachsub-sample was passed through a 2000µm synthetic nylonpolymer woven screen to retain only the sand and silt sizedparticles. After each screening, the nylon was rinsedthoroughly in deionised water. Analytical-grade (AnalaR)acids were used for all extraction solutions and cleaningprocedures to minimize metal contamination.

Bioavailable metal concentrations (Exchangeable andbound to carbonates) were determined using the Standard,Methods and Testing (SM&T) three step sequentialextraction technique reported by Quevauvillier et al. (1997).Metal content of the streambed sediment followingextraction was determined using Inductively CoupledPlasma Optical Emission Spectrometry (IAP-AES). Themetals sought in these analyses were cadmium, chromium,copper, iron, lead, nickel and zinc. Beasley (2001) detailsANOVA (Analysis of Variance) evaluations of relationshipsbetween particle size, extraction levels and heavy metals.

MacroinvertebratesMacroinvertebrates were preserved within five hours ofcollection using 95 percent ethanol and were sorted withina month, following the recommended standard procedurefor RIVPACS (River InVertebrate Prediction andClassification System)(Environment Agency, 1997). Thesamples were first placed in large white trays to ease sortingand specimens were placed in petri dishes for identificationto family level. Examples of each taxon were put in vialscontaining ethanol for quality assurance checking withEnvironment Agency (EA) scientists. The abundances ofeach taxon were recorded using EA audit sheets.

����������������

To determine the relative importance of environmentalfactors (metals and habitat) in explaining the variability incomposition of the macroinvertebrate community, partialcanonical correspondence analysis (pCCA) was used(ter Braak, 1987, 1994). Prior to analysis, the set ofexplanatory variables was subdivided into a set ofcovariables and a set of variables-of-interest. Thecovariables, which in this instance represent habitatvariables, are not the prime focus of the research and assuch did not enter the synthetic gradients (ter Braak andVerdonschot, 1995). The variables-of-interest, theconcentrations of heavy metals in the streambed sediments,are the remaining explanatory variables that construct thesynthetic gradients. The analysis follows that of CCA butwith the added requirement that each synthetic gradient mustbe uncorrelated with the covariables. Consequently, the

Investigating the influence of heavy metals on macroinvertebrate assemblages using Partial Canonical Correspondence Analysis

225

covariables represent gradients that have already beenextracted. The resultant ordination diagram displays theunimodal relationships between macroinvertebrates and thevariables-of-interest after the effects of the covariables havebeen partialled out (ter Braak, 1996).

The analysis used the programme CANOCO 4.0. All datasets were first converted into a CANOCO 4.0 format usingthe utility programme CanoImp. Macroinvertebrate families(response variables), environmental (predictors) andcovariable (concomitant variables) data were selected foranalysis using direct gradient analysis. While dissolvedoxygen, electrical conductivity and the pH of the watercolumn can vary with local conditions and macro-invertebrates respond to events at other times in the season,the consistency in values between May and September andthe decision not to sample during and in the days followingrain offers confidence that these data are valid. The datafiles inputted to each pCCA run are shown in Table 3. Run1 uses the May, run 2 the September and run 3 the combined1999 data set.

The unimodal response model was selected for this studybecause macroinvertebrates exhibit maximum abundancearound optimal conditions and because themacroinvertebrate family data files contained a large numberof zero values (ter Braak and Smilauer, 1998). PCCA wasused in this research to investigate the effects of thestreambed contaminants while acknowledging theimportance of the environmental data. Scaling focused uponinter-family distances as interpretation among families wasthe aim of the analysis using biplot scaling. All familyabundance values were log transformed to avoid undueinfluence of outliers on the ordination. Downweighting forrare families was not adopted.

Each run was completed by removing variables with highinflation factors, which permitted the ranking ofenvironmental variables in the order of their importance for

determining the macroinvertebrate families data using the‘forward selection’ option in CANOCO. Forward selectionwas also used to reduce to 10 the number of environmentalvariables to improve the clarity of the ordination diagrams.In ‘automatic selection’, the K best variables are selectedsequentially on the basis of maximum extra fit. The statisticalsignificance of each variable selected is judged by a Monte-Carlo permutation test.

The data are first displayed as rankings (Tables 4-6) andexplored further using ordination diagrams (Figs. 2-7). InTables 4-6, the top ten ranked environmental variables arepresented in terms of importance in explaining communitycomposition. These rankings and their statistical significanceindicate which of the elements exerts the greatest influenceon macroinvertebrate community structures. These data arethen plotted as ordination diagrams using the programmeCanoDraw 3.1, graphically representing the communitystructure and the community response to the environmental

Table 3. Details of each pCCA model run.

Model runs Data files Constituents

pCCA Run 1 MACMAY Macroinvertebrate Abundances (May)METMAY Metals & Water Chemistry Concentrations (May)COMAY Covariables (May)

pCCA Run 2 MACSEP Macroinvertebrate Abundances (September)METSEP Metals & Water Chemistry Concentrations (September)COSEP Covariables (September)

pCCA Run 3 MACMIX Macroinvertebrate Abundances (May & Sept)METMIX Metals & Water Chemistry Concentrations (May & Sept)COMIX Covariables (May & Sept)

Table 4. Top 10 rankings for heavy metal concentrations inMay (extract 1) using unrestricted Monte Carlo significancetest (p<0.05).

Variable Lambda-A F p

Zinc 0.10 3.31 0.005*Dissolved Oxygen 0.07 2.34 0.005*Electrical conductivity 0.05 1.76 0.020*Nickel 0.05 1.85 0.025*Lead 0.05 1.94 0.010*pH 0.03 1.01 0.415Iron 0.04 1.19 0.295Copper 0.02 0.91 0.470Chromium 0.02 0.64 0.900Cadmium 0.02 0.61 0.885

Gary Beasley and Pauline E. Kneale

226

Table 5. Top 10 rankings for heavy metal concentrations inSeptember (extract 1) using unrestricted Monte Carlosignificance test (p<0.05).

Variable Lambda-A F p

Nickel 0.08 2.01 0.020*Zinc 0.06 1.50 0.045*pH 0.06 1.31 0.115Lead 0.04 1.05 0.410Copper 0.04 0.91 0.515Electrical Conductivity 0.03 0.92 0.555Dissolved Oxygen 0.04 1.00 0.470Cadmium 0.05 1.20 0.260Chromium 0.03 0.66 0.900Iron 0.02 0.53 0.945

Table 6. Top 10 rankings for heavy metal concentrations(extract 1) using unrestricted Monte Carlo significance tests(p<0.05). Data from May and September.

Variable Lambda-A F p

Zinc 0.10 3.36 0.005*Nickel 0.09 3.14 0.005*Electrical Conductivity 0.05 1.54 0.060pH 0.04 1.45 0.075Iron 0.03 1.26 0.180Lead 0.04 1.17 0.280Copper 0.03 1.02 0.410Dissolved Oxygen 0.02 0.82 0.695Chromium 0.02 0.68 0.810Cadmium 0.01 0.53 0.965

Fig. 2. Site - environment biplot based on pCCA run 1.

Fig. 3. Site - environment biplot based on pCCA run 2.

variables. Modifications to the ordination diagrams in termsof improvement in clarity were made using the programmeCanoPost 1.0. Interpretation of the ordination diagramsfacilitates the ranking of sites in terms of communitycomposition in relation to each element and ofmacroinvertebrate families with respect to tolerance andsensitivity to each element.

���� ��

Tables 4 to 6 show that despite changes in rank orderbetween data sets, zinc and nickel exert a significantinfluence on the composition of macroinvertebratecommunities in the study streams. For each pCCA runcontaining trace metals, zinc and nickel are ranked highly

on three occasions confirming that these metals are widelydistributed in urban runoff. Moreover, at the concentrationsrecorded in this study, these results show that zinc is themajor heavy metal determinant of community compositionin these streams. The only other metal to exert a significantinfluence on community structure is lead in May (pCCArun 1 Table 4).

Investigating the influence of heavy metals on macroinvertebrate assemblages using Partial Canonical Correspondence Analysis

227

-1.0 +1.0-1.0

+1.0

NipH DOEc Fe

Pb

CuZn

CrCd

PhilPerl

Plan

Hept

Leuc

Rhya

Odon

Ephemeri

Hydrops

Valv

Hydrom

Simu

HydrobDyti

Hydroph

Nemo

Poly

Spha

Elmi

Hali

Chlo

Ancy

LymnaChir

Limne

Asel

Seri

Erpo

Gloss

Phys

Tipu

Ephemere

Gamm Olig

Leptop

Baet

-1.0 +1.0

-1.0

+1.0

Zn

DO

Cd

Pb

Cr

Cu

pH

Ni

Ec

Fe

Phil

Hali

Leptoc

Phys

Ephemeri

Leptop

Hept

Rhya

Erpo

Perl

NemoLymna

Seri

HydropsElmi

Hydrob

Asel

Leuc

Ephemere

Gloss

Tipu

Spha

Chir

Chlo

OligSimu

GammLimne

Dyti

Baet

Hydroph

Ancy

Poly

Valv

Fig. 4. Site - environmental biplot based on pCCA run 3.

Fig. 5. Macroinvertebrate families - environment biplot based onpCCA run 1. Macroinvertebrate families (see Table 10) are shown ascrosses.

-1.0 +1.0

-1.0

+1.0

Zn

CdCu

Pb

PH

CrEC

Ni

DO

Fe

Phil

HydromHept

Leptop

Perl

Leptoc

Rhya

Asel

Elmi

NemoErpo

OdonPhys

HydropsLeuct

SeriChlor

Lymna

Ephemeri

Hydroph

Spha

Ephemere

TipuHali

Hydrob

ChirGamm

BaetGloss

Dyti

Olig

Ancy

Simu

Valva

Limne

Poly

Fig. 6. Macroinvertebrate families - environment biplot based onpCCA run 2. Macroinvertebrate families are shown as crosses.

Fig. 7. Macroinvertebrate families - environmental biplot based onpCCA run 3. Macroinvertebrate families are shown as crosses.

����� ��������

Further detail can be extracted by inspecting the pCCAordination diagrams. In each, site and macroinvertebratefamily data are represented by points, environmentalvariables by arrows. The ordination diagrams display,simultaneously, the main patterns of community variations,

in so far as these reflect environmental variation and themain pattern of the tolerances of macroinvertebrate familieswith respect to the environmental variables. In Figs. 2–4,each site is located with respect to the sediment chemistryand site characteristic variables. In Figs. 5–7,macroinvertebrate family points correspond to their

Gary Beasley and Pauline E. Kneale

228

approximate optima in the two-dimensional environmentalsubspace based on their weighted average, which indicatesthe centre of a macroinvertebrate family distribution alongan environmental variable (ter Braak and Looman, 1986;ter Braak, 1986; ter Braak and Prentice, 1988). Differencesin weighted averages between macroinvertebrate familiesindicate differences in their tolerances along thatenvironmental variable. Environmental variables arerepresented by arrows, which point in the direction ofmaximum change of that variable across the ordinationdiagram. The length of the arrows is proportional to therate of change in this direction. Environmental variableswith long arrows display a stronger correlation with theordination axes than those with short arrows, as signifiedby the co-ordinates of the arrow head. Environmentalvariables that are strongly correlated with the ordinationaxes are more closely related to the pattern of communityvariation shown in the ordination diagram (ter Braak, 1987).The rule for quantitative interpretation is that each arrowrepresenting an environmental variable determines adirection or axis in the diagram on to which sites andmacroinvertebrate family points are projected. Sites ormacroinvertebrate families with their perpendicularprojection endpoints near to or beyond the tip of an arrowwill be strongly positively correlated with and influencedby the environmental variable represented by that arrow.Those sites or macroinvertebrate family whose projectionslie near the origin will be less strongly affected.

�������� ����� ����� ��������

Given the broad within and between site variability inenvironmental conditions both temporally and spatially, itwould be surprising if the pCCA results were clear-cut.Figures 2 to 4 show the results for the runs of sites in relationto metal and environmental variables and no clear sitegrouping along either ordination axis is evident. This isconsistent with the findings of Gower et al. (1995), whoexamined the impact of mine drainage on macroinvertebrateassemblages by comparing clean control sites withcontaminated mine sites; as there were no intermediate sites,clear site groupings could be distinguished. However, inthe present study, the temporal and graduated spatialvariations in environmental conditions are very differentfrom Gower’s.

The strong influence of nickel and zinc can be seenvisually on the metal ordination diagrams with their arrowhead co-ordinates signifying strong correlation with eitherordination axis 1 or 2 (Figs. 2–4).

In the current research, the data fall along a continuumwith a wide range of contamination levels arising from

diverse sources, land usage and variations in catchmentcharacteristics and history. The plots show definite trendsrelating land use to environmental variables andmacroinvertebrate family composition. Figures 2 and 4 showthat metal tolerant communities are associatedpredominantly with streams receiving runoff from motorwayor industrial land uses. This is reflected in the top fiverankings in Tables 7 and 9. The strong relationships withzinc and nickel suggest that these are major determinantsof community composition in these streams (Tables 4 and6). Sites in the lower left and upper right of Figs. 2 and 4,respectively, possess communities sensitive to metalcontamination. Hence, these sites are either clean referencesites or low-density suburban sites (Tables 7 and 9).

In Tables 7, 8 and 9, the sites with tolerantmacroinvertebrate communities include the sites below theindustrial and motorway runoff points but these do not totallydominate the data. Suburban residential sites are alsorepresented, particularly suburban terraces or councilhousing where there is no off-street parking. Although roadtraffic is light, the overnight parking of vehicles on bothsides of the road is associated with enhanced metal levelsin the sediments. Drainage from the carriageway in a ruralarea below a four-way junction with traffic lights, whereheavy traffic is braking sharply or accelerating to get uphill,has also enhanced metal levels as has a site which drains alay-by used by lorries. In contrast, the sites with the broadestmacroinvertibrate communities and the more sensitiveindicator species are typically rural and residential areaswhere each property includes a garage so that on-streetparking is rare.

Figure 3 shows a slightly different ordination of sites for

Table 7. Top and bottom 5 ranked sites (left to right) in termsof communitiy composition for pCCA run 1 (May data).

Ranked variable Top 5 sites : Bottom 5 sites :tolerant sensitivecommunities communities

Zinc � � � A � � � A A ADissolved Oxygen � � A � � A � � � �Electrical Conductivity � � � � A � A A � �Nickel � � � � � � A A � �Lead � � � A � � A A � �pH � � � � � A � � � AIron � � � A � � A A � �Copper � � � A � �A A � �Chromium � � � A � � A � A ACadmium � � � A � � A � A A

A Pristine, � Residential � Motorway ∆ Retail, � Industrial

Investigating the influence of heavy metals on macroinvertebrate assemblages using Partial Canonical Correspondence Analysis

229

September in relation to the environmental variables; thesereflect seasonal changes in family composition andenvironmental conditions at the sites. Communities in sitesat the top of the diagram are metal tolerant whilst those atthe bottom are metal sensitive. When individual sitelocations are inspected, the ‘changes in position’ aregenerally small, except downstream from motorway sites(Table 8) where the community composition in Septemberis less tolerant of heavy metals than in May. Nevertheless,the broad grouping of sites having tolerant or sensitivecommunities does not change with season; the familypatterns observed remain broadly stable from spring toautumn although the total numbers of individuals arereduced in September (Beasley, 2001). The influence of zinc and nickel remains high as shownin Tables 3–5. The communities at sites with low metalconcentrations plot consistently with those in Figs. 2 and 4confirming that differences between seasons are small.

Each pCCA run produces slightly different relationshipswithin the ordination diagrams but a common patternemerges. Pollution indicator families are identified from theordination diagrams for each contaminant and the five mosttolerant and sensitive families is presented for each (Tables11-13). Figure 5 plots metal and acid tolerant invertebratesin the upper right section of the plot, with metal sensitivefamilies in the lower left section. Although the rank orderdiffers slightly between metals, many of the same familiesappear in each set of rankings. The families most frequentlyidentified as tolerant are Hydrophilidae, Asellidae,Ephemerellidae, Philoptamidae and Chloroperlidae whilethe families identified as metal sensitive, Leptophlebiidae,Ephemeridae, Leuctridae, Hydrobiidae and Valvatidae aregenerally absent from streams with elevated concentrationsof one or more of the metals. These results are in agreementwith those found in other studies in which families from theorders Ephemeroptera, Tricoptera and Plecoptera were

Table 10. Abbreviations of macroinvertebrate family names used in Tables 11-13 and Figs. 5-7.

Ancy Ancylidae Hali Haliplidae Odon OdontoceridaeAsel Asellidae Hept Heptageniidae Olig OligachaetaBeat Beatidae Hydrob Hydrobiidae Perl PerlodidaeChir Chironomidae Hydrom Hydrometridae Phil PhilopotamidaeChlo Chloroperlidae Hydroph Hydrophilidae Phys PhysidaeDyti Dytiscidae Hydrops Hydropsychidae Plan PlanorbidaeElmi Elmidae Leptoc Leptoceridae Poly PolycentropodidaeEphemere Ephemerellidae Leptop Leptophlebiidae Rhya RhyacophilidaeEphemeri Ephemeridae Leuc Leuctridae Simu SimulidaeErpo Erpobdellidae Limne Limnephilidae Spha SphaeriidaeGamm Gammaridae Lymna Lymnaeidae Tipu TipulidaeGloss Glossiphonidae Nemo Nemouridae Valv Valvatidae

Table 8. Top and bottom 5 ranked sites (left to right) interms of community composition for pCCA run 2(September data).

Ranked variable Top 5 sites : Bottom 5 sites :tolerant sensitivecommunities communities

Nickel A � A � � � A ∆ � AZinc � A � � � A A A A ApH � A � � A A A � � �Lead � � � A � A A A A �Copper � A � � � A A A A �Electrical Conductivity � � � � � A � A � ADissolved Oxygen � A ∆ A � A � A � �Cadmium � A � � � A A A A AChromium � A � � � A A A A �Iron A � � A � � A ∆ � A

Table 9. Top and bottom 5 ranked sites (left to right) interms of community composition for pCCA run 3 (all data).

Ranked variable Top 5 sites : Bottom 5 sites :tolerant sensitivecommunities communities

Zinc � � A � � A � A A �Nickel � A � A A � � A A �Electrical Conductivity � � � A � � A A A �pH � � � � � A A A � AIron � � � A � A A � � ALead � � � � A A A � A �Copper � � � A � A A � A �Dissolved Oxygen � � � � � A A A � AChromium � � � A � � A A A �Cadmium � � � A � A A � � A

A Pristine, � Residential � Motorway ∆ Retail, � Industrial

A Pristine, � Residential � Motorway ∆ Retail, � Industrial

Gary Beasley and Pauline E. Kneale

230

Table 11. Top and bottom 5 ranked weighted averages of families in relation to the top 10 ranked environmental variablesfor run 1 (May data).

Variable Five most tolerant families Five most sensitive families

Zinc Asel, Lymna, Erop, Hydroph, Chlo Leptoc, Phil, Hali, Ephemer, LeptopDissolved Oxygen Erpo, Lymna, Asel, Gloss, Spha Phil, Hali, Leptoc, Phys, EphemeriElectrical Conductivity Hydroph, Asel, Ephemere, Lymna, Chlo Leptoc, Ephemeri, Leuc, Valv, HydrobNickel Phil, Hydroph, Asel, Ephemere, Chlo Leptoc, Leuc, Ephemeri, Valv, HydrobLead Asel, Hydroph, Lymna, Chlo, Ephemere Leptoc, Ephemeri, Leuc, Leptop, HydrobpH Erpo, Lymna, Valva, Chir, Gamm Phil, Hali, Phys, Hept, PerlIron Asel, Hydroph, Lymna, Chlo, Gloss Leptoc, Ephemeri, Leuc, Leptop, HydrobCopper Hydroph, Asel, Lymna, Chlo, Ephemere Leptoc, Ephemeri, Leuc, Leptop, HydrobChromium Erpo, Lymna, Asel, Gloss, Chlo Phil, Leptoc, Hali, Ephemeri, PhysCadmium Asel, Lymna, Hydroph, Chlo, Erpo Leptoc, Phil, Ephemeri, Leptop, Phys

Table 12. Top and bottom 5 ranked weighted averages for families in relation to the top 10 ranked environmental variablesfor run 2 (September data).

Variables Five most tolerant families Five most sensitive families

Nickel Phil, Perl, Chlo, Rhya, Hept Plan, Ephemeri, Valv, Hydrob, DytiZinc Chlo, Phys, Hali, Hydroph, Rhya Hydrom, Plan, Nemo, Ephemeri, EphemerepH Plan, Ephmeri, Valv, Phys, Hali Perl, Phil, Hydrom, Leuc, HeptLead Chlo, Phys, Hali, Hydroph, Rhya Hydrom, Nemo, Plan, Ephemeri, EphemereCopper Chlo, Phys, Hali, Hydroph, Rhya Hydrom, Plan, Nemo, Ephemeri, EphemereElectrical Conductivity Chlo, Phys, Hali, Plan, Lymna Perl, Phil, Hydrom, Nemo, LeucDissolved Oxygen Plan, Ephemeri, Valv, Hydrob, Dyti Plan, Peri, Chlo, Rhya, HeptCadmium Chlo, Phys, Hali, Rhya, Hydroph Hydrom, Plan, Ephemeri, Nemo, EphemereChromium Chlo, Phys, Rhya, Hydroph, Hali Hydrom, Plan, Ephemeri, Nemo, EphemereIron Phil, Perl, Chlo, Rhya, Hydroph Plan, Ephemeri, Hydrom, Valv, Limne

Table 13. Top and bottom 5 ranked weighted averages for families in relation to the top 10 ranked environmental variablesfor run 3 (all data).

Variables Five most tolerant families Five most sensitive families

Zinc Spha, Lymna, Asel, Erpo, Simu Leptoc, Hydrom, Ephemeri, Leptop, OdonNickel Phil, Spha, Perl, Hydrom, Hept Leptoc, Ephemeri, Phys, Leptop, LimneElectrical Conductivity Spha, Lymna, Asel, Erpo, Simu Leptoc, Hydrom, Ephemeri, Leptop, PhilPH Asel, Erpo, Phys, Chlo, Lymna Phil, Hydrom, Hept, Leptop, PerlIron Leptoc, Ephemeri, Phys, Limne, Dyti Phil, Hydrom, Hept, Perl, RhyaLead Spha, Lymna, Asel, Erpo, Chlo Phil, Leptoc, Hydrom, Leptop, EphemeriCopper Spha, Lymna, Asel, Erpo, Simu Leptoc, Hydrom, Phil, Leptop. EphemeriDissolved Oxygen Phil, Hydrom, Hept, Perl, Leptop Phil, Hydrom, Hept, Leptop, PerlChromium Spha, Lymna, Asel, Erpo, Simu Leptoc, Hydrom, Phil, Leptop, EphemeriCadmium Spha, Lymna, Asel, Erpo, Simu Leptoc, Hydrom, Ephemeri, Leptop, Phil

Investigating the influence of heavy metals on macroinvertebrate assemblages using Partial Canonical Correspondence Analysis

231

absent from metal polluted streams. (Whiting and Clifford,1983; Casper, 1994; Gower et al., 1994, 1995)

In Fig. 6, metal tolerant families are located in the upperhalf and metal sensitive in the lower half of the diagram. InFig. 7 tolerant families and sensitive families are located inthe lower left and upper right sections respectively. Themetal sensitive families are largely the same as thoseidentified in May. However, the families tolerant inSeptember and in the combined data set (May andSeptember) differ somewhat, (Tables 11 and 12) perhapsbecause of differences in the life cycle between families,when abundance varies naturally between seasons. As withcommunity composition, inspection of the three familyordination diagrams for heavy metals shows that the samefamilies are plotted in roughly the same ordination space inrelation to the heavy metal variables. From these plots andthe rankings, several overall indicator families of metaltolerance emerge, notably, Aselidae, Ephemerellidae,Hydrophilidae, Lymnaeidae, Sphaeridae, Physidae,Haliplidae, Rhyacophilidae, Erpobdellidae and Simuliidae.Similarly, despite the fact that rankings change slightlybetween runs and between metals, Leptoceridae,Ephemeridae, Leuctridae, Valvatidae, Hydrobiidae,Hydrometridae, Planorbidae, Nemouridae, Leptophlebiidaeand Odontoceridae are usually absent from streams withelevated concentrations of one or more of the heavy metals.

����������

The results of the pCCA show that heavy metals accountfor approximately 24% of the variation in macroinvertibratecommunity composition while the physical variables inRIVPACS account for a further 30%. Of course, both bioticand abiotic factors affect the distribution of macro-invertebrates in streams and some 50% of the overallvariation is unexplained. Many possible reasons have beenexplored elsewhere (Whiting and Clifford, 1983; Maltby etal., 1995a); they include altered substrate compositioncaused by an influx of fine particulates, altered flow regimeas a consequence of bridges and channelisation, changes infood availability, differences in surrounding terrestrialhabitats and additional contaminants such as PAHs(polycyclic aromatic hydrocarbons) form vehicle emissionsand other sources (Beasley and Kneale, 2002).

On the relative importance of different heavy metals andhabitat on macroinvertebrate community compositions, itis often difficult to know which of the metal elements hasthe greatest effects on specific macroinvertebrates since theyoften occur in high concentrations simultaneously. However,pCCA was used, successfully, to identify zinc and nickel asthe main metal influences regardless of the time of sampling.

This must be compared with the findings of Armitage (1980)and Malmqvist and Hoffsten (1999) that copper and zinchad the strongest negative effects on taxonomic richness.Indeed, Armitage (1980) concluded that Ephemeropteraspecies were particularly sensitive to zinc because they wererestricted to sites with zinc concentrations below300 µm l–1. Copper has been blamed for changes incommunity structure in streams elsewhere (Gower et al.,1994; Nimmo et al., 1996) but, in the present research,copper exerted no significant influence on communitycomposition. It may be that, in these bed sediments,bioavailable concentrations of copper are generally belowtheir toxic threshold.

In identifying the macroinvertebrate families that aretolerant and sensitive to metal loading, the pCCA analysesdemonstrate that Ephemeroptera families are particularlysensitive to elevated metal levels; this agrees with the resultsof Kiffney and Clements (1994), Gower et al. (1994, 1995),Schultheis et al. (1997), Malmqvist and Hoffsten (1999)and Clements et al. (2000). The present research foundLeptoceridae (caddis flies) and Ephemerellidae (mayflies)particularly sensitive because of their absence from streamswith moderate metal contamination. The widespreaddistribution of Baetidae (mayflies) and their central positionin the families-environment ordination diagrams (Figs. 5–7) supports the findings of Gower et al. (1994, 1995) thatthey have moderate metal tolerance. Similarly, results fromthe pCCA analysis show that the tolerance of Plecopterafamilies (stoneflies) varies; Leuctridae are slightly moretolerant than Nemouridae, and Chloroperlidae appear to beparticularly tolerant to metals. Metal tolerance of Tricopterafamilies (caddis flies). is also low, especiallyPolycentropodidae, Rhyacophilidae, and Hydropsychidae;the latter two families prefer high dissolved oxygenconditions (Figs. 5–7). More sensitive are the cased caddis,Limnephilidae and Sericostomatidae. Asselidae (waterlice)are identified in the present research as being tolerant ofsome heavy metals as has been found by Brown (1976) andGower et al. (1994). However, within each family, tolerancesvary between species, illustrating the importance of speciesidentification.

���� ������

pCCA analysis has been shown to be capable ofdiscriminating, usefully, between sites and familyrelationships with metal pollution loading. As anticipated,there is a loss in family numbers as land use changes fromrural to suburban to urban. The spatial distribution of thesites and the diversity of land uses demonstrates the valueof this type of monitoring for forecasting and for indicating

Gary Beasley and Pauline E. Kneale

232

streams where the ecology is potentially particularlyvulnerable to metal-rich runoff. Adding parallel informationfor other surface runoff contaminants (Sansalone andBuchberger, 1997) such as petrol, oil and tar products,dioxins, oxygenated compounds, halogenated phenols,hydrocarbons, de-icing salts and asbestos is likely toimprove the level of explanation in the model. Detailedmodelling of such sources and compounds offers asignificant arena for future research.

����������Andoh, R.Y.G., 1994. Urban runoff: nature, characteristics and

control. J. Inst.Water Environ. M anage., 8, 371–378.Argyraki, A., Ramsey, M.H. and Thompson, M., 1995. Proficiency

testing in sampling: Pilot study on contaminated land. Analyst,120, 2799–2803.

Armitage, P.D., 1980. The effects of mine drainage and organicenrichment in benthos in the river Nent system, northernPennines. Hydrobiologia, 74, 119–128.

Beasley, G.E., 2001. Investigating the relations between streambedsediment contaminants and macroinvertebrate assemblages inurban headwater streams in Yorkshire. Unpublished PhD thesis,University of Leeds, UK.

Beasley, G. and Kneale, P.E., 2002. Reviewing the impact of metalsand PAHs on macroinvertebrates in urban watercourses. Prog.Phys.Geog., 26, 236–270.

Brown, B.E., 1976. Observation on the tolerance of the isopodAsellus meridianus Rac. to copper and lead. Water Res., 10,555–559.

Brown, L. R. and May, J. T., 2000. Macroinvertebrate assemblageson woody debris and their relations with environmental variablesin the Lower Sacramento and San Joaquin River drainages,California. Environ. Monit. Assess., 64, 211–239.

Casper,A. F., 1994. Population and community effects of sedimentcontamination from residential urban runoff on benthicmacroinvertebrate biomass and abundance. Bull. Environ.Contam.Toxicol., 53, 796–799.

Clements, W.H., Carlisle, D.M., Lazorchak, J.M. and Johnson,P.C., 2000. Heavy metals structure benthic communities inColorado mountain streams. Ecol. Appl., 10, 626–638.

Environment Agency 1997. Assessing Water Quality – GeneralQuality Assessment (GQA) Scheme for Biology, EnvironmentAgency, Bristol, UK.

Farag, A.M., Woodward, D.E., Brumbaugh, W., Goldstein, J.N.,MacConnell, E., Hogstrand, C. and Barrows, F.T., 1999. Dietaryeffects of metals-contaminated invertebrates from the Coeurd’Alene River, Idaho, on cutthroat trout. Trans. Amer. Fish. Soc.,128, 578–592.

Furse, M.T., 2000. The application of RIVPACS procedures inheadwater streams – an extensive and important nationalresource. In: Assessing the biological quality of fresh waters:RIVPACS and other techniques, The Freshwater BiologicalAssociation, Ambleside, UK, 77–91.

Garcia-Criado, F., Tome, A., Vega, F.J. and Antolin, C., 1999.Performance of some diversity and biotic indices in riversaffected by coal mining in northwestern Spain. Hydrobiologia,394, 209–217.

Gower, A.M., Myers, G., Kent, M. and Foulkes, M.E., 1994.Relationships between macroinvertebrate communities andenvironmental variables in metal-contaminated streams in south-west England. Freshwater Biol., 32, 199–21.

Gower, A.M., Myers, G., Kent, M. and Foulkes, M.E., 1995. Theuse of marcroinvertebrate assemblages in the assessment ofmetal – contaminated streams. In: The Ecological Basis for RiverManagement, D.M. Harper and A.J.D. Ferguson, (Eds.), Wiley,Chichester, UK, 181–191.

Griffith, M.B., Kaufmann, P.R., Herlihy, A.T. and Hill, B.H., 2001.Analysis of macroinvertebrate assemblages in relation toenvironmental gradients in Rocky Mountain streams. Ecol.Appl., 11, 489–505.

Hickey, C.W. and Clements, W.H., 1998. Effects of heavy metalson benthic macroinvertebrate communities in New Zealandstreams. Environ. Toxicol. Chem., 17, 2338–2346.

Karouna-Renier, N.K. and Sparling, D.W., 2001. Relationshipsbetween ambient geochemistry, watershed land-use and tracemetal concentrations in aquatic invertebrates living instormwater treatment ponds, Environ. Pollut., 112, 183–192.

Kiffney, P.M. and Clements, W.H., 1994. Structural responses ofbenthic macroinvertebrate communities from different streamorders to zinc. Environ. Toxicol. Chem., 13, 389–395.

LaPoint, T.W., Melancon, S.M. and Morris, M.K., 1984.Relationships among observed metal concentrations, criteria andbenthic community structural responses in 15 streams. J. WaterPollut. Control Fed., 56, 1030–1038.

Malmqvist, B. and Hoffsten, P., 1999. Influence of drainage fromold mine deposits on benthic macroinvertebrate communitiesin central Swedish streams. Water Res., 33, 2415–2423.

Maltby, L., Boxall, A.B.A., Forrow, D.M., Calow, P. and Betton,C.I., 1995a. The effects of motorway runoff on freshwaterecosystems: 2. Identifying major toxicants. Environ. Toxicol.Chem., 14, 1093–1101.

Maltby, L., Forrow, D.M., Boxall, A.B.A., Calow, P. and Betton,C.I., 1995b. The effects of motorway runoff on freshwaterecosystems: 1. Field study. Environ. Toxicol. Chem., 14,.1079–1092.

Marsalek, J., Rochfort, Q., Brownlee, B., Mayer, T. and Servos,M., 1999. An explanatory study of urban runoff toxicity. WaterSci. Technol., 39, 33–39.

Nelson, S.M. and Roline, R.A., 1999. Relationships betweenmetals and hyporheic invertebrate community structure in a riverrecovering from metals contamination. Hydrobiologia, 397,211–226.

Nimmo, D., Lafrancios, T., Willox, M., Farag, I. and Woodward,D., 1996. Macroinvertebrate metrics, toxicity studies and metalsanalysis to assess impacts of mine tailings above YellowstoneNational Park. Bull. N. Amer. Bentholog. Soc., 13, 144–145.

Novotny, V., 1995. Diffuse sources of pollution by toxic metalsand impact on receiving waters. In : Heavy Metals, problemsand solutions, W. Salomons, V. Fostner, and P. Mader, P., (Eds.),Springer, New York, 33–52.

Perdikaki, K. and Mason, C.F., 1999. Impact of road run-off onreceiving streams in Eastern England. Water Res., 33, 1627–1633.

Pitt, R. and Barron, P., 1989. Assessment of urban and industrialstormwater runoff toxicity and the evaluation/development oftreatment for runoff toxicity abatement phase. A report to theUS EPA, Edison, New York.

Quevauviller, P.H., Rauret, G., Lopez-Sanchez, J.F., Rubio, R.,Ure, A. and Muntau, H., 1997. The Certification of the EDTA-Extractable Contents (Mass Fractions) of Cd, Cr, Ni, Pd andZn in Sediment Following a Three-Step Sequential ExtractionProcedure – CRM 601, Office for Official Publications of theEuropean Communities, Luxembourg.

Reinhold-Dudok van Heel, H.C. and den Besten, P.J. 1999. Therelation between macroinvertebrate assemblages in the Rhine-Meuse Delta (the Netherlands) and sediment quality. Aquat.Ecosyst. Health Manage., 2, 19–38.

Investigating the influence of heavy metals on macroinvertebrate assemblages using Partial Canonical Correspondence Analysis

233

Ruse, L.P. and Herrmann, S.J., 2000. Plecoptera and Trichopteraspecies distribution related to environmental characteristics ofthe metal-polluted Arkansas River, Colorado. Western N. Amer.Naturalist, 60, 57–65.

Sansalone, J.J. and Buchberger, S.G., 1997. Partitioning and firstflush of metals in urban roadway storm water. J. Environ. Eng.,123, 134–143.

Schultheis, A.S., Sanchez, M. and Hendricks, A.C., 1997.Structural and functional responses of stream insects to copperpollution. Hydrobiologia, 346, 85–93.

Shutes, R.B.E., 1984. The influence of surface runoff on themacroinvertebrate fauna of two north London streams. Environ.Technol. Lett., 6, 395–404.

Sriyaraj, K. and Shutes, R.B.E., 2001. An assessment of the impactof motorway runoff on a pond, wetland and stream. Environ.Int., 26, 433–439.

ter Braak, C.F.J., 1986. Canonical correspondence analysis; a neweigenvector technique for multivariate direct gradient analysis.Ecol., 67, 1167–1179.

ter Braak, C.F.J., 1987. CANOCO – A FORTRAN Program forCanonical Community Ordination by [Partial][Detrended][Canonical] Correspondence Analysis, PrincipalComponents Analysis and Redundancy Analysis (version 2.1),DLO-Agricultural Mathematics Group, Wageningen, TheNetherlands.

ter Braak, C.F.J., 1994. Canonical community ordination. Part I:Basic theory and linear methods. Ecosci., 1, 127–140.

ter Braak, C.F.J., 1996. Unimodal Models to Relate Species toEnvironment. Agricultural Mathematics Group, Wageningen,The Netherlands.

ter Braak, C.F.J. and Looman, C.W.M., 1986. Weighted averaging,logistic regression and the Gaussian response model. Math.Biosci., 78, 57–72.

ter Braak, C.F.J. and Prentice, I.C., 1988. A theory of gradientanalysis. In: Advances in Ecological Research, 18, 271–317.

ter Braak, C.F.J. and Smilauer, P., 1998. CANOCO ReferenceManual and User’s Guide to Canoco for Windows: Softwarefor Canonical Community Ordination (version 4.0),Microcomputer Power, Ithaca, New York.

ter Braak, C.F.J. and Verdonschot, P.F.M., 1995. Canonicalcorrespondence analysis and related multivariate methods inaquatic ecology. Aquat. Sci., 57, 255–289.

Tucker, K.A. and Burton, G.A., 1999. Assessment of nonpoint-source runoff in a stream using in situ and laboratory approaches.Environ. Toxicol. Chem., 18, 2797–2803.

Whiting, E.R. and Clifford, H.F., 1983. Invertebrate and urbanrunoff in a small northern stream, Edmonton, Alberta, Canada.Hydrobiologia, 102, 73–80.

Wright, J.F., 2000. An introduction to RIVPACS. In: Assessingthe biological quality of fresh waters: RIVPACS and othertechniques. The Freshwater Biological Association, Ambleside,UK. 1–24.

Gary Beasley and Pauline E. Kneale

234