pesticide effects on the structure and function of stream ecosystems · anthropogenic stressors to...

PhD thesisJes J. Rasmussen

2012

PESTICIDE EFFECTS ON THE STRUCTURE AND FUNCTION OF STREAM ECOSYSTEMS

AARHUS UNIVERSITYDEPARTMENT OF BIOSCIENCE

AU

2012PhD thesisJes J. Rasmussen

PESTICIDE EFFECTS ON THE STRUCTURE AND FUNCTION OF STREAM ECOSYSTEMS

He who constantly and unafraid

confronts the darkness will be the first to

see a glimpse of light

Dmitry Glukhovsky, 2002

Preface

This PhD thesis is submitted to the Faculty of Science, Aarhus University and is a result of a three year PhD project. The work has mainly been conducted at the department of Bioscience and partly at the department of Environmental Engineering, Technical University of Denmark. The PhD study was partly funded by and conducted as part of the RISKPOINT project which was financed by the Danish Research Council (grant no. 2104-07-0035). Moreover, this PhD project was partly funded by the National Environmental Research Institute, Aarhus University. The objective of this PhD was to examine the ecological effects of environmentally agricultural pesticides in surface water ecosystems in Denmark. The work conducted during this PhD includes building data bases, modelling, two laboratory studies and two major field campaigns. The major parts of this work have been included in the thesis as papers, paper drafts or planned papers. This thesis includes an introduction that provides background information for the scientific questions that have been addressed in the studies of this PhD along with nine papers. Of these papers, five have been published in peer-reviewed journals (Papers I – IV and IX), one is published in a non peer-reviewed journal (Paper VI) and three have been submitted to peer-reviewed journals (Papers V, VII and VIII).

Several people have been involved in the studies in my PhD, and to even more I need to express my deepest gratitude. I am so grateful for getting the opportunity to work on this PhD project and receiving such a large degree of freedom to plan, conduct, analyse and report results from my studies. I am grateful for the constant confidence and support of my supervisors, Annette Baattrup-Pedersen, Peter Wiberg-Larsen, Erik Baatrup and Brian Kronvang. This team has given me the best supervision and encouragement I could possible wish for. I have also been so fortunate to develop a warm and friendly scientific relationship with Matthias Liess and Nina Cedergreen who have also been great sparring partners and sources of inspiration to my work.

Huge thanks to the technical personnel in the stream ecology group who provided much and diverse help in the field and in the laboratory. Marlene Venø Skærbæk, Uffe Mensberg, Dorte Nedergaard, Johnny Nielsen and Henrik Stenholt; you have all been great with your continuous enthusiasm and friendly nature. I would also like to thank the freshwater ecology group in general for the very nice working environment despite increasingly rough times for many: Esben Kristensen, John Bøhme Dybkjær, Peter Mejlhede, Carlos Hoffmann, Nikolai Friberg, Joachim Audet, Rosemberg Menezes, Korhan Özkan, Daniel Gräber, Hans Thodsen, Dennis Trolle, Tom Davidsson, Shenglan Lu, Juliane and Jacob Zaccharias, Susanne Lildal Amsick, Hans Estrup Andersen, Ulrik Nørum, Søren Erik Larsen and Torben Lauridsen. Also a thanks to the ‘Fantasy football group for fun harassment and nerdy football conversations.

I owe my deepest thanks to Ursula McKnight who is the best collaboration partner I could have wished for, a friend. Also I would like to thank my RISKPOINT colleagues in general for bringing many different scientific fields into my world. I have been so fortunate to work with several of the RISKPOINT participants, all of whom I appreciate and deeply respect: Poul Løgstrup Bjerg, Phillip John Binning, Maria Loinaz, Nanna Kruuse and Peter Bauer Gottwein.

The warmest thoughts and hugs go to my dear friends: Nils Schrødder, Line Hansen, John Bøhme Dybkjær, Kristian B. Laursen, Kristian Slater, Kasper Grønning, Jost L. Hansen, Niels Egense Møller, Lena Nansen, the ‘whist club’ and the ‘role play nerds’ for creating time and space for all sorts of non-scientific fun in dark and bright periods.

My deepest gratitude to my family who supported me in all my enterprises and always has given me the desire to go further – still knowing that they are there behind me…with a smile. Especially my girlfriend, Trine and my little boy, Anton have always given me a place of joy and happiness to which I could return when things were tough. Also a special thanks to my parents in law for their interest in my studies and many constructive talks and limitless support. Building a house during my PhD was tough, but fun!

Contents

Summary 7

Dansk resumé 9

1. Introduction 11

Agricultural pesticides 12

Objectives and approach 13

2. Direct effects of pesticides on macroinvertebrate communities in streams 14

Pesticide effects on stream macroinvertebrates 14

Characterising the extent of the agricultural pesticide problem in Danish streams 14

Linking toxic stress from pesticides to macroinvertebrate community responses 16

The SPEAR index 17

Importance of physical habitat properties in stream ecosystems 18

Interactions between physical habitat characteristics and pesticide stress 19

3. Pesticide effects on ecosystem processes and indirect effects on macroinvertebrates 21

Decomposition of coarse particulate organic matter – stimulating and inhibiting factors 21

Pesticide effects on freshwater fungi 21

Pesticide effects on macroinvertebrate shredding activity 23

Indirect effects of fungicides on macroinvertebrate shredding activity 23

4. Perspectives 25

References 26

Included papers 32

I. Buffer strip width and agricultural pesticide contamination in Danish lowland streams: Implications for stream and riparian management

II. Local physical habitat quality clouds the effect of predicted pesticide runoff from agricultural land in Danish streams

III. Stream habitat structure influences macroinvertebrate responses to pesticide stress IV. Impacts of pesticides and natural stressors on leaf litter decomposition in agricultural

streams V. Effects of a triazole fungicide and a pyrethroide insecticide on the decomposition of leaves

in the presence or absence of macroinvertebrate shredders VI. Ny viden om effekter af pesticider i vandløb VII. Thresholds for the effects of pesticides on invertebrate communities and leaf breakdown in

stream ecosystems VIII. Integrated assessment of chemical stressors on surface water ecosystems IX. An integrated model for assessing the risk of TCE groundwater contamination to human

receptors and surface water ecosystems

List of included papers

I) Rasmussen, J.J., Baattrup-Pedersen, A., Wiberg-Larsen, P., McKnight, U.S., Kronvang, B. 2011. Buffer strip width and agricultural pesticide contamination in Danish lowland streams: Implications for stream and riparian management. Ecological Engineering 37, 1990-1997.

II) Rasmussen, J.J., Baattrup-Pedersen, A., Larsen, S.E., Kronvang, B. 2011. Local physical habitat quality clouds the effect of predicted pesticide runoff from agricultural land in Danish streams. Journal of Environmental Monitoring 13, 943-950.

III) Rasmussen, J.J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Friberg, N., Kronvang, B. 2012. Stream habitat structure influences macroinvertebrate response to pesticides. Environmental Pollution 164, 142-149.

IV) Rasmussen, J.J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Monberg, R.J., Kronvang, B. 2012. Impacts of pesticides and natural stressors on leaf litter decomposition in agricultural streams. Science of the Total Environment 416, 148-155.

V) Rasmussen, J.J., Monberg, R.J., Baattrup-Pedersen, A., Cedergreen, N., Wiberg-Larsen, P., Strobel, B., Kronvang, B. 2012. Effects of a triazole fungicide and a pyrethroid insecticide on the decomposition of leaves in the presence or absence of macroinvertebrate shredders. Submitted to Aquatic Toxicology.

VI) Rasmussen, J.J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Monberg, R.J., McKnight, U.S., Kronvang, B. 2011. Ny viden om effekter af pesticider i vandløb. Vand & Jord 18, 143-147.

VII) Schäfer, R., von der Ohe, P.C., Rasmussen, J.J., Kefford, B., Beketov, M., Schulz, R., Liess, M. 2012. Thresholds for the effects of pesticides on invertebrate communities and leaf breakdown in stream ecosystems. Submitted to Environmental Science and Technology.

VIII) McKnight, U.S., Rasmussen, J.J., Kronvang, B., Bjerg, P.L., Binning, P.J. 2012. Integrated assessment of chemical stressors on surface water ecosystems. Submitted to Science of the Total Environment.

IX) McKnight, U.S., Funder, S.G., Rasmussen, J.J., Finkel, M., Bjerg, P.L., Binning, P.J. 2010. An integrated model for assessing the risk of TCE groundwater contamination to human receptors and surface water ecosystems. Ecological Engineering 36, 1126-1137.

Papers in preparation A. McKnight, U.S., Rasmussen, J.J., Kronvang, B., Bjerg, P.L., Binning, P.J. Pesticide

occurrence in streams is a fingerprint for the history of applications in the catchment. B. Rasmussen, J.J., McKnight, U.S., Baattrup-Pedersen, A., Loinaz, M., Kruuse, N.,

Kronvang, B., Bjerg, P.L., Binning, P.J. Evaluating ecological impacts from different stressors at the catchment scale.

C. Rasmussen, J.J., Friberg, N., Baattrup-Pedersen, A., Kronvang, B. Species traits reveal importance of anthropogenic controls in agricultural streams

7

Streams and rivers do not contain more than a fragment of the global water resources, but they are unique and complex habitats due to the extremely close connectivity between terrestrial and aquatic habitats. They host important fish stocks and a unique set of organisms that provides invaluable consumer services, such as chemical water purification and organic matter processing. Stream ecosystems are, moreover, impacted by numerous anthropogenic stressors, and they remain some of the most impaired ecosystems on earth (MEA, 2005). The contamination of streams with pesticides applied in agricultural production is widely acknowledged as one of the greatest anthropogenic stressors to stream ecosystems (Schulz, 2004), and agricultural pesticides are known to pose a threat to all living organisms in stream ecosystems (Liess et al., 2005). The general objective of this PhD thesis was to study the effects of agricultural pesticides on the structure and function of stream ecosystems. Benthic macroinvertebrates were primarily selected as study organisms since they 1) respond strongly to pesticide contamination; 2) they are thoroughly studied in the field of ecology which provides more available relevant information for the studies; 3) they are highly important for a series of ecosystem processes; 4) there is a scarcity of field studies evaluating effects of pesticide contamination resulting from normal agricultural practice on stream ecosystems, and 5) there is a lack of studies focusing on the indirect effects of pesticides – e.g. pesticide effects on food resources of macroinvertebrates. The project aimed to address the study objective through a series of activities, including field campaigns, laboratory studies and interpretations of historical data. More specifically, the project involved activities that were launched to:

• Characterise the potential problem of

pesticide contamination in Danish streams for the macroinvertebrate communities

• Evaluate if it is possible to disentangle the effects of pesticide pollution from effects of all other anthropogenic stressors on macroinvertebrates

• Evaluate pesticide effects on selected microorganism based ecosystem processes

• Study the effects of changes in microorganism communities, that is the result of pesticide exposure, on the foraging

behaviour of macroinvertebrates using this food resource

Our results indicate that pesticide concentrations often reach concentrations during storm flow episodes that potentially impact the macroinvertebrate community (Paper I). Moreover, we found that pesticides adsorbed in bed sediments, especially pyrethroids, may have a stronger impact on the benthic macroinvertebrate community than the freely diluted (or adsorbed to suspended particles) fraction that only occur briefly during rain episodes (Papers VIII and A). The dimensions of vegetated buffer strips were found to be of high importance for the total amount of pesticides that reached the stream recipient and their toxicity to macroinvertebrates, since total concentrations and toxicity were strongly intercorrelated. Using data from the Danish stream monitoring program, however, we found that it was not possible to detect effects of predicted pesticide runoff contamination in 212 Danish streams (Paper II). We used different macroinvertebrate based metrics to evaluate the impact of a series of different anthropogenic stressors and found the local physical habitat characteristics to be clouding the effects from other stressors, including the recently developed pesticide indicator SPEAR (Liess & von der Ohe, 2005). Applying field data collected on the Danish islands, Funen and Zealand, we were, however, able to detect the effects of pesticides on the macroinvertebrate community structure (using the SPEAR index) when macroinvertebrate samples were collected on the physically best locations in the sampled streams (Paper III, VII and B). Compiling similar datasets (also using macroinvertebrate data collected from the physically best locations in the streams) from other biogeographical regions we fitted dose-response curves to the field data in order to obtain field-based effect thresholds for changes in the macroinvertebrate community structure (Paper VII). In paper III, IV and C, we used a paired reach field set-up to study the effects of pesticides on stream ecosystems in reaches with differing physical characteristics. In paper III, we showed that the SPEAR index response was different between reaches with differing physical properties

Summary

8

but similar degree of pesticide pollution. This difference was found to be primarily due to the presence or absence of species with specific requirements for hard substrate types or macrophyte species as habitat. These species were therefore absent in even unpolluted streams if their specific habitat requirements were not fulfilled. Typically, streams that are characterised by degraded physical habitats and substrate types dominated by fine particles are physically unstable systems. Moreover, these stream characteristics are dominating in areas that are heavily utilised for conventional agriculture, and other side-effects of conventional agriculture, such as pesticide pollution and eutrophication are therefore often present in these stream types. Similar ecological traits are consequently required for species that successfully can survive in such highly disturbed habitats, and over-representation of these traits are therefore not only a signal for pesticide pollution (Paper C). We also evaluated the relative importance of different chemical pollutions that originated from contaminated sites (papers VIII, IX and B) compared to other physical and chemical stressors (e.g. diffuse source pesticides pollution). In these studies we found no indications that contamination plumes that discharge into adjacent streams were significantly contributing to the sum effect of all measured anthropogenic stressors. However, this is not a general conclusion but need to be evaluated from case to case. In paper IV we showed that the microbial decomposition of leaf litter was reduced in the field when pesticides, especially fungicides, were detected in water samples during storm flow. This is intuitively logic, because the primary group of microbial decomposers are freshwater fungi, and fungicides have modes of actions that specifically affect these organisms. We also showed that the rates of microbial litter decomposition was further reduced by 50% in stream sites that were characterised by degraded physical habitats compared to sites that were characterised by heterogeneous physical conditions. These findings probably show that increased siltation and reduced boundary layers curtail the organisms in their ability to acquire nutrients and oxygen from the surrounding environment. The reduction in microbial decomposition may or may not be the result of decreased microbial biomass, but macroinvertebrate shredding activity was not significantly affected

by the potentially reduced microbial biomass, by pesticides or by physical conditions, primarily due to very high abundances of the freshwater amphipod, Gammarus pulex (paper IV). In other biogeographical regions, however, the shredding activity was clearly reduced in streams where pesticides with high toxicity for benthic macroinvertebrates were detected in stream water during storm flow (paper VII). In the laboratory, we studied potential indirect effects of fungicides and insecticides on macroinvertebrate shredders (paper V). Our results showed that fungicide treatments reduced the microbial biomass on leaves and their decomposition rate. The reduced microbial biomass resulted in lower nutritional value of the leaves for macroinvertebrate shredders, and in compensation for decreased nutritional value they increased their ingestion of leaf material in order to maintain basic body functions. This picture was evident even though some of the leaves were additionally treated with a pyrethroid insecticide. Our results suggest that pesticides indeed pose a threat to ecosystems in Danish streams, but more emphasis needs to be given to the characterisation of transport routes and pesticides adsorbed in the bed sediments in order to improve mitigation strategies and risk assessment. We found that the SPEAR index was a strong tool for detecting effects of pesticides on macroinvertebrate communities, but physical properties of streams are equally important to consider in risk assessment, sampling strategies for macroinvertebrates and assessment of ecosystem processes. We also highlight the importance of indirect effects of pesticides. The effect of pesticides on the level of ecosystems is a combination of direct and indirect effects and feed-back mechanisms, which all need to be considered in the interpretation of ecosystem effects. The major conclusions and future perspectives are additionally summarised in popular scientific paper (VI).

9

Den samlede mængde vand i alle jordens vandløb udgør ikke mere end 0,006 % af jordens samlede vandressourcer, men vandløb rummer unikke og komplekse habitater, fordi der er så tæt forbindelse og stor kontaktflade imellem det akvatiske og terrestriske miljø. Vandløbene rummer en række yderst vigtige ressourcer for mennesker, såsom fisk, samt en række organismer, der udøver essentielle services for mennesker – såsom vandløbenes selvrensende effekter og stofomsætning. Vandløbsøkosystemer er påvirket af en lang række menneskeskabte stressfaktorer, og de udgør nogle af de mest påvirkede økosystemer overhovedet (MEA, 2005). Forurening af vandløb med de i konventionelle landbrug anvendte pesticider er bredt anerkendt som en af de største menneskeskabte stressfaktorer, der påvirker vandløbsøkosystemerne (Schulz, 2004), og pesticider er påvist at kunne påvirke alle vandløbets organismer (Liess et al., 2005). Det overordnede formål med denne PhD var at studere effekter af pesticider på vandløbsøkosystemers struktur og funktion. Primær fokus har været rettet mod vandløbsmakroinvertebraterne, fordi de 1) har påvist høj følsomhed overfor pesticider; 2) de er en grundig undersøgt organismegruppe, og der er derfor generelt meget tilgængelig viden til anvendelse i denne PhD; 3) makroinvertebraterne er en vigtig bestanddel i vandløbsøkosystemet og er helt centrale i en række biologiske processer; 4) effekter af pesticidforurening, der stammer fra almindelig landbrugspraksis, på makroinvertebrater er utilstrækkeligt undersøgt, og 5) der er generel mangel på studier, der fokuserer på indirekte effekter af pesticider – f.eks. via effekter på makroinvertebraters fødeemner. En række aktiviteter har været udført som del af denne PhD for at imødekomme projektets formål, herunder studer i felten og laboratoriet samt analyser af historiske data, indsamlet i forbindelse med NOVANA overvågningsprogrammet. Mere specifikt har delstudierne adresseret følgende delformål: • Karakterisere det potentielle problem for

makroinvertebratsamfund som udgøres af pesticidforurening i danske vandløb

• Evaluere muligheden for at udrede effekterne pesticider fra andre menneskeskabte stressfaktorer

• Vurdere pesticideffekter på udvalgte mikrobiologiske økosystemprocesser

• Studere effekter af forandringer i mikrobiologiske samfund, der er forårsaget af pesticideksponering, på makroinvertebraters fouragering på denne fødekilde

Vores studier viser, at pesticider ofte forefindes i tilstrækkeligt høje koncentrationer i vandløbsvand under høj vandføring i forbindelse med kraftige nedbørshændelser til at kunne påvirke makroinvertebratsamfund (artikel I). Desuden fandt vi, at sorberede pesticider i sedimenter i vandløb, ofte pyrethroide insekticider, kan have større effekt på bundlevende makroinvertebrater end pesticider associeret med vandfasen under høj vandføring (artikel VIII og A). Bredden af udyrkede randzoner langs vandløb blev desuden fundet at have høj beskyttelsesværdi i forhold til mængden af pesticider, der blev transporteret til vandløbene, samt deres giftighed for makroinvertebrater. Ved analyser af data fra 212 vandløbsstationer fra NOVANA overvågningsprogrammet fandt vi, at det ikke var muligt at påvise effekter af estimeret pesticidforurening (artikel II). Vi anvendte forskellige makroinvertebratbaserede indikatorer (herunder det nyligt udviklede SPEAR indeks for pesticidforurening, Liess & von der Ohe, 2005) for at evaluere den relative betydning af målte menneskeskabte stressfaktorer, og vi fandt, at de lokale fysiske forhold, især substrat karakteristika, overskyggede effekten af andre stressfaktorer. Ved brug af samme SPEAR indeks gav anvendelsen af data for pesticider og makroinvertebrater indsamlet i feltstudier på Fyn og Sjælland til gengæld en stærk sammenhæng mellem pesticidernes giftighed for makroinvertebrater og disses artssammensætning, når prøveindsamlingen var foretaget på de fysisk set bedste lokaliteter i vandløbene (artikler III, VII og B). Anvendelsen af data fra flere forskellige biogeografiske regioner, indsamlet med samme prøvetagningsstrategi, gav samme billede (artikel VII). I artikel VII udnyttede vi denne stærke

Dansk resumé

10

sammenhæng mellem SPEAR indeks scores og pesticiders giftighed til at fitte dosis-respons kurver på data for at fastsætte felt-baserede grænseværdier for, hvornår der forekommer signifikante forandringer i makroinvertebraters samfundsstruktur. I artiklerne III, IV og C anvendte vi parrede vandløbsstrækninger, hvor strækningerne udelukkende adskilte sig på deres fysiske karakteristika, til at studere, hvorvidt pesticideffekter på vandløbsøkosystemer var afhængig af vandløbsfysikken. Vi viste, at SPEAR indeks scores var signifikant forskellige imellem de parrede strækninger primært pga. fravær eller tilstedeværelse af arter med specifikke habitatkrav for hårde substrattyper. Disse arter var således fraværende selv i uforurenede vandløb på de fysisk forarmede strækninger, fordi deres habitat krav ikke var opfyldt. Typisk er fysisk forarmede vandløb domineret af fine substrattyper og er fysisk ustabile systemer. Derudover er netop de fysisk forarmede vandløb overrepræsenteret i landbrugsdominerede oplande, hvorfor der i netop disse vandløb også ofte findes høje koncentrationer af pesticider og næringsstoffer. Organismer, der residerer i disse vandløb, er derfor karakteriseret ved biologiske karaktertræk, der generelt er tilpasset særligt ustabile miljøer – hvilket ikke nødvendigvis kun afspejler effekter af pesticider (artikel C). Vi evaluerede også den relative betydning af kemiske forureninger relateret til forurenede grunde (artikler VIII, IX og B) i forhold til andre fysiske og kemiske stressfaktorer, såsom landbrugsrelateret pesticidforurening. I disse studier fandt vi ingen indikationer på, at forurenede grunde har nævneværdig effekt på smådyrssamfundene sammenlignet med fysisk forarmelse og pesticidforurening. Disse scenarier skal dog typisk vurderes fra sag til sag. I artikel IV viste vi, at den mikrobiologiske nedbrydning af organisk materiale (bøgeblade) var reduceret i vandløb, hvor især svampemidler blev fundet i vandløbsvandet under forøget vandføring. Dette giver intuitivt mening, fordi den primære gruppe af mikrobielle nedbrydere er mikrosvampe, og svampemidler har virkningsmekanismer, der netop er rettet med denne organismegruppe. Vi viste derudover, at på strækninger med fysisk forarmede forhold var den mikrobielle omsætning yderligere reduceret med 50% sammenlignet med strækninger med bedre fysiske forhold. Dette afspejler formentlig, at de

mikrobielle organismer bliver reduceret i deres evne til at tilegne sig næringsstoffer og ilt fra de omgivende vandmasser på strækninger karakteriseret af øget sedimenttransport og tykkere grænselag. Denne påviste reduktion i mikrobiel aktivitet kunne være en indikation på nedsat mikrobiel biomasse og dermed nedsat fødemæssig kvalitet for makroinvertebraterne, men dette afspejlede sig imidlertid ikke i ituriveraktivitet hos makroinvertebraterne, der klart var domineret af ferskvandstangloppen, Gammarus pulex (artikel IV). I andre biogeografiske områder var der til gengæld en klar reduceret ituriveraktivitet i vandløb, hvor der var detekteret pesticider med høj giftighed for makroinvertebrater (artikel VII). I et laboratoriestudium undersøgte vi potentielt indirekte effekter af svampe- og insektmidler på ituriveraktivitet (artikel V). Vores resultater bekræftede, at svampemidler reducerede både aktivitet og biomasse af mikrobielle nedbrydere. Den reducerede mikrobielle biomasse kom ligeledes til udtryk i form af nedsat næringsværdi af bladene for makroinvertebrater, og makroinvertebraterne reagerede på denne næringsmæssige forringelse ved at øge fødeindtaget, formentlig en kompensation for at opretholde basale fysiologiske processer i deres kroppe. Dette overordnede billede var tydelig på trods af, at halvdelen af bladene ligeledes var behandlet med et pyrethroid insektmiddel. Vores samlede resultater viser, at der generelt forefindes økosystemeffekter af pesticider i danske vandløb, men også at der i fremtiden skal lægges mere vægt på de sedimentbundne fraktioner og disses transportruter til vandløbene for at kunne optimere virkemidler og indsatsområder. Vi fandt, at SPEAR indekset var et brugbart supplement for vurderingen af pesticid effekter i vandløb, men også at de fysiske forhold i vandløb er ligeligt vigtigt for forståelsen af den samlede effekt af menneskeskabte stressfaktorer. Prøveindsamlingsstrategier og kvalitetsbedømmelser skal vægte begge faktorer for at få meningsfyldt udbytte af analyser på økosystem effekter for biologisk struktur og funktion. Herunder fremhæver vi også vigtigheden af indirekte effekter af pesticider, der kan fremkalde kaskadeeffekter igennem flere trofiske niveauer. De overordnede konklusioner og perspektiver af denne PhD afhandling er ligeledes opsummeret i artikel VI.

11

World-wide, running waters (streams and rivers) drain an area of approximately 150 million km2 of land. Apart from hosting important fish stocks and enriching us with their scenic beauty, they are also characterised by conspicuously rich and complex environments due to their close connectivity between terrestrial and aquatic habitats. Running waters therefore accommodate a broad spectrum of different life forms which, additionally, provide a series of consumer services such as water purification and the organic matter processing. These processes eventually play an important role in the supply of drinking water for humans and human related activities. Moreover, they are conduits for effluents from industrial, domestic and agricultural sources. The water quality requirements for these processes and the additional requirements for conservation of biodiversity have led to conflicts in the management and conservation of streams. The problem of globalisation increases the competition on the international market for food production and industry. One side effect is the increasing use of xenobiotic compounds (chemicals that are not natural products or contents of living organisms) in food production and industry with the risk of introducing them into the human and/or natural environment. The quantity of xenobiotic compounds that are

deliberately or accidentally introduced to the natural environment is a serious threat to all forms of life and the direct, indirect and combined effects of these compounds in natural environments is of great concern in terms of biodiversity, ecosystem services (resources and processes that are supplied by natural ecosystems, such as the processing of organic matter in streams and stream self-purification) and environmental sustainability in general (Liess et al., 2005; MEA, 2005). In total, more than 14 million chemicals exist, and at least 100,000 of them are industrially produced. Many of these compounds are not characterised with respect to their ecotoxicity or characterised in terms of occurrence in the natural environment due to e.g. insufficient analytical methods or only temporally brief occurrences (like some agricultural pesticides) (von der Ohe et al., 2009). In consequence, a number of different screening and prioritisation exercises have been conducted aiming to improve the effectiveness and usefulness of environmental risk assessment (ERA) (e.g. Baun et al., 2004; Carafa et al., 2011; Götz et al., 2010; Kreuger, 1998; Muir and Howard, 2006; von der Ohe et al., 2011). Results from one of these studies revealed that approximately 75% of all xenobiotic chemicals that were found to exceed the predicted no effect

Figure 1. Major transport routes into receiving streams for pesticides that are applied to agricultural fields.

1. Introduction

12

concentrations (the concentration below which exposure to a substance is not expected to cause adverse effects) for non-target organisms in streams were pesticides (von der Ohe et al., 2011). Agricultural pesticides

Pesticides that are applied to agricultural fields are transported to surface water recipients via wind drift, surface runoff, tile drains, groundwater and leakage from point sources (e.g. washing places for spraying equipment and farm yards) (Schulz, 2004) (Fig. 1). The transport pathways are strongly governed by local climatic and geological conditions and by the physicochemical properties of the pesticide compound. Surface runoff and transport through tile drains are widely acknowledged as the most important transport routes for pesticides from fields to receiving streams (Kreuger, 1998; Kronvang et al., 2004; Liess et al., 1999; Schulz et al., 1998; Wauchope, 1978). Consequently, peak pesticide

concentrations are mostly short and occur during heavy precipitation events. Considering peak concentrations is essential as the maximum exposure concentration is proposed to be more important for estimating ecological effects than i.e. exposure duration (Schulz and Liess, 2000). In Denmark, 62% of the total area is used for conventional agriculture, and the length of the stream network amounts to more than 60,000 km – the majority of which are small streams (< 2 m wide). The combination of extensive agricultural activities and the close connectivity between land and stream emphasises that streams are at high risk of receiving pesticide input. Especially in small agricultural streams where the connectivity between land and stream is high, and the dilution factor is low (low discharge), the risk of high pesticide concentrations is enormous (Probst et al., 2005a; Schulz, 2004). This landscape scenario with primarily small streams draining catchments dominated by conventional agriculture is not unique for Denmark but is more a general picture

Figure 2. The stream catchment, illustrating different anthropogenic stressors that impact the stream ecosystem structure and function. The font size of the stressor nomenclature is congruent with the amount of focus that the specific stressor has received in the papers that are appended in the PhD thesis. Pesticide pollution is addressed in papers I-VIII, physical stream properties in papers II-IV and VIII, pollution from contaminated sites in papers VIII and IX, nutrients and organic pollution is briefly addressed in papers II-IV and VIII. The remaining identified stressors have only marginally been discussed in papers II and VIII but will be further discussed in papers B and C.

13

for many highly developed agricultural countries like Holland, United Kingdom, Germany, France, Sweden and United States emphasising the magnitude of the potential problem (Brown et al., 2006; Kreuger, 1998; Probst et al., 2005b; Schäfer et al., 2007; Weston et al., 2008). Overall, a multitude of environmental parameters that act on stream ecosystems on different spatial and temporal scales co-occur and are directly or indirectly correlated (fig. 2). As an example, agricultural streams are potentially impacted by a series of local scale stressors that primarily occur in short time intervals such as pollution with pesticides and nutrients. Moreover, agricultural streams may be impacted by other local stressors that are more permanent in time such as physical habitat degradation (loss of physical habitat complexity and quality) resulting from channelisation and dredging. On a larger spatial scale, the distance to nearest colonisation pool is an important parameter that may act as a strong dispersal barrier on agricultural stream ecosystems, since they can be rather isolated islands of water in a land of sea (MacArthur and Wilson, 1967). Larger scale ecological filters (landscape structure, climate etc.) shape and constrain the regional species pool, but in order to join a local community, a species in a regional pool must possess appropriate biological traits to “pass” through nested filters at the local scale (Poff, 1997). The relative influence of local and regional factors that govern the structure of stream ecosystems is not fully understood, but there is some evidence that the local scale factors are more important ecological filters for e.g. benthic stream macroinvertebrates than regional scale filters (Pedersen and Friberg, 2009; Pedersen et al., 2004; Sandin, 2009). Appreciating the relative contribution of single stressors, such as pesticides, to the sum effect of all co-occurring stressors is, however, a major and difficult task. In consequence, only 5 studies (representing 0.6% of the total amount of papers addressing pesticide effects on freshwater organisms) attempted to interpret the effects of pesticides in larger field set-up’s (Beketov and Liess, 2012). Objectives and approach

The general objective of this thesis was to study effects of pesticides on stream ecosystem structure and function with specific focus on benthic macroinvertebrates and ecosystem processes

related to this group of organisms. The project aimed to address the study objective through a series of field and laboratory studies and analyses on historical data sets that were collected as part of Danish monitoring program, NOVANA. More specifically, we aimed to:

• Characterise the potential problem of pesticide contamination in Danish streams for the macroinvertebrate communities

• Evaluate if it is possible to disentangle the effects of pesticide pollution from effects of all other anthropogenic stressors on macroinvertebrates

• Evaluate pesticide effects on selected ecosystem processes of relevance to macroinvertebrates

• Study the (indirect) effects of changes in microorganism communities, that is the result of pesticide exposure, on the foraging behaviour of macroinvertebrates that use this food resource

The papers presented in this thesis, together with some unpublished results discussed here, aim at providing empirical insight to the overall understanding of how pesticides impact stream ecosystems and how these effects may be detected in multistressor environments, such as streams.

14

The diffuse source contamination of agriculturally applied pesticides is widely acknowledged as one of the most important anthropogenic stressors in stream ecosystems (Neumann et al., 2002; Schulz, 2004), and they have a wide range of impacts on organisms from all trophic levels ranging from algae to fish (Davies and Cook, 1993; Guasch et al., 1997; Kreutzweiser and Kingsbury, 1987; Liess et al., 2005). However, benthic macroinvertebrates have attracted considerable attention as they respond strongly to pesticide exposure, especially insecticides (e.g. Nørum et al., 2010; Schulz and Liess, 1999; Wallace et al., 1989). Moreover, macroinvertebrate community composition is used widely as indicator for the ecological quality of streams (Metcalfe-Smith, 1996). In consequence, macroinvertebrate ecology is well researched, and benthic macroinvertebrates are keystone organisms for detecting and interpreting impacts of various anthropogenic stressors, including pesticides. Pesticide effects on stream macroinvertebrates

The vast majority of ecotoxicological studies concerning pesticide impacts on stream macroinvertebrates are focusing on the insecticides, since this group of pesticides has modes of actions that directly impact non-target aquatic insects and arthropods. Examples are pyrethroid insecticides which primarily have an effect on the nervous system decreasing the decay of the action potential by inhibiting closure of the sodium channel (Vijverberg and van den Bercken, 1990) and organophosphate insecticides which irreversibly inactivate the enzyme acetyl cholinesterase – an essential enzyme in the nervous system of insects (and humans and many other animals) (Straus and Chambers, 1995). In several studies, effects of insecticides on benthic stream macroinvertebrates have been documented using mortality, immobilisation and the offset of catastrophic drift as endpoints in field- or mesocosm experiments using controlled exposure scenarios (Beketov and Liess, 2008a; Heckmann and Friberg, 2005; Kreutzweiser and Kingsbury, 1987; Liess, 1994; Muirhead-Thomson, 1978; Nørum et al., 2010; Rasmussen et al., 2008; Sibley et al., 1991; Wallace et al.,

1989). Despite severely reduced abundances of especially taxa belonging to Ephemeroptera, Plecoptera and Trichoptera in stream sections that were exposed to insecticides resulting from direct toxic effects on the macroinvertebrates or their active escape behaviour (catastrophic drift), the communities appeared to recover within few months to a year after the contamination (Heckmann and Friberg, 2005; Kreutzweiser and Kingsbury, 1987; Liess and Schulz, 1999; Sibley et al., 1991; Wallace et al., 1989). However, the remaining question is still whether normal agricultural practices result in pesticide contamination that significantly impacts the macroinvertebrate communities. Characterising the extent of the agricultural pesticide problem in Danish streams

Measuring pesticide contamination that actually results from normal agricultural practices and linking the contamination to community responses of stream macroinvertebrates is of utmost importance in order to produce sound and realistic risk assessment. Not only is it important to characterise realistic exposure scenarios, it is, additionally, important to link exposure scenarios to agricultural practices, climatic conditions and landscape characteristics in order to gain knowledge that can be used to produce and prioritise the best management and mitigation strategies. In fact, according to the European Water Framework Directive (WFD), member states are obliged to not only assess the ecological quality and meet ecological quality requirements, but also to identify and characterise the major environmental and anthropogenic drivers that cause impairment of the surface water ecosystems (European Commission, 2000). Through the last decades, an increasing number of studies report measurements of pesticide concentrations in stream water resulting from normal agricultural practices in different parts of the world using different measuring techniques (Berenzen et al., 2005; Carafa et al., 2011; Castillo et al., 2006; Feo et al., 2010; Hladik and Kuivila, 2009; Jergentz et al., 2004; Kreuger and Tornqvist, 1998; Leonard et al., 2000; Liess et al., 1996, 1999; Liess and Von der

2. Direct effects of pesticides on macroinvertebrate communities in streams

15

Ohe, 2005; Neumann et al., 2002, 2003; Schletterer et al., 2010; Schriever et al., 2007; Schulz, 2004; Schulz et al., 1998; Schäfer et al., 2007, 2011c; Thiere and Schulz, 2004; Wiggers, 1999). Nevertheless, in only few of these studies, the authors have been successful in linking catchment based characteristics to measured in-stream pesticide concentrations, mainly because the transport of pesticides from fields to streams is extremely complex and strongly governed by a series of climatic, geological and hydromorphological factors that are also characterised by a high degree of local variation. Different approaches have been attempted to link measured concentrations in stream water to catchment based parameters ranging from simple correlations to complex models. Roughly, all of these studies support that the strongest governing factor determining high pesticide concentrations in stream water is the applied amount of pesticides in the catchment (Berenzen et al., 2005; Kreuger and Tornqvist, 1998; Schriever et al., 2007; Sørensen et al., 2003). This consensus also receives support from results in paper I. More importantly, since the majority of pesticide compounds that are highly toxic to benthic macroinvertebrates, in addition are highly lipophilic (especially pyrethroids), their preferential transport routes to receiving streams could be dominated by surface runoff. Vegetated buffer strips may therefore be an effective mitigation tool obstructing this transport route (Lacas et al., 2005). In paper I, we showed that the width of vegetated buffer strips in the near upstream area indeed was the catchment parameter that was most strongly correlated to the toxicity (log sumTUD. magna)

1 of the measured pesticide concentrations. The use of pesticide toxicity to benthic macroinvertebrates as alternative to pesticide concentrations is further justified by a very strong correlation between measured total pesticide concentrations in stream water and the calculated log sumTUD. magna (Fig. 3). The log sumTUD. magna was, furthermore, found to be stronger correlated to catchment based parameters than other measures for pesticide contamination (i.e. total concentration and number

1 Toxic Units (TU) for 48h acute toxicity tests on Daphnia magna is used as surrogate measure for pesticide toxicity to stream macroinvertebrates in general (Tomlin, C.D.S., 2000. The pesticide manual, a wold compendium. Crop Protection Publications, Farnham, Surrey, UK.).

of pesticide compounds) (paper I). The pesticide toxicities obtained during storm flow in several of the Funen streams were above the threshold for expected ecosystem effects (suggested by Schäfer et al., 2007). In paper I, we additionally attempted to characterise and quantify pesticides remaining in the stream system as particle-associated complexes after the runoff episodes. Unfortunately, this approach was unsuccessful due to too high detection limits at the Omegam Laboratories whose services were purchased for this task. This alternative exposure scenario is, however, probably highly relevant as pesticides remaining in the stream systems are adsorbed to habitat surfaces and food for benthic macroinvertebrates. Especially the strongly lipophilic compounds, of which many are highly toxic to macroinvertebrates, are partly transported

to surface water recipients as sorption complexes that will eventually be deposited on the stream bed and partly as freely diluted molecules which are likely to rapidly adsorb to organic surfaces like macrophyte patches and sediments (Domagalski et al., 2010; Fojut and Young, 2011; Gan et al., 2005; Moore et al., 2001; Palmquist et al., 2011; Stehle et al., 2011; Wauchope, 1978). The mechanisms determining bioavailability of compounds adsorbed to organic surfaces are, unfortunately, not yet fully understood, but the

Figure 3. Sum toxicity of measured pesticides for Daphnia magna (as surrogate measure for benthic macroinvertebrates) as a function of the measured total pesticide concentrations in 14 Danish streams.

16

organic carbon content of the substrate is important for binding affinity of the pesticides (Cui et al., 2010; Maund et al., 2002; Palmquist et al., 2011; Yang et al., 2006a; Yang et al., 2006b; Åkerblom et al., 2008). However, the fraction of adsorbed pesticides constitutes an alternative and more chronic exposure scenario that, relevantly, deserves attention. This is furthermore important, because the degradation and volatilisation of pesticide compounds is significantly reduced in their adsorbed forms. In streams on Zealand, we found that the most toxic compounds for benthic macroinvertebrates, mainly pyrethroid and organo-chlorine insecticides were only found in bed sediments and not in stream water during storm flow (papers VIII and A) which emphasises the importance of this exposure scenario. With few exceptions, only studies from the United States (primarily California) have focused on characterising the pesticide pollution that is associated with sediments in agricultural streams, and, overall, the studies clearly show that sediment-associated pesticides indeed have an impact on growth and mortality for sediment dwelling macroinvertebrates (Anderson et al., 2006; Bacey and Spurlock, 2007; Cover et al., 2008; Ding et al., 2010; Domagalski et al., 2010; Ensminger et al., 2011; Fojut and Young, 2011; Hladik and Kuivila, 2009; Palmquist et al., 2011; Weston et al., 2005, 2009; Weston and Lydy, 2010). However, common for all these studies is that the pesticide quantifications were not evaluated according to expected sources of origin and possible mitigation methods. Kronvang et al. (2003) linked pesticide concentrations in bed sediments to land-use activities and Weston et al. (2008) showed that one drain primarily was responsible for the observed contamination of stream bed sediment concentrations of pyrethroids. The study of Weston et al. (2008) additionally emphasises the importance of soil macro-pores and tile drains as transport routes for even the most lipophilic pesticide compounds. More detailed studies are needed to elucidate the importance of tile drains and macro-pores for the transport of lipophilic pesticides as single compounds and sorption complexes. Linking measured toxic stress from pesticides to macroinvertebrate community responses

Only a limited amount of large scale field studies have focused on linking responses of

macroinvertebrate communities to measured pesticide exposure originating from normal agricultural practice (Beketov and Liess, 2012). This is probably because 1) high costs are associated with the chemical analyses 2) large natural variation strongly complicates the interpretation of observed community dynamics and 3) the co-occurrence of multiple anthropogenic stressors, all of which contribute to the overall community response, complicates the possibilities to disentangle the effects of separate stressors (Beketov and Liess, 2012). Moreover, as pesticide concentrations rarely reach the levels of acute mortality for stream macroinvertebrates, sublethal effects may be very important, and the overall community response may, in consequence, only be visible after several generations.

Small scale (single site) and semi-controlled field studies attempting to link diffuse source pesticide contamination to effects on macroinvertebrates in streams have been conducted for several decades with varying degrees of success focusing on separate species to entire communities using absence/presence, abundance, diversity, mortality and drift as endpoints, but clear congruency of the different results is not certain (reviewed by Schulz, 2004). However, recently the field of ecology has been increasingly integrated into the field of ecotoxicology in order to improve the understanding of biological responses to contaminants. Especially the use of biological traits to appraise the effects of pesticides on non-target freshwater macroinvertebrates is considered promising for the detection and interpretation of pesticide contamination (Baird et al., 2011; Relyea and Hoverman, 2006). This interest has been particularly strong for freshwater macroinvertebrates because 1) they respond strongly to pesticide stress, 2) trait-based approaches are most developed in freshwater ecology for macroinvertebrates and 3) there is a paucity of field studies evaluating effects of pesticides on other groups of organisms (Schäfer et al., 2011a).

The distribution and abundance of specific traits of freshwater macroinvertebrates have been shown to remain relatively constant among similar ecosystem types across different continents whereas the species composition is clearly more variable (Baird and Van den Brink, 2007; Rubach et al., 2011; Van Den Brink et al., 2011). In addition, biological traits have been shown to

17

respond rather specifically to the different anthropogenic stressors such as brief pulses of agricultural pesticides (Liess and Von der Ohe, 2005; Rubach et al., 2010). The SPEAR index

One example of such trait-based approaches is the SPEcies At Risk index (SPEAR) which was developed for detecting pesticide effects on freshwater macroinvertebrates (Liess and Von der Ohe, 2005). The SPEAR index applies physiological (sensitivity to pesticides and other organic toxicants) and biological (generation time, timing and duration of terrestrial life stages and dispersal capacity) traits to determine the relative abundance of the species that are at risk of being impacted by the periodic exposure to pesticides (Fig. 4). The SPEAR index has been demonstrated to remain relatively constant among reference streams and, more importantly, it responds selectively to pesticide exposure in small streams across different biogeoregions (Beketov and Liess, 2008b; Liess and Von der Ohe, 2005; Schäfer et al., 2007; Schäfer et al., 2011c; von der Ohe et al., 2007).

In paper VII, we used the SPEAR indicator and measured pesticide toxicity in the water phase

of streams during the main insecticide application season to establish field based dose-response curves in order to estimate effect thresholds. In this study, we compiled field based data from Denmark, Germany, France, Finland, Belgium, Spain, Russia and Australia, and we found that significant changes in the macroinvertebrate communities occurred around log maxTU(D.magna) = -3. This threshold value is consistent with previous suggestions (Schäfer et al., 2007). The macroinvertebrate samples in all biogeoregions were collected on the physically best location in the sampled stream which may reduce the broad applicability of these findings.

The SPEAR index was, additionally, identified as the ecological indicator that explained the highest proportions of variations in macroinvertebrate community structure in paper VIII, in which we evaluated ecological effects of a TCE point source (described in paper IX, and ecosystem effects of the TCE plume were proposed to be unimportant for all stream dwelling organisms) and diffuse source pesticide pollution in Skensved stream (Zealand) along 12 kilometres of the stream that was characterised by high physical quality. The TCE plume discharging into the stream was found to be an insignificant anthropogenic stressor compared to diffuse source

Figure 4. Schematic overview of how macroinvertebrate species are classified as SPEcies At Risk (SPEAR) or not at risk (SPEnotAR). This figure is modified after (Liess et al., 2008)

18

pesticide pollution, and the SPEAR index was identified as the strongest ecological indicator tool for characterising ecological quality, and the SPEAR index responded most strongly to diffuse source pesticide contamination. However, as previously mentioned, in a set of streams on Zealand the sediment samples were found to contain significantly higher concentrations of the pesticides with highest toxicity for macroinvertebrates (e.g. pyrethroids and organochlorine insecticides) than storm flow water samples, and the strong correlations between the SPEAR index and measured pesticide toxicity was only obtained when using toxicity data from the sediment samples (papers VIII and B). Figure 5 summarises the results from all field studies conducted during this PhD in terms of macroinvertebrate community responses to diffuse source pesticide pollution. Remarkably, the SPEAR index totally explains 76% of the data variability which clearly emphasises the strong potential of this index as supplementary indicator tool to the existing indicators. However, on Zealand, the most impacted streams in terms of pesticide pollution (except Skensved stream) were, additionally, characterised by heavy

physical modifications due to channelisation, dredging and ground water abstraction (paper B). The poor physical properties of these streams may have imposed additional stress to the resident organisms, hereby further decreasing the SPEAR score. One relevant question is therefore if there is an interaction between physical habitat quality and pesticide impact on stream ecosystems. Importance of physical habitat properties in stream ecosystems

In spite of the clear strength in using biological traits for assessing risk and impact of pesticides, such as SPEAR, the potential interaction of different anthropogenic stressors may prompt altered effects (Matthaei et al., 2010; Wagenhoff et al., 2011). Since the effect of one stressor can cause reduced fitness of an organism, the effect of another stressor may increase. Conversely, the effect of one stressor can cause improved growth conditions for one organism (e.g. through the elimination of predators or competitors for food and space), which ultimately may increase the fitness of this organism. In consequence, the effect of another stressor on this organism may be reduced (Pedersen and Friberg, 2009).

One stressor that potentially interacts with diffuse source pesticide pollution in agricultural streams is the intensive physical maintenance of streams. Physical maintenance comprises activities like channelisation and dredging which is performed in order to optimise flow conveyance into the stream channel and within it. One side-effect of such maintenance activities is the heavy degradation of physical habitat diversity and quality in the streams. The physical conditions in heavily maintained streams are often characterised by high proportions of soft substrates and low variability in the depth and width of the streams (Pedersen, 2009). Moreover, this physical habitat degradation prompts generally more laminar flow conditions, but probably also produces more flow variability on the seasonal level which, consequently make the physical environment highly unstable. In addition, the generally increased low flow dominated by laminar flow reduces the reaeration capacity of the stream which at least periodically will also reduce the oxygen concentrations in the streams (Friberg et al., 2009; Pedersen, 2009).

The close correlation between stream macroinvertebrate communities and physical

Figure 5. %SPEAR abundance as a function of summed pesticide toxicity to Daphnia magna (log sum TU(D. magna). The two reaches that were sampled in each of the 14 stream sites on Funen were collapsed and treated as one sampling site. Measured pesticide toxicity is based on storm flow water samples for the Funen stream sites and on sediment samples for the Zealand stream sites (papers I, VIII and unpublished data). All macroinvertebrate samples were collected during summer (from June to August). The blue dashed lines indicate the 95% confidence intervals.

19

habitat characteristics such as substratum types and habitat heterogeneity has been documented in several studies (Brunke et al., 2001; Pedersen and Friberg, 2009). Especially species of Ephemeroptera, Plecoptera and Trichoptera (EPT) often require high current velocities, well mixed and oxygenated water and coarse substrates or the presence of specific species of submergent macrophytes (Giller and Malmqvist, 1998). In addition, many EPT species have narrow niche requirements that are associated with more stable hydromorphological environments, whereas macroinvertebrate taxa that are related to more disturbed environments (anthropogenic disturbance) with low habitat heterogeneity and high proportions of soft substrates often are generalists with less specific niche requirements (Dunbar et al., 2010). These traits are identical with the traits that have been identified as useful for organisms that can cope with periodic pesticide pollution (Liess and von der Ohe, 2005).

Interaction between physical habitat characteristics and pesticide stress

The previous paragraph emphasises the essentiality of high habitat heterogeneity if the stream is to support high biodiversity of especially the EPT taxa. The majority of EPT species are additionally very sensitive to pesticide pollution (Wogram and Liess, 2001). Moreover, the majority of EPT taxa are characterised as at risk in the SPEAR index (Liess and Von der Ohe, 2005). The pool of potential recolonising species is therefore severely reduced in reaches with degraded physical conditions. Moreover, resident species may become increasingly sensitive to pesticide stress due to sub-optimal physical habitat conditions. In consequence, the quality and heterogeneity of physical habitats could clearly influence species sensitivity to pesticide stress and, furthermore the outcome of an ecological pesticide indicator, such as the SPEAR index.

In paper II, we analysed an existing data set collected as part of the Danish monitoring program, NOVANA, to investigate whether it was possible to detect the effects of predicted pesticide exposure in streams that, additionally, were subject to multiple physical and chemical stressors. We used different biological metrics to evaluate the potential impact from different types of anthropogenic stress, and we found that local physical habitat quality was the overall governing

factor for the macroinvertebrate community structure. All applied biological metrics, including SPEAR, responded most strongly to local physical habitat characteristics, and none of them responded to predicted pesticide exposure. Schletterer et al. (2010) additionally insinuated that surber samples collected on soft substrates may be characterised by much lower abundance of SPEAR compared to surber samples collected on other types of substrate. Several factors could explain these results (see paper II for detailed discussion). However, since measured pesticide concentrations in Danish streams indeed did obtain sufficiently high concentrations to significantly impact the macroinvertebrate communities in Danish streams (papers I, VIII and A), poor physical habitat conditions may have been a factor influencing the effect of pesticides – or at least the SPEAR index score.

Riffle sections of streams were prioritised for macroinvertebrate sampling in previous studies developing and using the SPEAR index (Liess and Von der Ohe, 2005; Schletterer et al., 2010; Schäfer et al., 2007; Schäfer et al., 2011c; von der Ohe et al., 2007) which partly explains the tight correlation that was observed between measured pesticide toxicity and SPEAR index scores (see

Figure 6. %SPEAR abundance as a function of the log maximum TU for D. magna. Data points represent macroinvertebrate data from June and pesticides detected in storm flow water samples in the main insecticide application season in the region. The solid regression line indicates the correlation between data points for sites with heterogeneous physical habitat structure and the dashed line represent sites with homogeneous physical habitat structure. The figure is modified after Paper III.

20

also paper VII). In paper III, we investigated macroinvertebrate community responses to diffuse source pesticide contamination at stream sites with low or high habitat complexity and quality. We showed that the SPEAR index response to pesticides was different between stream reaches differing in physical habitat complexity and quality (Fig. 6). We found that this difference in the response of the SPEAR index was primarily due to the presence or absence of macroinvertebrate taxa with specific requirements for stream microhabitats characterised by hard substrate, turbulent flow and high oxygen concentrations. This suggests that the primary stressor acting on these taxa was related to physical habitat properties rather than diffuse source pesticide contamination. We also found that the difference in SPEAR index scores between sites with differing physical complexity was highest at intermediate levels of pesticide contamination whereas the difference was insignificant at highest levels of pesticide contamination (log mTU(D. magna) > -3). This suggests that sufficiently high levels of pesticide contamination overrule the possible positive effects of high habitat complexity and quality, whereas high habitat complexity and quality may remediate some of the effects of intermediate pesticide contamination.

21

Several previous studies document effects of diffuse source pesticides on the macroinvertebrate community structure in streams as discussed in the previous chapters. However, only few studies have addressed how changes in ecosystem structure resulting from pesticide contamination potentially translates into changed rates of ecosystem functions (Ecosystem services).

The conversion of coarse particulate organic matter into fine particulate organic matter is one such important ecosystem process that is mediated partly by microbial organisms (fungi and bacteria) and partly by shredding macroinvertebrates (Abelho, 2001). Microorganisms, especially freshwater hyphomycetes, are essential for the conversion of leaf litter into more palatable food resources for macroinvertebrate shredders, filtrators and collectors/gatherers (Bärlocher and Kendrick, 1975; Gessner et al., 2007). Macroinvertebrate shredders continue the conversion leaf litter into secondary production and finer particles of organic matter (Graca, 2001). Moreover, the food preference of shredders and their ingestion rate of the food material are strongly regulated by the nutritional quality of the food source which is governed by the internal nutrient concentrations of the leaves and, more importantly, the biomass of microbial decomposers (Graca, 2001). Moreover, some macroinvertebrates clearly prefer certain species of hyphomycetes colonisers, and the presence of such species stimulates the food ingestion rates for the shredding macroinvertebrates (Arsuffi and Suberkropp, 1989; Bundschuh et al., 2011).

Increased nitrate and phosphate concentrations that are side effects from agricultural production have been shown to stimulate fungal biomass and decomposition rates, since aquatic fungi and bacteria supplement their nutrient uptake from leaves with nutrients that are assimilated from the stream water (Gulis et al., 2006; Maharning and Bärlocher, 1996). However, other environmental parameters, such as pesticide contamination and increased sediment transport, being additional side effects from agricultural activities, may counteract the stimulating effects from eutrophication (Young et al., 2008). Aquatic hyphomycetes have storage organs in the form of

spores which can remain dormant in sediments for decades waiting for optimal substrate and growth conditions (Maharning and Bärlocher, 1996). Physical stream maintenance activities as channelisation, dredging and the removal of woody vegetation in the riparian areas may all be factors reducing the volume and diversity of remaining dormant spores in the sediments. This may ultimately affect the total conversion of leaf litter since a reduced species richness of aquatic fungi is known to reduce the overall macroinvertebrate shredding activity (Bärlocher et al., 2010; Gessner et al., 2007).

A series of field studies and laboratory studies have confirmed the stimulating effects of nitrate and phosphate on the growth and function of aquatic fungi (e.g. Chung and Suberkropp, 2008; Gulis and Suberkropp, 2003; Suberkropp, 1998). In a recent field study, however, different results were reported showing that, in spite of high nutrient concentrations, microbial activity was not increased in a set of Panamanian streams – which may be the result of additional toxic stress by e.g. pesticide pollution (Bärlocher et al., 2010).

Pesticide effects on freshwater fungi

Recent studies report that especially fungicides pose a significant threat for freshwater micro-organisms (Dijksterhuis et al., 2011; Maltby et al., 2009). Dijksterhuis et al. (2011), furthermore, showed that the potency of azole fungicides was particularly high for non-target aquatic fungi.

Moreover, azole fungicides are particularly interesting because 1) they are widely applied on agricultural fields and is the most frequently applied group of fungicides on Danish agricultural fields (except for mancozeb that is only applied to potatoes on sandy soils) (Danish EPA, 2010) and 2) they are often applied in tank mixtures with pyrethroid insecticides and are known to synergise the effect of pyrethroids on non-target crustaceans (Belden et al., 2007; Cedergreen et al., 2008). Azole fungicides are C14α-demethylase inhibitors that obstruct the biosynthesis of ergosterol (a significant part of the fungi cell walls) by inhibiting the activity of an enzyme belonging to the group of P450 monooxygenases (Copping and Hewitt, 1998).

3. Pesticide effects on ecosystem processes and indirect effects on macroinvertebrates

22

In paper IV, we investigated the leaf mass loss induced by microbial organisms in the 14 streams that were also used to study macroinvertebrate community responses to pesticide contamination (paper III). To present, only two field studies have been conducted investigating the effect of measured pesticide contamination on microbial organisms in streams (paper IV and Schäfer et al., 2011b). In Schäfer et al. (2011b), the authors used a modified Toxic Unit (TU) approach based on the unicellular green algae, Selenastrum capricornutum, as a surrogate measure for pesticide toxicity to freshwater fungi, because no toxicity data existed for freshwater fungi at the time of the study. In a set of Australian streams the authors found that microbial leaf decomposition significantly decreased with increasing log maxTU(Selenastrum capricornutum). In paper

IV we modified this TU approach since new ecotoxicity data for freshwater hyphomycetes was published in Dijksterhuis et al. (2011). We therefore used freshwater hyphomycetes toxicity data whenever available and supplied with toxicity data for S. capricornutum. For clarity this TU approach was called TUmicrobial. This measure for pesticide toxicity for microorganism decomposers was then linked to measured leaf mass loss at two reaches differing in physical complexity and quality in each of 14 streams (see also fig. 7). We found a strong correlation between log maxTUmicrobial and leaf decomposition rates induced by microbial organisms. High values of TUmicrobial were primarily due to fungicides (azoxystrobin and azole fungicides). Furthermore, we found that the decomposition rates were further reduced at sites with low physical habitat complexity and quality (fig. 8.). Sites with low physical habitat complexity and quality were characterised by low current velocities and high proportions of sand and silt

probably reflecting high sedimentation rates. Rates of microbial decomposition rates may decrease as a direct consequence of sedimentation, because microorganisms become curtailed in their ability to use external nutrient sources and oxygen (Bärlocher, 2005; Imberger et

Figure 7. Schematic presentation of the two sampling sites in each of the study streams. The sampling sites represent uniform (upstream) and diverse (downstream) physical habitat structure. Each site represents a 50 m reach. Current velocity is used as a surrogate measure for the spatial distribution of physical habitat types as current velocity has a direct effect on substrate types, turbulence and reaeration. This figure is from paper III.

Figure 8. The rate of microbial leaf decomposition (kmicrobial) as a function of the maximum Toxic Unit for microorganisms (mTUmicrobial) measured in 14 Danish streams with event triggered samplers. Data represents paired sites from each stream with the upstream site being characterised by uniform physical habitats and the downstream site being characterised by diverse physical habitats. This figure is from paper IV.

23

al., 2010). In addition, the effects of reduced oxygen availability may be further promoted by increased thickness of the boundary layer due to low current velocity and more laminar flow conditions (Friberg et al., 2009).

Pesticide effects on macroinvertebrate shredding activity

The other group of leaf decomposers, shredding macroinvertebrates, has also been shown to respond to pesticide stress. Schäfer et al. (2007) showed that macroinvertebrate shredding activity decreased with decreasing proportions of the macroinvertebrates belonging to SPEAR. Since the majority of SPEAR species belong to Ephemeroptera, Plecoptera and Trichoptera (EPT), and since many of these taxa, additionally, belong to the feeding functional group of shredders, this correlation was not surprising.

In paper IV, however, we did not find a similar correlation. The shredding activity in the 14 Funen streams did not respond to pesticide contamination or physical habitat properties. The shredding activity was primarily governed by the density of shredding macroinvertebrates at the sampling site which did not reflect the abundance of SPEAR species. The amphipod, Gammarus pulex, which is not classified as SPEAR, was dominating in terms of abundance at all sampling sites. This species is an omnivore but primarily act as a shredder (Marchant, 1981). The combination of diverse feeding strategies and relatively high recolonisation and reproduction potential (Marchant, 1981) may entail that Gammaridae has important competitive advantages in frequently disturbed and multistressor dominated environments. Other authors have reported that shredding activity primarily reflects the abundance of shredders and not the level of anthropogenic stress that is imposed upon the stream ecosystem (Ferreira et al., 2006; Gessner and Chauvet, 2002; Hagen et al., 2006). In contrast, e.g. Pascoal et al. (2003) and Young and Collier (2009) found significant responses of shredding activity to anthropogenic stressors. In paper VII we compiled field based data for shredding activity from the three existing studies where measured pesticide concentrations were also available (paper IV, Schäfer et al., 2007, 2011b). We attempted to find similarities of responses in ecosystem functions across different biogeoregions. However, whereas shredding

activity was closely linked to the abundance of SPEAR in French and Australian streams, this correlation was not significant for the Danish streams and was therefore excluded from the analysis (see discussion in paper VII).

Using microbial leaf litter decomposition rates as a proxy for microbial biomass (Bärlocher, 2005), the results in paper IV, additionally, indicate that macroinvertebrate shredding activity was independent of the total microbial biomass on the leaves (reflecting nutritional quality). However, in the literature it is consistently reported that shredding macroinvertebrates selectively choose highly palatable leaves with high microbial biomass over those with lower palatability and lower microbial biomass (Abelho, 2001; Graca, 2001). A few laboratory studies have reported reduced shredding activity of leaves that were exposed to azole fungicides due to changed fungal community structure (Bundschuh et al., 2011; Zubrod et al., 2011) or pyrethroid insecticides (Lauridsen et al., 2006). However, common for all three laboratory studies is that applied exposure concentrations and durations are not environmentally realistic. In consequence, controlled laboratory studies using environmentally realistic exposure scenarios are needed.

Indirect effects of fungicides on macroinvertebrate shredding activity

In paper V, we studied the decomposition of leaves that were previously exposed to the triazole fungicide propiconazole (two concentrations), the pyrethroid insecticide alpha-cypermethrin (two concentrations) or any combination of the two using environmentally realistic exposure concentrations and duration. Subsequent to the exposure, the leaf decomposition was studied in clean stream water in the presence or absence of an assemblage of macroinvertebrate shredders.

Microbial leaf decomposition rates were significantly reduced in the propiconazole treatments, and we showed that the reduction in leaf decomposition rates reflected loss of microbial biomass (C:N ratio was used as a proxy for microbial biomass) probably reflecting direct toxic effects of the fungicide. Not surprisingly, shredding activity was reduced in the alpha-cypermethrin treatments reflecting a direct toxic or a repelling effect. More importantly, we found that the macroinvertebrate assemblage,

24

additionally, responded to propiconazole treatments by increasing their ingestion rates of leaf litter with lower microbial biomass (higher C:N ratio) which is probably a compensatory behaviour that is initiated in order to meet energy requirements for basic metabolism and body functions (fig. 9). Similar behaviour was also observed by Zubrod et al. (2010), but the authors additionally showed that increased intake of leaf material by Gammaridae did not reflect increased uptake of digested food. Conversely, the actual energy uptake was decreased in spite of increased food acquisition probably reflecting increased stress that was imposed by the pesticide (Zubrod et al., 2010). These findings are highly important for the mechanistic understanding of ecotoxicological effects of pesticides in the field. Synergistic effects between compounds – even when exposure incidents are temporally separated – may occur through a combination of direct and indirect effects. When a macroinvertebrate species is exposed to sublethal concentrations of a pesticide, the basic metabolism may increase in response to the physiological stress that is imposed by the pesticide. Simultaneously, the nutritional quality of food resources (e.g. leaf material) may be reduced due to simultaneous or previous exposure to another pesticide, such as fungicides. As a consequence, the direct effect of the pesticide may increase due to decreased fitness of the organism. Indeed, the direct effect of insecticides on macroinvertebrates have been shown to increase when the fitness of the organisms is reduced due to competitive stress or food limitation (Beketov and Liess, 2005; Liess and Foit, 2010).

Figure 9. The rate of leaf decomposition (k) facilitated by microbial organisms (A) and an assemblage of macroinvertebrate shredders (B) as a function of the average C:N ratio of leaves that were previously exposed to one of eight different treatments with propiconazole, alpha-cypermethrin or the combination of the two (and control). Dashed lines represent 95% confidence intervals. Figure and caption are from paper V.

25

Several scientific questions have been addressed in this PhD thesis, and we have been able to provide answers for some of them. However, as is often the case, we have generated new questions during the process of answering questions. In paper VI we have focused on the perspectives of this PhD thesis, and we have addressed some of the obvious questions that arose during our studies and which should be addressed in future scientific research.

In the preceding decades, research focusing on the pesticide transport from fields to stream recipients has increased markedly, but still the linkage between normal agricultural practice and the amount of pesticides that is found in receiving streams is not sufficiently researched. For example, a recent study from California, United States , has shown that few sub-patches of fields adjacent to streams are responsible for most of the observed in-stream contamination, and some of this observation may be ascribed to the fact that even highly lipophilic substances are transported via macropores to tile drains at specific places on the fields where soil surface and drains are sufficiently connected via macropores (Weston et al., 2008). However, more studies are needed that investigate the patchy distribution of sub-field-blocks that significantly contribute to the in-stream pesticide contamination, and especially the contribution from tile drains needs to investigated more deeply. Moreover, more emphasis needs to be attributed to pesticides adsorbed to suspended particles and sediment since these complexes govern different exposure scenarios and bioavailability of the pesticides.

Indirect effects of pesticides on different groups of organisms in freshwater ecosystems have only been the focus of a limited number of studies (see e.g. Bundschuh et al., 2011; Rasmussen et al., 2008; Schulz and Liess, 2001a, b; Zubrod et al., 2010, 2011). However, it is clear that indirect effects are relevant for the interpretation of overall ecosystem effects, because the observed effects are the sum of direct and indirect effects and feed-back mechanisms. Fungicidal effects on non-target freshwater fungi, herbicidal effects on micro-algae and pyrethroid sorption to habitat surfaces and food sources but still being potentially bioavailable for macroinvertebrates are some important aspects that deserve focus in future research.

The significance of multiple stressors is one additionally important aspect in the pesticide research needing more substance. The field studies in this PhD provide the first comprehensive field studies considering diffuse source pesticide contamination and the loss of physical habitat complexity which is an additional side effect of agricultural practice. More effort should be channelled into the research on multiple stressor effects which in this PhD has been shown to potentially overrule the effects of pesticide contamination in terms of ecosystem structure as well as function (papers II, III and IV). Using biological traits to disentangle the effects of one stressor from another provides a very promising approach, such as SPEAR. The SPEAR approach is surely strong, as has also been documented in this PhD (papers III, VII and VIII), but more formal guidelines for sampling strategy and macroinvertebrate identification level are needed to optimise the approach which may overcome its current limitations in terms of disentangling pesticide effects from those of poor physical conditions.

Considering all of the study areas above, the combined results from these should be used for improving the quality and geographic prioritisation of mitigation instruments that sufficiently reduce the inadvertent effects of pesticides on non-target freshwater organisms. New or modified indicators and risk assessment tools are needed for assisting catchment managers in selecting appropriate mitigation strategies for pesticide application and losses to surface water ecosystems.

4. Perspectives

26

Abelho, M., 2001. From litterfall to breakdown in streams: A review. The scientific world 1, 656-680.

Anderson, B.S., Phillips, B.M., Hunt, J.W., Connor, V., Richard, N., Tjeerdema, R.S., 2006. Identifying primary stressors impacting macroinvertebrates in the Salinas River (California, USA): Relative effects of pesticides and suspended particles. Environmental Pollution 141, 402-408.

Arsuffi, T.L., Suberkropp, K., 1989. Selective feeding by shredders on leaf-colonizing stream fungi - Comparison of macroinvertebrate taxa. Oecologia 79, 30-37.

Bacey, J., Spurlock, F., 2007. Biological assessment of urban and agricultural streams in the California Central Valley. Environmental Monitoring and Assessment 130, 483-493.

Baird, D.J., Baker, C.J.O., Brua, R.B., Hajibabaei, M., McNicol, K., Pascoe, T., De Zwart, D., 2011. Toward a knowledge infrastructure for Traits-based ecological risk assessment. Integrated Environmental Assessment and Management 7, 209-215.

Baird, D.J., Van den Brink, P.J., 2007. Using biological traits to predict species sensitivity to toxic substances. Ecotoxicology and Environmental Safety 67, 296-301.

Baun, A., Ledin, A., Reitzel, L.A., Bjerg, P.L., Christensen, T.H., 2004. Xenobiotic organic compounds in leachates from ten Danish MSW landfills - chemical analysis and toxicity tests. Water Research 38, 3845-3858.

Beketov, M., Liess, M., 2012. Ecotoxicology and macroecology - Time for integration. Environmental Pollution 162, 247-254.

Beketov, M.A., Liess, M., 2005. Acute contamination with esfenvalerate and food limitation: Chronic effects on the mayfly, Cloeon dipterum. Environmental Toxicology and Chemistry 24, 1281-1286.

Beketov, M.A., Liess, M., 2008a. Potential of 11 pesticides to initiate downstream drift of stream macroinvertebrates. Archives of Environmental Contamination and Toxicology 55, 247-253.

Beketov, M.A., Liess, M., 2008b. Variability of pesticide exposure in a stream mesocosm system: Macrophyte-dominated vs. non-vegetated sections. Environmental Pollution 156, 1364-1367.

Belden, J.B., Gilliom, R.J., Martin, J.D., Ludy, M.J., 2007. Relative toxicity and occurrence patterns of pesticide mixtures in streams draining agricultural watersheds dominated by corn and soybean production. Integrated Environmental Assessment and Management 3, 90-100.

Berenzen, N., Lentzen-Godding, A., Probst, M., Schulz, H., Schulz, R., Liess, M., 2005. A comparison of predicted and measured levels of runoff-related pesticide concentrations in small

lowland streams on a landscape level. Chemosphere 58, 683-691.

Brown, C.D., Turner, N., Hollis, J., Bellamy, P., Biggs, J., Williams, P., Arnold, D., Pepper, T., Maund, S., 2006. Morphological and physico-chemical properties of British aquatic habitats potentially exposed to pesticides. Agriculture Ecosystems & Environment 113, 307-319.

Brunke, M., Hoffmann, A., Pusch, M., 2001. Use of mesohabitat-specific relationships between flow velocity and river discharge to assess invertebrate minimum flow requirements. Regulated Rivers-Research & Management 17, 667-676.

Bundschuh, M., Zubrod, J.P., Kosol, S., Maltby, L., Stang, C., Duester, L., Schulz, R., 2011. Fungal composition on leaves explains pollutant-mediated indirect effects on amphipod feeding. Aquatic Toxicology 104, 32-37.

Bärlocher, F., 2005. Freshwater fungal communities, in: Deighton, J., Oudemans, P., White, J. (Eds.), The fungal communitiy: its organization and role in the ecosystem. CRC Press, Florida, United States, pp. 39-59.

Bärlocher, F., Helson, J.E., Williams, D.D., 2010. Aquatic hyphomycete communities across a land-use gradient of Panamanian streams. Fundamental and Applied Limnology 177, 209-221.

Bärlocher, F., Kendrick, B., 1975. Leaf-conditioning by microorganisms. Oecologia 20, 359-362.

Carafa, R., Faggiano, L., Real, M., Munné, A., Ginebreda, A., Guasch, H., Flo, M., Tirapu, L., Von der Ohe, P.C., 2011. Water toxicity assessment and spatial pollution patterns identification in a Mediterranean River Basin district. Tools for water management and risk analysis. Science of the Total Environment 409, 4269-4279.

Castillo, L.E., Martinez, E., Ruepert, C., Savage, C., Gilek, M., Pinnock, M., Solis, E., 2006. Water quality and macroinvertebrate community response following pesticide applications in a banana plantation, Limon, Costa Rica. Science of the Total Environment 367, 418-432.

Cedergreen, N., Christensen, A.M., Kamper, A., Kudsk, P., Mathiassen, S.K., Streibig, J.C., Sorensen, H., 2008. A review of independent action compared to concentration addition as reference models for mixtures of compounds with different molecular target sites. Environmental Toxicology and Chemistry 27, 1621-1632.

Chung, N., Suberkropp, K., 2008. Influence of shredder feeding and nutrients on fungal activity and community structure in headwater streams. Fundamental and Applied Limnology 173, 35-46.

Copping, L.G., Hewitt, H.G., 1998. Chemistry and mode of action of crop protection agents. The Royal Society of Chemistry, Cambridge.

References

27

Cover, M.R., May, C.L., Dietrich, W.E., Resh, V.H., 2008. Quantitative linkages among sediment supply, streambed fine sediment, and benthic macroinvertebrates in northern California streams. Journal of the North American Benthological Society 27, 135-149.

Cui, X., Hunter, W., Yang, Y., Chen, Y., Gan, J., 2010. Bioavailability of sorbed phenanthrene and permethrin in sediments to Chironomus tentans. Aquatic Toxicology 98, 83-90.

Danish, Environmental, Protection, Agency, 2010. Pesticide Statistics, Orientering fra Miljøstyrelsen nr. 6 (In Danish), Copenhagen, Denmark.

Davies, P.E., Cook, L.S.J., 1993. Catastrophic macroinvertebrate drift and sublethal effects on brown trout, Salmo trutta caused by cypermethrin spraying on a Tasmanian stream. Aquatic Toxicology 27, 201-224.

Dijksterhuis, J., van Doorn, T., Samson, R., Postma, J., 2011. Effects of seven fungicides on non-target aquatic fungi. Water Air and Soil Pollution 222, 421-425.

Ding, Y.P., Harwood, A.D., Foslund, H.M., Lydy, M.J., 2010. Distribution and toxicity of sediment-associated pesticides in urban and agricultural waterways from Illinois, USA. Environmental Toxicology and Chemistry 29, 149-157.

Domagalski, J.L., Weston, D.P., Zhang, M.H., Hladik, M., 2010. Pyrethroid insecticide concentrations and toxicity in streambed sediments and loads in surface waters of the San Joaquin Valey, California, USA. Environmental Toxicology and Chemistry 29, 813-823.

Dunbar, M.J., Warren, M., Extence, C., Baker, L., Cadman, D., Mould, D.J., Hall, J., Chadd, R., 2010. Interaction between macroinvertebrates, discharge and physical habitat in upland rivers. Aquatic Conservation-Marine and Freshwater Ecosystems 20, S31-S44.

Ensminger, M., Bregin, R., Spurlock, F., Goh, K.S., 2011. Pesticide concentrations in water and sediment and associated invertebrate toxicity in Del Puerto and Orestimba Creeks, California, 2007-2008. Journal of Environmental Monitoring and Assessment 175, 573-587.

Feo, M.L., Ginebreda, A., Eljarrat, E., Barcelo, D., 2010. Presence of pyrethroid pesticides in water and sediments of Ebro River Delta. Journal of Hydrology 393, 156-162.

Ferreira, V., Gulis, V., Graca, M.A.S., 2006. Whole-stream nitrate addition affects litter decomposition and associated fungi but not invertebrates. Oecologia 149, 718-729.

Fojut, T.L., Young, T.M., 2011. Pyrethroid sorption to Sacramento river suspended solids and bed sediments. Environmental Toxicology and Chemistry 30, 787-792.

Friberg, N., Sandin, L., Pedersen, M.L., 2009. Assessing the effects of hydromorphological degradation on macroinvertebrate indicators in rivers: Examples, constraints and outlook. Integrated Environmental Assessment and Management 5, 86-96.

Gan, J., Lee, S.J., Liu, W.P., Haver, D.L., Kabashima, J.N., 2005. Distribution and persistence of pyrethroids in runoff sediments. Journal of Environmental Quality 34, 836-841.

Gessner, M.O., Chauvet, E., 2002. A case for using litter breakdown to assess functional stream integrity. Ecological Applications 12, 498-510.

Gessner, M.O., Gulis, V., Kuehn, K.A., Chauvet, E., Suberkropp, K., 2007. Fungal decomposers of plant litter in aquatic ecosystems, in: Kubicek, C.P.a.D., I.S. (Ed.), Environmental and microbial relationships, 2nd edition. Springer Verlag, Berlin.

Giller, P.S., Malmqvist, B., 1998. The biology of streams and rivers, 1st ed. Oxford University Press, Oxford, UK.

Graca, M.A.S., 2001. The role of invertebrates on leaf litter decomposition in streams - A review. International Review of Hydrobiology 86, 383-393.

Guasch, H., Munoz, I., Roses, N., Sabater, S., 1997. Changes in atrazine toxicity throughout succession of stream periphyton communities. Journal of Applied Phycology 9, 137-146.

Gulis, V., Ferreira, V., Graca, M.A.S., 2006. Stimulation of leaf litter decomposition and associated fungi and invertebrates by moderate eutrophication: implications for stream assessment. Freshwater Biology 51, 1655-1669.

Gulis, V., Suberkropp, K., 2003. Interactions between stream fungi and bacteria associated with decomposing leaf litter at different levels of nutrient availability. Aquatic Microbial Ecology 30, 149-157.

Götz, C.W., Stamm, C., Fenner, K., Singer, H., Schärer, M., Hollender, J., 2010. Targeting aquatic microcontaminants for monitoring: exposure categorization and application to the Swiss situation. Environmental Science and Pollution Research 17, 341-354.

Hagen, E.M., Webster, J.R., Benfield, E.F., 2006. Are leaf breakdown rates a useful measure of stream integrity along an agricultural landuse gradient? Journal of the North American Benthological Society 25, 330-343.

Heckmann, L.H., Friberg, N., 2005. Macroinvertebrate community response to pulse exposure with the insecticide lambda-cyhalothrin using in-stream mesocosms. Environmental Toxicology and Chemistry 24, 582-590.

Hladik, M.L., Kuivila, K.M., 2009. Assessing the Occurrence and Distribution of Pyrethroids in

28

Water and Suspended Sediments. Journal of Agricultural and Food Chemistry 57, 9079-9085.

Imberger, S.J., Thompson, R.M., Grace, M.R., 2010. Searching for effective indicators of ecosystem function in urban streams: assessing cellulose decomposition potential. Freshwater Biology 55, 2089-2106.

Jergentz, S., Mugni, H., Bonetto, C., Schulz, R., 2004. Runoff-related endosulfan contamination and aquatic macroinvertebrate response in rural basins near Buenos Aires, Argentina. Archives of Environmental Contamination and Toxicology 46, 345-352.

Kreuger, J., 1998. Pesticides in stream water within an agricultural catchment in southern Sweden, 1990-1996. The Science of the Total Environment 216, 227-251.

Kreuger, J., Tornqvist, L., 1998. Multiple regression analysis of pesticide occurrence in streamflow related to pesticide properties and quantities applied. Chemosphere 37, 189-207.

Kreutzweiser, D.P., Kingsbury, P.D., 1987. Permethrin treatments in Canadian forests. Part 2: Impact on stream macroinvertebrates. Pesticide Science 19, 49-60.

Kronvang, B., Laubel, A., Larsen, S.E., Friberg, N., 2003. Pesticides and heavy metals in Danish streambed sediment. Hydrobiologia 494, 93-101.

Kronvang, B., Strom, H.L., Hoffmann, C.C., Laubel, A., Friberg, N., 2004. Subsurface tile drainage loss of modern pesticides: field experiment results. Water Science and Technology 49, 139-147.

Lacas, J.G., Voltz, M., Gouy, V., Carluer, N., Gril, J.J., 2005. Using grassed strips to limit pesticide transfer to surface water: a review. Agronomy for Sustainable Development 25, 253-266.

Lauridsen, R.B., Kronvang, B., Friberg, N., 2006. Occurrence of sediment-bound pyrethroids in danish streams and their impact on ecosystem function. Water Air and Soil Pollution 6, 423-432.

Leonard, A.W., Hyne, R.V., Lim, R.P., Pablo, F., Van den Brink, P.J., 2000. Riverine endosulfan concentrations in the Namoi River, Australia: Link to cotton field runoff and macroinvertebrate population densities. Environmental Toxicology and Chemistry 19, 1540-1551.

Liess, M., 1994. Pesticide impact on macroinvertebrate communities of running waters in agricultural ecosystems, in: Sladeckova, A. (Ed.), International Association of Theoretical and Applied Limnology - Proceedings, Vol 25, Pt 4, pp. 2060-2062.

Liess, M., Brown, C.D., Dohmen, G.P., Duquesne, S., Hart, A., Heimbach, F., Kreuger, J., Lagadic, L., Maund, S., Reinert, W., Streloke, M., Tarazona, J.V., 2005. Effects of pesticides in the field. SETAC press, Berlin, Germany.

Liess, M., Foit, K., 2010. Intraspecific competition delays recovery of population structure. Aquatic Toxicology 97, 15-22.

Liess, M., Schafer, R.B., Schriever, C.A., 2008. The footprint of pesticide stress in communities-Species traits reveal community effects of toxicants. Science of the Total Environment 406, 484-490.

Liess, M., Schulz, R., 1999. Linking insecticide contamination and population response in an agricultural stream. Environmental Toxicology and Chemistry 18, 1948-1955.

Liess, M., Schulz, R., Liess, M.H.D., Rother, B., Kreuzig, R., 1999. Determination of insecticide contamination in agricultural headwater streams. Water Research 33, 239-247.

Liess, M., Schulz, R., Neumann, M., 1996. A method for monitoring pesticides bound to suspended particles in small streams. Chemosphere 32, 1963-1969.

Liess, M., Von der Ohe, P.C., 2005. Analyzing effects of pesticides on invertebrate communities in streams. Environmental Toxicology and Chemistry 24, 954-965.

MacArthur, R.H., Wilson, E.O., 1967. The theory of Island Biogeography. Princeton University Press, Princeton, N.J.

Maharning, A.R., Bärlocher, F., 1996. Growth and reproduction in aquatic hyphomycetes. Mycologia 88, 80-88.

Maltby, L., Brock, T.C.M., van den Brink, P.J., 2009. Fungicide Risk Assessment for Aquatic Ecosystems: Importance of Interspecific Variation, Toxic Mode of Action, and Exposure Regime. Environmental Science & Technology 43, 7556-7563.

Marchant, R., 1981. The ecology of Gammarus in running water, in: Lock, M.A., Williams, D.D. (Eds.), Perspectives in running water ecology. Plenum Press, New York, USA, pp. 225-250.

Matthaei, C.D., Piggott, J.J., Townsend, C.R., 2010. Multiple stressors in agricultural streams: interactions among sediment addition, nutrient enrichment and water abstraction. Journal of Applied Ecology 47, 639-649.

Maund, S.J., Hamer, M.J., Lane, M.C.G., Farrelly, E., Rapley, J.H., Goggin, U.M., Gentle, W.E., 2002. Partitioning, bioavailability, and toxicity of the pyrethroid insecticide cypermethrin in sediments. Environmental Toxicology and Chemistry 21, 9-15.

Metcalfe-Smith, J.L., 1996. Biological water-quality assessment of rivers: use of macroinvertebrate communities, in: Petts, G., Calow, P. (Eds.), River Restoration. Blackwell Science, Oxford, UK.

Millenium, Ecosystem, Assessment, 2005. Millenium Ecosystem Assessment. Ecosystems and human

29

well-being: synthesis. Island Press, Washington, DC.

Moore, M.T., Bennett, E.R., Cooper, C.M., Smith, S., Shields, F.D., Milam, C.D., Farris, J.L., 2001. Transport and fate of atrazine and lambda-cyhalothrin in an agricultural drainage ditch in the Mississippi Delta, USA. Agriculture Ecosystems & Environment 87, 309-314.

Muir, D.C.G., Howard, P.H., 2006. Are there other persistent organic pollutants? A challenge for environmental chemists. Environmental Science & Technology 40, 7157-7166.

Muirhead-Thomson, R.C., 1978. Lethal and behavioral impact of chlorpyrifos methyl and temephos on select stream macroinvertebrates - experimental studies on downstream drift. Archives of Environmental Contamination and Toxicology 7, 139-147.

Neumann, M., Liess, M., Schulz, R., 2003. A qualitative sampling method for monitoring water quality in temporary channels or point sources and its application to pesticide contamination. Chemosphere 51, 509-513.

Neumann, M., Schutz, R., Schafer, K., Muller, W., Mannheller, W., Liess, M., 2002. The significance of entry routes as point and non-point sources of pesticides in small streams. Water Research 36, 835-842.

Nørum, U., Friberg, N., Jensen, M., Pedersen, J., Bjerregaard, P., 2010. Behavioural changes in three species of freshwater macroinvertebrates exposed to the pyrethroid lambda-cyhalothrin: Laboratory and stream microcosm studies. Aquatic Toxicology 98, 328-335.

Palmquist, K., Fairbrother, A., Salatas, J., Guiney, P.D., 2011. Environmental Fate of Pyrethroids in Urban and Suburban Stream Sediments and the Appropriateness of Hyalella azteca Model in Determining Ecological Risk. Integrated Environmental Assessment and Management 7, 325-335.

Pascoal, C., Pinho, M., Cassio, F., Gomes, P., 2003. Assessing structural and functional ecosystem condition using leaf breakdown: studies on a polluted river. Freshwater Biology 48, 2033-2044.

Pedersen, M.L., 2009. Effects of channelisation, riparian structure and catchment area on physical habitats in small lowland streams. Fundamental and Applied Limnology 174, 89-99.

Pedersen, M.L., Friberg, N., 2009. Influence of disturbance on habitats and biological communities in lowland streams. Fundamental and Applied Limnology 174, 27-41.

Pedersen, M.L., Friberg, N., Larsen, S.E., 2004. Physical habitat structure in Danish lowland streams. River Research and Applications 20, 653-669.

Poff, N.L., 1997. Landscape filters and species traits: Towards mechanistic understanding and prediction in stream ecology. Journal of the North American Benthological Society 16, 391-409.

Probst, M., Berenzen, N., Lentzen-Godding, A., Schulz, R., 2005a. Scenario-based simulation of runoff-related pesticide entries into small streams on a landscape level. Ecotoxicology and Environmental Safety 62, 145-159.

Probst, M., Berenzen, N., Lentzen-Godding, A., Schulz, R., Liess, M., 2005b. Linking land use variables and invertebrate taxon richness in small and medium-sized agricultural streams on a landscape level. Ecotoxcology and Environmental Safety 60, 140-146.

Rasmussen, J.J., Friberg, N., Larsen, S.E., 2008. Impact of lambda-cyhalothrin on a macroinvertebrate assemblage in outdoor experimental channels: Implications for ecosystem functioning. Aquatic Toxicology 90, 228-234.

Relyea, R., Hoverman, J., 2006. Assessing the ecology in ecotoxicology: a review and synthesis in freshwater systems. Ecology Letters 9, 1157-1171.

Rubach, M.N., Ashauer, R., Buchwalter, D.B., De Lange, H.J., Hamer, M.J., Preuss, T.G., Töpke, K., Maund, S., 2011. Framework for traits-based assessment in ecotoxicology. Integrated Environmental Assessment and Management 7, 172-186.

Rubach, M.N., Baird, D.J., Van den Brink, P.J., 2010. A new method for ranking mode-specific sensitivity of freshwater arthorpods to insecticides and its relationship to biological traits. Environmental Toxicology and Chemistry 29, 476-487.

Sandin, L., 2009. The relationship between land-use, hydromorphology and river biota at different spatial and temporal scales: a synthesis of seven case studies. Fundamental and Applied Limnology 174, 1-5.

Schletterer, M., Fureder, L., Kuzovlev, V.V., Beketov, M.A., 2010. Testing the coherence of several macroinvertebrate indices and environmental factors in a large lowland river system (Volga River, Russia). Ecological Indicators 10, 1083-1092.

Schriever, C.A., von der Ohe, P.C., Liess, M., 2007. Estimating pesticide runoff in small streams. Chemosphere 68, 2161-2171.

Schulz, R., 2004. Field studies on exposure, effects, and risk mitigation of aquatic nonpoint-source insecticide pollution: A review. Journal of Environmental Quality 33, 419-448.

Schulz, R., Hauschild, M., Ebeling, M., Nanko-Drees, J., Wogram, J., Liess, M., 1998. A qualitative field method for monitoring pesticides in the edge-of-field runoff. Chemosphere 36, 3071-3082.

30

Schulz, R., Liess, M., 1999. A field study of the effects of agriculturally derived insecticide input on stream macroinvertebrate dynamics. Aquatic Toxicology 46, 155-176.

Schulz, R., Liess, M., 2000. Toxicity of fenvalerate to caddisfly larvae: chronic effects of 1-vs 10-h pulse-exposure with constant doses. Chemosphere 41, 1511-1517.

Schulz, R., Liess, M., 2001a. Acute and chronic effects of particle-associated fenvalerate on stream macroinvertebrates: A runoff simulation study using outdoor microcosms. Archives of Environmental Contamination and Toxicology 40, 481-488.

Schulz, R., Liess, M., 2001b. Toxicity of aqueous-phase and suspended particle-associated fenvalerate: Chronic effects after pulse-dosed exposure of Limnephilus lunatus (Trichoptera). Environmental Toxicology and Chemistry 20, 185-190.

Schäfer, R., Van den Brink, P.J., Liess, M., 2011a. Impacts of pesticides on freshwater ecosystems, in: Sanchez-Bayo, F., van den Brink, P.J., Mann, R.M. (Eds.), Ecological impacts of toxic chemicals. Bentham, Bussum, Netherlands.

Schäfer, R.B., Bundschuh, M., Rouch, D.A., E., S., Von der Ohe, P.C., Pettigrove, V., Schulz, R., Nugegoda, D., Kefford, B., 2011b. Effects of pesticide toxicity, salinity and other environmental variables on selected ecosystem functions in streams and the relevance for ecosystem services. Science of the Total Environment 415, 69-78.

Schäfer, R.B., Caquet, T., Siimes, K., Mueller, R., Lagadic, L., Liess, M., 2007. Effects of pesticides on community structure and ecosystem functions in agricultural streams of three biogeographical regions in Europe. Science of the Total Environment 382, 272-285.

Schäfer, R.B., Pettigrove, V., Rose, G., Allinson, G., Wightwick, A., von der Ohe, P.C., Shimeta, J., Kuhne, R., Kefford, B., 2011c. Effects of Pesticides Monitored with Three Sampling Methods in 24 Sites on Macroinvertebrates and Microorganisms. Environmental Science & Technology 45, 1665-1672.

Sibley, P.K., Kaushik, N.K., Kreutzweiser, D.P., 1991. Impact of a pulse application of permethrin on the macroinvertebrate community of a headwater stream. Environmental Pollution 70, 35-55.

Stehle, S., Elsaesser, D., Gregoire, C., Imfeld, G., Niehaus, E., Passeport, E., Payraudeau, S., Schaefer, R.B., Tournebize, J., Schulz, R., 2011. Pesticide Risk Mitigation by Vegetated Treatment Systems: A Meta-Analysis. Journal of Environmental Quality 40, 1068-1080.

Straus, D.L., Chambers, J.E., 1995. Inhibition of actetylcholinesterase and aliesterases of fingerling channel catfish by chlorpyrifos, parathion and

S,S,S-tributyl phosphorottrithioate (DEF). Aquatic Toxicology 33, 311-324.

Suberkropp, K., 1998. Effect of dissolved nutrients on two aquatic hyphomycetes growing on leaf litter. Mycological Research 102, 998-1002.

Sørensen, P.B., Bruggemann, R., Carlsen, L., Mogensen, B.B., Kreuger, J., Pudenz, S., 2003. Analysis of monitoring data of pesticide residues in surface waters using partial order ranking theory. Environmental Toxicology and Chemistry 22, 661-670.

Thiere, G., Schulz, R., 2004. Runoff-related agricultural impact in relation to macroinvertebrate communities of the Lourens River, South Africa. Water Research 38, 3092-3102.

Tomlin, C.D.S., 2000. The pesticide manual, a wold compendium. Crop Protection Publications, Farnham, Surrey, UK.

Van Den Brink, P.J., Rubach, M.N., Culp, J.M., Pascoe, T., Maund, S.J., Baird, D.J., 2011. Traits-based ecological risk assessment (TERA): Realizing the potential of ecoinformatics approaches in ecotoxicology. Integrated Environmental Assessment and Management 7, 169-171.

Vijverberg, H.P.M., van den Bercken, J., 1990. Neurotoxicological effects and the mode of action of pyrethroid insecticides. Critical Reviews in Toxicology 21, 105-126.

von der Ohe, P.C., De Deckere, E., Prüss, A., Muñoz, I., Wolfram, G., Villagrasa, M., Ginebreda, A., Hein, M., Brack, W., 2009. Toward an integrated assessment of the ecological and chemical status of European river basins. Integrated Environmental Assessment and Management 5, 50-61.

von der Ohe, P.C., Dulio, V., Slobodnik, J., De Deckere, E., Kühne, R., Ebert, R., Ginebreda, A., De Cooman, W., Schüürmann, G., Brack, W., 2011. A new risk assessment approach for the prioritization of 500 classical emerging organic microcontaminants as potential river basin specific pollutants under the European Water Framework Directive. Science of the Total Environment 409.

von der Ohe, P.C., Prub, A., Schafer, R.B., Liess, M., de Deckere, E., Brack, W., 2007. Water quality indices across Europe - a comparison of the good ecological status of five river basins. Journal of Environmental Monitoring 9, 970-978.

Wagenhoff, A., Townsend, C.R., Phillips, N., Matthaei, C.D., 2011. Subsidy-stress and multiple-stressor effects along gradients of deposited fine sediment and dissolved nutrients in a regional set of streams and rivers. Freshwater Biology 56, 1916-1936.

Wallace, J.B., Lugthart, G.J., Cuffney, T.F., Schurr, G.A., 1989. The impact of repeated insecticidal treatments on drift and benthos of a headwater stream. Hydrobiologia 179, 135-147.

31

Wauchope, R.D., 1978. Pesticide content of surface-water draining from agricultural fields - review. Journal of Environmental Quality 7, 459-472.

Weston, D.P., Holmes, R.W., You, J., Lydy, M.J., 2005. Aquatic toxicity due to residential use of pyrethroid insecticides. Environmental Science & Technology 39, 9778-9784.

Weston, D.P., Lydy, M.J., 2010. Urban and Agricultural Sources of Pyrethroid Insecticides to the Sacramento-San Joaquin Delta of California. Environmental Science & Technology 44, 1833-1840.

Weston, D.P., You, J., Harwood, A.D., Lydy, M.J., 2009. Whole sediment toxicity identification evaluation tools for pyrtethroid insecticides: III. temperature manipulation. Environmental Toxicology and Chemistry 28, 173-180.

Weston, D.P., Zhang, M.H., Lydy, M.J., 2008. Identifying the cause and source of sediment toxicity in an agriculture-influenced creek. Environmental Toxicology and Chemistry 27, 953-962.

Wiggers, L., 1999. Pesticides in streams, brooks and lakes in the municipality of Aarhus (in Danish), Technical Report. Aarhus County.

Wogram, J., Liess, M., 2001. Rank ordering of macroinvertebrate species sensitivity to toxic compounds by comparison with that of Daphnia magna. Bulletin of Environmental Contamination and Toxicology 67, 360-367.

Yang, W., Gan, J., Hunter, W., Spurlock, F., 2006a. Effect of suspended solids on bioavailability of pyrethroid insecticides. Environmental Toxicology and Chemistry 25, 1585-1591.

Yang, W., Spurlock, F., Liu, W., Gan, J., 2006b. Inhibition of aquatic toxicity of pyrethroid insecticides by suspended sediment. Environmental Toxicology and Chemistry 25, 1913-1919.

Young, R.G., Collier, K.J., 2009. Contrasting responses to catchment modification among a range of functional and structural indicators of river ecosystem health. Freshwater Biology 54, 2155-2170.

Young, R.G., Matthaei, C.D., Townsend, C.R., 2008. Organic matter breakdown and ecosystem metabolism: functional indicators for assessing river ecosystem health. Journal of the North American Benthological Society 27, 605-625.

Zubrod, J.P., Bundschuh, M., Feckler, A., Englert, D., Schulz, R., 2011. Ecotoxicological impact of the fungicide tebuconazole on an aquatic decomposer-detritivore system. Environmental Toxicology and Chemistry 30, 2718-2724.

Zubrod, J.P., Bundschuh, M., Schulz, R., 2010. Effects of subchronic fungicide exposure on the energy processing of Gammarus fossarum (Crustacea;

Amphipoda). Ecotoxicology and Environmental Safety 73, 1674-1680.

Åkerblom, N., Arbjork, C., Hedlund, M., Goedkoop, W., 2008. Deltamethrin toxicity to the midge Chironomus riparius Meigen - Effects of exposure scenario and sediment quality. Ecotoxicology and Environmental Safety 70, 53-60.

32

Included papers

Paper I Rasmussen, J.J., Baattrup-Pedersen, A., Wiberg-Larsen, P., McKnight, U.S., Kronvang, B. 2011. Buffer strip width and agricultural pesticide contamination in Danish lowland streams: Implications for stream and riparian management. Ecological Engineering 37, 1990-1997.

Paper II Rasmussen, J.J., Baattrup-Pedersen, A., Larsen, S.E., Kronvang, B. 2011. Local physical habitat quality clouds the effect of predicted pesticide runoff from agricultural land in Danish streams. Journal of Environmental Monitoring 13, 943-950.

Paper III Rasmussen, J.J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Friberg, N., Kronvang, B. 2012. Stream habitat structure influences macroinvertebrate response to pesticides. Environmental Pollution 164, 142-149.

Paper IV Rasmussen, J.J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Monberg, R.J., Kronvang, B. 2012. Impacts of pesticides and natural stressors on leaf litter decomposition in agricultural streams. Science of the Total Environment 416, 148-155.

Paper V Rasmussen, J.J., Monberg, R.J., Baattrup-Pedersen, A., Cedergreen, N., Wiberg-Larsen, P., Strobel, B., Kronvang, B. 2012. Effects of a triazole fungicide and a pyrethroid insecticide on the decomposition of leaves in the presence or absence of macroinvertebrate shredders. Submitted to Aquatic Toxicology.

Paper VI Rasmussen, J.J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Monberg, R.J., McKnight, U.S., Kronvang, B. 2011. Ny viden om effekter af pesticider i vandløb. Vand & Jord 18, 143-147.

Paper VII Schäfer, R., von der Ohe, P.C., Rasmussen, J.J., Kefford, B., Beketov, M., Schulz, R., Liess, M. 2012. Thresholds for the effects of pesticides on invertebrate communities and leaf breakdown in stream ecosystems. Submitted to Environmental Science and Technology.

Paper VIII McKnight, U.S., Rasmussen, J.J., Kronvang, B., Bjerg, P.L., Binning, P.J. 2012. Integrated assessment of chemical stressors on surface water ecosystems. Submitted to Science of the Total Environment.

Paper IX McKnight, U.S., Funder, S.G., Rasmussen, J.J., Finkel, M., Bjerg, P.L., Binning, P.J. 2010. An integrated model for assessing the risk of TCE groundwater contamination to human receptors and surface water ecosystems. Ecological Engineering 36, 1126-1137.

Paper I

Bs

JBa

b

a

ARRAA

KBPRWN

1

atpicflpKAipcP

0d

Ecological Engineering 37 (2011) 1990– 1997

Contents lists available at ScienceDirect

Ecological Engineering

j o ur nal homep age : www.elsev ier .com/ locate /eco leng

uffer strip width and agricultural pesticide contamination in Danish lowlandtreams: Implications for stream and riparian management

es J. Rasmussena,∗, Annette Baattrup-Pedersena, Peter Wiberg-Larsena, Ursula S. McKnightb,rian Kronvanga

Department of Freshwater Ecology, National Environmental Research Institute, Aarhus University, Vejlsøvej 25, 8600 Silkeborg, DenmarkDepartment of Environmental Engineering, Technical University of Denmark, Miljøvej, Building 113, 2800 Kgs. Lyngby, Denmark

r t i c l e i n f o

rticle history:eceived 10 June 2011eceived in revised form 26 July 2011ccepted 7 August 2011vailable online 9 September 2011

eywords:uffer stripesticidesunoffater framework directive

a b s t r a c t

According to the European Water Framework Directive, member states are obliged to ensure that allsurface water bodies achieve at least good ecological status and to identify major anthropogenic stressors.Non-point source contamination of agricultural pesticides is widely acknowledged as one of the mostimportant anthropogenic stressors in stream ecosystems.

We surveyed the occurrence of 31 pesticides and evaluated their potential toxicity for benthic macroin-vertebrates using Toxic Units (TU) in 14 Danish 1st-and 2nd-order streams in bed sediments and streamwater during storm flow and base flow. Total pesticide concentrations and toxic potential were highestduring storm flow events with maximum TU ranging from −6.63 to −1.72. We found that minimumbuffer strip width in the near upstream area was the most important parameter governing TU. Further-more, adding a function for minimum buffer strip width to the Runoff Potential (RP) model increased its

on-point sources power to predict measured TUs from 46% to 64%. However, including a function for tile drainage capacityis probably equally important and should be considered in future research in order to further optimisethe RP model. Our results clearly emphasise the importance of considering buffer strips as risk mitigationtools in terms of non-point source pesticide contamination. We furthermore apply our results for dis-cussing the minimum dimensions that vegetated buffer strips should have in order to sufficiently protectstream ecosystems from pesticide contamination and maintain good ecological status.

© 2011 Elsevier B.V. All rights reserved.

mfaawaodie

p

. Introduction

Non-point source contamination of streams with pesticidespplied in agricultural production is widely acknowledged as one ofhe greatest stressors to stream ecosystems, and various routes foresticide transport from the field to stream recipients have been

dentified (Neumann et al., 2002; Schulz, 2004). There is a clearonsensus in the existing literature verifying surface runoff andow through tile-drains as the most important pathways for non-oint pesticide losses in agricultural catchments (Kreuger, 1998;ronvang et al., 2004; Neumann et al., 2002; Wauchope, 1978).s a consequence, the highest pesticide concentrations occur dur-

ng heavy precipitation events, and the footprint of pesticides is
roposed to be more distinct in small streams due to a closeronnectivity between land and stream (Kreuger and Brink, 1988;robst et al., 2005; Schulz, 2004).
∗ Corresponding author. Tel.: +45 89201757.E-mail address: [email protected] (J.J. Rasmussen).

amcetoi

925-8574/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.ecoleng.2011.08.016

According to the European Water Framework Directive (WFD),ember states are obliged to measure and ensure that all sur-

ace water bodies achieve at least good ecological status within defined timetable (European Commission, 2000). Requirementsre not only to assess the overall ecological quality of surfaceaters, but also to identify the major environmental and/or

nthropogenic drivers of ecological degradation and the extentf impairment. Several biotic indices and multi-metric proce-ures have been developed attempting to robustly characterise the

mpact of selected stressors that result in the deviation from goodcological status (Furse et al., 2006).

Non-point source pesticide contamination of rivers potentiallyoses a threat to all stream dwelling organisms (Liess et al., 2005),nd there is a growing interest to develop and provide field-basedodels to assist in characterising the non-point source pesticide

ontamination that originates from agricultural practices (Friberg
t al., 2003; Schäfer et al., 2007, 2011a; Schulz, 2004). However,here is still a need for additional studies that investigate the loss,ccurrence and fate of agricultural pesticides in streams and theirmpact on stream biota. Establishing causal relationships between
dx.doi.org/10.1016/j.ecoleng.2011.08.016

http://www.sciencedirect.com/science/journal/09258574

http://www.elsevier.com/locate/ecoleng

mailto:[email protected]


Engin

ptc(vpitFtcrlisbkspbrt

fpcmmabb

2

2

smmdtca

2

fnaaoospmo(beid

abuwdwscob

2

fTacp2

cOtttiflwpscdscT

Jctoswe

fliposcmot

csaub

J.J. Rasmussen et al. / Ecological

esticides and their impact on flora and fauna is difficult dueo natural variability in stream ecosystem communities and theo-existing pressures from several other anthropogenic stressorsLiess et al., 2005; Rasmussen et al., 2011). However, Liess andon der Ohe (2005) introduced the SPEcies At Risk indicator foresticides (SPEARpesticides), which has been validated as a selective

ndicator that successfully separates the effects of pesticides fromhose of other anthropogenic stressors (Schäfer et al., 2007, 2011a).urthermore, Schriever et al. (2007b) found that SPEARpesticides washe biological parameter best describing stream macroinvertebrateommunity responses to a modelled indicator of pesticide surfaceunoff (RP). In contrast, Rasmussen et al. (2011) were unable toink RP with SPEARpesticides using a large dataset of small Dan-sh streams, which could be due to the presence of wider buffertrips along Danish streams compared to German streams. Sinceuffer strip information is not integrated into the RP model but isnown to significantly influence pesticide runoff, different buffertrip characteristics between the two sets of study streams canlausibly explain the different results. Implementing a function foruffer strip width (representing a simplified measure for pesticideunoff retaining capacity) might, therefore, significantly improvehe predictive power of the RP model.

In this study we screened 14 Danish 1st-and 2nd-order streamsor pesticides that are frequently applied in normal agriculturalractices in their respective catchments. The study aims were to (1)haracterise pesticide occurrence and potential toxicity for benthicacroinvertebrates in Danish streams, (2) identify the environ-ental parameters that most strongly govern pesticide occurrence

nd toxicity, and (3) improve the predictive power of the RP modely using detailed environmental data and by adding a function foruffer strip width.

. Materials and methods

.1. Study area

The field campaign was conducted in 2009 in a set of studytreams that is located on Funen, Denmark (Fig. 1), where catch-ents are characterised by low elevation and loamy soils withedium to low infiltration capacity. Agriculture and forest are the

ominant types of land use. Climatic conditions are temperate andhe average regional precipitation is 700 mm year−1. Dominatingrop types in the studied catchments were rye, wheat, barley, grassnd oilseed rape (Appendix A).

.2. Stream characteristics

Fourteen 1st-or 2nd-order streams were selected based on theollowing selection criteria: year-round water flow, no mainte-ance activities conducted during the sampling period (dredgingnd weed-cutting) and no sources of pollution other than fromgricultural non-point sources. The streams represent a gradientf potential pesticide contamination predicted from the proportionf adjacent agricultural land. In order to optimise the selection oftreams, the pesticide runoff was predicted by applying the runoffotential (RP) model (see also Schriever et al., 2007a,b). The RP-odel is a generic indicator that was developed to quantify the risk

f pesticide runoff contamination to streams from agricultural landSchriever et al., 2007a). Calculated RP for site selection support was
ased on the assumption that any runoff-triggering precipitationvent would be evenly distributed among the studied streams. Datanput for grown crops and pesticide application was based on 2008ata (Danish EPA, 2009).
lqcv

eering 37 (2011) 1990– 1997 1991

Using aerial photographs, buffer strip dimensions (minimumnd average buffer strip width) were determined for each streamy digitalising buffer strips in 500, 1000 and 2000 m sectionspstream of the sampling sites in ArcGis 9.2. Average buffer stripidth was calculated by simple mathematical integration of theigitalised buffer strip area. The outer boundaries of buffer stripsere characteristically visible using summer photos, since buffer

trips are relatively unmaintained compared to conventional agri-ultural fields and fallow land. Consequently, the different typesf vegetation found in the buffer strips clearly defined their outeroundaries.

.3. Quantification of pesticide contamination

The selection of analysed pesticides was based on applicationrequency and total applied amounts in 2008 (Danish EPA, 2009).his list was augmented with a series of banned pesticides thatre commonly found in drinking water wells. In total, 19 herbi-ides, 6 fungicides and 6 insecticides were included in the samplingrogram (Appendix B). The sampling campaign was conducted in009.

We used event-triggered samplers to characterise pesticideontamination during heavy precipitation events (Liess and von derhe, 2005). The sampling system consisted of two 1 L glass bottles

hat were deployed in the flowing part of the stream channel. Bot-les were filled passively through small (0.5 cm in diameter) glassubes when the water level increased above the glass tube open-ng. The two bottles were positioned 5 cm and 10 cm above baseow water level, respectively. Filled water samples were retrievedithin 24 h after each heavy precipitation event. During the sam-ling period, two precipitation episodes triggered the samplingystem. The first episode occurred on the 28th of May and washaracterised by a precipitation depth ranging from 7 mm to 10 mmepending on the site. This episode triggered samplers in only sixtreams. The second episode occurred on the 12th of June and washaracterised by a precipitation depth ranging from 19 to 47 mm.he latter triggered the sampling system in all streams.

Bed sediment was sampled (stratified sampling) on the 20th ofuly using a kajak corer (8 cm diameter). All sediment samples wereollected within a 50 m stream section extending upstream fromhe event triggered samplers. One sample consisted of a minimumf 30 sub-samples from the top layer (1–2 cm) of newly depositedediment at in order to obtain sediment samples that generallyere representative for the respective reaches (see also Friberg

t al., 2003).Water samples were collected manually in August during low

ow conditions in order to characterise the potential ‘backgroundnput’ of pesticides originating from groundwater inflow. Bannedesticides were detected in all streams indicating the importancef groundwater input as a source of pesticides. However, in ourtudy, pesticides in the August samples were characterised by aombination of low concentrations and low toxicity to benthicacroinvertebrates. Consequently, we assumed that pesticides

riginating from groundwater input were of minor importance inhe studied streams.

The pesticide analyses (including solid phase extraction) wereonducted by OMEGAM laboratories in Amsterdam; unfilteredamples were sent to the laboratories in coolers immedi-tely after collection. The final extract of each sample wassed in different analysis programs. Analysis programs wereased on gas-chromatography mass-spectrometry (GC-MS) or

iquid-chromatography mass-spectrometry (LC–MS). The limit ofuantification for each compound was determined as the lowestoncentration that can be reliably quantified (95% confidence inter-al) (Appendix B). Detection limits were 0.01–0.1 �g L−1 for water

1992 J.J. Rasmussen et al. / Ecological Engineering 37 (2011) 1990– 1997

the 14

sRsw

2

trrwttpaTaa

citsotiWu

g

Fig. 1. Schematic map of

amples and 0.01–0.1 mg kg−1 (dry weight) for sediment samples.esults were corrected for recovery, which was determined bypiked samples. For all compounds, recovery was reported to beithin 85–110% of actual concentrations.

.4. Predicted pesticide exposure

The runoff potential model was produced to predict runoff con-amination of a generic compound instead of predicting actualunoff losses for a specific compound. However, due to the highesolution and quality input data (field block-specific crop data) weere able to meet data requirements for a more detailed version of

he model in terms of grown crops (Eq. (1)). Due to the high resolu-ion of crop data, we could additionally improve our estimates foresticide application rates using the average compound-specific
pplication rate for each crop type in 2009 (Danish EPA, 2010).hus, we could calculate the runoff potential for the compoundsssociated with each crop type instead of just predicting runoff for
generic compound. For further details on the original RP model,

wtt

study stream locations.

onsult Schriever et al. (2007a). We calculated RP for all sites apply-ng a two-sided corridor of 100 m extending 500 m upstream ofhe sampling location. Modification of the considered catchmentize, i.e. implementation of other corridor lengths (1000 or 2000 m)r utilisation of the total catchment had no significant effect onhe results. For convenience the two-sided 100 m corridor extend-ng 500 m upstream will be referred to as the stream corridor.

e calculated pesticide runoff by first applying the runoff modelnderlying the RP (modified after Schriever et al. (2007a)):

LOAD =n∑

i=1

m∑j=1

k∑l=1

Ai,j · Dl ·(

1 − Ij100

)· 1

1 + Kocl · OCi/100

·f (si) · f (Pi, Ti)Pi

(1)

here index i refers to the respective field blocks, index j referso different crop types present on the fields, and index l referso specific pesticides. Ai,j is the size of agricultural land (ha), Dl is

J.J. Rasmussen et al. / Ecological Engineering 37 (2011) 1990– 1997 1993

Table 1Pesticides detected in stream water from 14 Danish streams in the period from April to August, 2009. Three samples were collected in each stream of which two werecollected with event-triggered samplers during May and June (high precipitation events), and one sample was collected manually during base-flow conditions in August.Pesticide groups are indicated by letters H, F and I representing herbicides, fungicides and insecticides, respectively.

Compound Min concentration (�g L−1) Max concentration (�g L−1) Highest TUa Detection frequency (%)

Desethylterbutylazine (H) 0.01 0.11 −4.65 100Atrazine (H) 0.01 0.02 −6.63 7Dimethoate (I) 0.01 0.18 −4.05 14Metachlor (H) 0.01 0.05 −5.82 57Diflufenican (H) 0.02 0.15 −3.20 29Metamitron (H) 0.12 0.12 −4.68 7Pendimethaline (H) 0.02 0.97 −2.46 14Aclinofen (H) 0.14 0.14 −3.93 7Propyzamide (H) 0.01 0.43 −4.11 21Prosulfocarb (H) 0.01 0.07 −3.86 21Terbutylazine (H) 0.01 0.6 −4.55 57Hexazinone (H) 0.06 0.06 −6.15 7Simazine (H) 0.03 0.03 −4.56 7DEET (H) 0.05 0.05 −6.18 7Boscalid (F) 0.07 0.72 −3.87 36Azoxystrobin (F) 0.05 0.51 −2.77 43Propiconazole (F) 0.04 0.27 −4.58 43Tebuconazole (F) 0.02 0.24 −4.24 50Dimethomorf (F) 0.01 0.08 −5.12 14

2

, 2001

tgttcfdarv

R

se2D9s2docaiErdpae

2

c

ai

T

wcLmtmanaa

acamupaovwio

mwmoi

3

Pirimicarb (I) 0.01 0.3

a Based on LC50 values for 48 h acute toxicity tests with Daphnia magna (Tomlin

he application rate of the pesticide compound, Ij is the crop- androwth phase-specific plant interception of the substance at theime of the precipitation event (%), Kocl is the organic carbon sorp-ion coefficient of the pesticide compound, OCi is the soil organicarbon content of a field patch (%), si is the mean slope of a field (%),(si) describes the influence of the field slope. Pi is the precipitationepth (mm) of the considered event, Ti refers to the soil texture of

field (sandy/loamy), f(Pi, Ti) is a function describing the surfaceunoff volume for vegetated soils in the middle or late period foregetation growth. RP (Eq. (2)) is then calculated as:

P = log(

nmaxi=1

(gLOADi))

(2)

The runoff potential model was parameterised as follows: field-pecific crop types for each field block in the stream corridor werextracted from a national Danish database (LOOP) (Grant et al.,006). Soil slope in the stream corridor was estimated using aigitalised Elevation Map (DEM) with 1.6 m resolution in ArcGis.2. Soil texture composition (including humus content) within thetream corridor was extracted from the Hair database (Greve et al.,007). According to Thomas and Goudie (2000), sandy soil wasefined as soils containing <10% clay and >85% sand. The relativerganic carbon content of soils was calculated as 57% of the humusontent (Thomas and Goudie, 2000). The average crop-specificpplication rate for each pesticide compound potentially appliedn 2009 was extracted from national pesticide statistics (DanishPA, 2010). Precipitation data were provided by the Danish Meteo-ological Institute (http://www.dmi.dk) (100 km2 resolution). Theaily recorded precipitation was assumed to result from a singlerecipitation event. Plant interception values (Ij) were assigned toll crop types that were present during the considered precipitationvent according to Linders et al. (2000).

.5. Data analysis

We applied toxic units (TU) as a measure for pesticide toxicity,alculating TU for all pesticides detected in each sample. TU values

3

c

−1.72 21

).

re based on the acute 48 h LC50 value for Daphnia magna, as givenn Tomlin (2001) (Eq. (3)).

U(D.magna) = log(

Ci

LC50i

)(3)

here TU(D.magna) is the toxic unit for pesticide i, Ci is the measuredoncentration of pesticide i and LC50i is the corresponding 48 hC50 value for D. magna exposed to pesticide i. We identified theaximum TU for each water sample, and additionally calculated

he summed TU for all pesticides in each water sample. The sum-ation of all TUs is based on the assumption that all compounds

ct under the principle of toxic additivity. As the number of compo-ents in a toxic mixture increases, the range of deviation from toxicdditivity is proposed to decrease (the Funnel hypothesis) (Warnend Hawker, 1995).

All environmental parameters considered (minimum and aver-ge buffer strip width, proportion of agriculture in the streamorridor, crop types, estimated pesticide application, field slopesnd soil texture) were then correlated to the summed TU, maxi-um TU, number of pesticides and sum concentration of pesticides

sing Spearman rank order (r) correlations (P < 0.05). All tests wereerformed using the software SAS enterprise guide 4.2. Leveragend Cook’s Distance were calculated for all fitted regressions inrder to evaluate the contributed weight of each data point. Noalues for Cook’s Distance exceeded 0.1 and no leverage valuesere greater than 2*(p/n), where p is the number of parameters

n the model including the intercept, and n is the total number ofbservations. R2 values are given for all presented regressions.

In addition, we attempted to improve the RP model by imple-enting various functions of minimum and average buffer stripidth in the stream corridor. A fitted regression of the modified RPodel as a function of calculated TUs was compared to that of the

riginal RP model using Analysis of Covariance (ANCOVA) (P < 0.05)n SAS 9.2.

. Results

.1. Pesticides and TU

The results of the field campaign disclosed a total of 13 herbi-ides, 5 fungicides and 2 insecticide that were actually detected in

http://www.dmi.dk/

1994 J.J. Rasmussen et al. / Ecological Engineering 37 (2011) 1990– 1997

F he tott ed on1

wcpfIrbTchp

eT

aAcFpFsmnw

ig. 2. The summed TU of all pesticides (A and D), the maximum TU (B and E) and the proportion of agricultural land (D, E and F, respectively). Presented data are bas4 Danish streams in spring, 2009.

ater samples from the 14 study streams (Table 1). Summed con-entrations ranged from 0.01 to 3.17 �g L−1, the number of detectedesticides per sample ranged from 1 to 13, maximum TU rangedrom −6.63 to −1.72, and summed TU ranged from −6.63 to −1.57.n total, five of the nine streams at risk for receiving pesticideunoff (proportion of agricultural land ≥50%) were characterisedy at least one sample with summed and maximum TUs ≥ −3.he carbamate insecticide Pirimicarb and the Strubilurine fungi-ide Azoxystrobin were the pesticides primarily responsible for theigh TU values due to corresponding low LC50(D. magna) values. No
esticides were detected in the sediment samples.
Minimum buffer strip width was the environmental param-ter most strongly correlated with summed TU and maximumU (r = 0.80, P < 0.0001, Fig. 2a), followed by the proportion of

(iwc

al number of pesticides (C and F) as a function for minimum buffer strip width and water samples collected during storm flow conditions (two storm flow events) in

gricultural land in the stream corridor (r = 0.48, P < 0.05, Fig. 2d).pplying the maximum TU generated a comparable significantorrelation with minimum buffer strip width (r = 0.78, P < 0.0001,ig. 2b) and a slightly stronger significant correlation with theroportion of agriculture in the stream corridor (r = 0.66, P < 0.01,ig. 2e). Applying the average buffer strip width generated aignificant but weaker correlation with summed TU and maxi-um TU (r = 0.61, P < 0.01 and r = 0.65, P < 0.01, respectively) (data

ot shown). Furthermore, the number of pesticide compoundsas significantly correlated to the minimum buffer strip width

r = 0.72, P < 0.001, Fig. 2c) and the proportion of agricultural landn the stream corridor (r = 0.49, P < 0.05, Fig. 2f). Autocorrelations

ere found between the summed TU and the number of pesti-ides (r = 0.82, P < 0.0001), as well as total pesticide concentration

J.J. Rasmussen et al. / Ecological Engineering 37 (2011) 1990– 1997 1995

F(

(c(tbrwoo

3

wTtcrts(apiSt

4

4i

cpwdirtLwsgs

Fig. 4. The RP (A) and a modified version of RP (additionally considering minimumbuffer strip width) (B) as a function for summed TU. The RP and modified RP werebased on a series of environmental parameters deriving from a 2 m × 100 m streamcti

ur(tstcssTaa

4s

sr

ig. 3. Proportion of agricultural land as a function of minimum (�) and average©) buffer strip width. Data represent 14 Danish low-order streams.

r = 0.71, P < 0.001) (data not shown). Furthermore, total pesticideoncentration was autocorrelated with the number of pesticidesr = 0.90, P < 0.0001) (data not shown). The proportion of agricul-ural land was significantly correlated to minimum and averageuffer strip width in the stream corridor (r = 0.66, P < 0.01 and

= 0.73, P < 0.001, respectively), as shown in Fig. 3. No correlationas found between estimated compound-specific applied amounts

f pesticides in the stream corridor and in-stream concentrationsf the respective compounds.

.2. Predicted pesticide exposure

The runoff potential model (RP) was significantly correlatedith the summed TU (r = 0.70, P < 0.001, Fig. 4a) and the maximum

U (r = 0.63, P < 0.01) (data not shown). Adding the inverse func-ion for minimum buffer strip width (within a 2 × 100 m streamorridor extending 500 m upstream from a sampling point) to theunoff model (underlying RP) by simple multiplication improvedhe significance of the correlation found between the RP and theummed TU (r = 0.83, P < 0.0001, Fig. 4b) and the maximum TUr = 0.70, P < 0.001) (data not shown), reflected by reduced data vari-bility around the fitted regression. In other words, the explanatoryower of the model increased from 46% to 64% by adding the

nverse function for minimum buffer strip width to the RP model.lope and intercept were not significantly different between thewo regression lines (P < 0.05).

. Discussion

.1. The influence of buffer strips on the occurrence of pesticidesn streams

Minimum buffer strip width within a two-sided 100 m streamorridor extending 500 m upstream from the pesticide samplingoint was the environmental parameter most strongly correlatedith summed and maximum TUs for pesticides in stream wateruring storm flow. Decreasing summed and maximum TUs with

ncreasing minimum buffer strip width probably reflects runoffeduction, due especially to infiltration and pesticide adsorptiono organic matter within the buffer strip (Anbumozhi et al., 2005;acas et al., 2005; Vidon et al., 2010). Minimum buffer strip width
as autocorrelated with the proportion of agricultural land in the
tream corridor and hence buffer strip width may act as a surro-ate for the proportion of agricultural land. However, numerousite-specific studies document clear effects of buffer strips as a

aaRs

orridor extending 500 m upstream from the sampling points. Pesticide concentra-ions were measured during two storm flow events in 14 Danish low-land streamsn spring, 2009.

seful tool for reducing pesticide transport from fields to streamecipients. For example, both Lacas et al. (2005) and Schriever et al.2007a) found that precipitation intensity and local field charac-eristics (field slopes and crop types/growth phases) were moreensitive parameters than the proportion of agricultural land inhe sub-catchments when predicting pesticide runoff. The strongorrelations between minimum buffer strip width and TU mea-ures (and pesticide concentrations) that were observed in thistudy, additionally suggest that the site properties only affectedU measures marginally. This probably reflects comparable site-nd climatic- and agricultural (e.g. crop types and growth phasest the time of the storm events) properties in the region.

.2. Improving pesticide runoff predictability by adding buffertrip information

Applying high-resolution data, the runoff potential (RP) modeluccessfully predicted the toxicity of agricultural pesticides occur-ing in stream water during storm events. We found, however, that
dding a function for the minimum buffer strip width – within
two-sided 100 m corridor extending 500 m upstream – to theP model markedly improved the power of the model to predictummed TUs from 46% to 64% by reducing the data variability

1 Engin

asmtcosfwrfittwrs

fiHtcse2iid

4

wTNslgptsssport

hvcSidAn-alRhe

rsc

≥b(nicoFafits(Imt

4s

mYAb(TiewafeTavdrbescweetttoefobcLiowsb(

996 J.J. Rasmussen et al. / Ecological

round the regression line. The slope and intercept of the regres-ion line did not significantly change by adding the function forinimum buffer strip width to the RP model, which reflects that

he overall correlation between the RP and summed TUs remainsonstant with or without buffer strip information. However,ur results clearly emphasise that minimum buffer strip widthhould be added to the model whenever data are available, andurthermore underline the importance of considering buffer stripidth in upstream environments of stream sites potentially at

isk of being impacted by agricultural pesticides. Moreover, thesendings lend support to Rasmussen et al. (2011) who were unableo confirm the correlation between the RP and SPEARpesticideshat was found by Schriever et al. (2007b) in German streamsithout buffer strips. Rasmussen et al. (2011) suggested that their

esults were probably confounded by the presence of buffer stripsurrounding the study streams.

No data were available in terms of tile drainage intensity for theelds surrounding the streams that were examined in this study.owever, loamy and clayey agricultural soils are often intensively

ile drained, and the tile drains serve as a direct route for pesti-ides from field to surface waters underneath the buffer strip. Suchites have been found to be extremely vulnerable to pesticide loss,specially if macropores have developed in the soil (Kronvang et al.,004; Lewan et al., 2009; Renaud and Brown, 2008). We therefore

nfer that incorporating information about tile drainage conditionsn the considered (sub-) catchment would further improve the pre-ictive power of the RP model.

.3. Pesticide characteristics and their potential ecological impact

In this study, the summed toxic units (TU) based on storm flowater samples ranged from −6.63 to −1.57. Applying the maximum

U for single pesticides did not significantly change this spectrum.o pesticides were detected in any of the stream bed sediment

amples taken in this study, which could reflect too high detectionimits and/or an inappropriate sampling technique. More strate-ic sampling using a stationary suspended sediment sampler isroposed to further optimise the detection success of adsorbed pes-icides (Liess et al., 1996). However, Friberg et al. (2003) detectedeveral lipophilic pesticides adsorbed to bed sediments in Danishtreams applying a technique similar to the one used in the presenttudy. An additional factor that potentially explains the absence ofesticides in newly deposited bed sediments was the occurrencef several heavy precipitation events during July, which could haveeduced the residence time for the pesticides that were adsorbedo fine particulate organic matter.

Nevertheless, the range of TUs measured in this study doesave the potential to impair stream ecosystems. Benthic macroin-ertebrates have been shown to respond strongly to pesticideontamination (Norum et al., 2010; Rasmussen et al., 2008;chäfer et al., 2007), and they have successfully been applied asndicator organisms for pesticide contamination in the recentlyeveloped SPEARpesticides index (Liess and von der Ohe, 2005).pplying the SPEARpesticides index, macroinvertebrate commu-ity changes have been observed at maximum TUs down to3 in field studies (Schäfer et al., 2011b). The recommendednd currently applied threshold value characterising good eco-ogical status in the online SPEAR calculator (33% SPEcies Atisk) corresponds to a maximum TU value of −3 (see alsottp://www.systemecology.eu/SPEAR/calculator/index.php?lang=n).

We found that the maximum TU and summed TUs concur-ently exceeded the threshold value for ecosystem effects in fivetreams representing more than 50% of the streams at risk of beingontaminated by agricultural pesticides (proportion of agriculture

5

w

eering 37 (2011) 1990– 1997

50% in the stream corridor). Other anthropogenic stressors maye of higher importance than non-point pesticide contaminationRasmussen et al., 2011), but our results clearly emphasise thaton-point source pesticide contamination is a potential problem

n small Danish streams. Not surprisingly, the insecticide Pirimi-arb represented the primary risk for benthic fauna due to its modef action, which acts selectively against this group of organisms.ungicides having a less specific mode of action were, addition-lly, relevant stressors for the benthic macroinvertebrates. Ourndings are congruently supported by a large body of evidencehat identifies insecticides and fungicides as the primary pesticidetressors directly impacting benthic macroinvertebrates in streamssee e.g. Liess et al., 2005; Schäfer et al., 2007, 2011a; Schulz, 2004).n addition, we found that the herbicide, Pendicmethalin (inhibits

itosis), might also act as a potentially important stressor for ben-hic macroinvertebrates.

.4. Implications for stream management and the protection oftream ecosystems

The regression line in Fig. 2b represents the maxi-um TU as a function for minimum buffer strip width;

= −6.586(±0.681) + 6.235(±1.24) × exp(−0.249(±0.105)x).ssuming that the relationship is causative, the minimumuffer strip width necessary for obtaining good ecological statusmaximum TU ≤ −3), as required by the European WFD, is 6.6 m.his is strongly contrasted by present legislative requirementsn Denmark where only natural streams or streams with a highcological objective (approximately 40% of the total stream net-ork) are required to have 2 m of uncultivated buffer strips. The

im of buffer strips in Denmark is only to protect stream banksrom erosion, and pesticide application restrictions are currentlynforced only via application guidelines for specific compounds.he vast majority of Danish streams are therefore still unprotectedgainst pesticide contamination. However, considering the largeariability in data around the fitted regression and the precedingifficulties in predicting optimal dimensions for the buffer stripetaining capacity, we recommend that the suggested minimumuffer strip width is considered with care. Furthermore, Schäfert al. (2007) detected very high maximum TUs in small Frenchtreams that were flanked by buffer strips exceeding 11 m. Thisould indicate that the correlation between minimum buffer stripidth and the TU obtained in this study is not applicable for gen-

ral extrapolation in time or space. However, the results of Schäfert al. (2007) could be confounded by intensive tile-drainage, asile drains introduce an important transport route underneathhe vegetated buffer strips. Only a few authors have attemptedo describe the dimensions that buffer strips should have forptimum performance in terms of pesticide retention (Johnsont al., 2007), probably reflecting the numerous highly variableactors influencing pesticide runoff, including timing and volumef rainfall events occurring subsequent to pesticide application,uffer strip vegetation types and growth phases, soil infiltrationapacity, soil moisture and runoff velocity (Klöppel et al., 1997;acas et al., 2005; Pot et al., 2005). Depending on the site character-stics, climatic conditions and local pesticide application practices,ptimal buffer strip width change. As a consequence, buffer stripsider than 6.6 m could be necessary for sufficient protection of

tream ecosystems from pesticide surface runoff, as it has alsoeen found for different phosphorus forms and other pollutantsHoffmann et al., 2009; Mander, 2005; Uusi-Kämppä, 2005).

. Conclusions

The minimum width of buffer strips in the near upstream areaas found to be the most important environmental parameter

http://www.systemecology.eu/SPEAR/calculator/index.php%3Flang=en

Engin

gsbfsm(pcrse

PUstadbaptmm

A

naHatm

A

i

R

A

D

D

E

F

F

G

G

H

J

K

K

K

K

L

L

L

L

L

L

M

N

N

P

P

R

R

R

S

S

S

S

S

S

T

T

U

V

J.J. Rasmussen et al. / Ecological

overning measured summed and maximum TUs in Danishtreams. This suggests that the prevalence and dimensions foruffer strips currently required by Danish legislation is, in general,ar from sufficient in protecting stream ecosystems from non-pointource pesticides. Despite the fact that small streams with catch-ent sizes under 10 km2 are disregarded within the European WFD

European Commission, 2000), we believe it is still essential torotect the upper branches of streams with buffer strips espe-ially since these systems serve as sources for recolonisation to theeaches further downstream (targeted in the WFD). Providing suchources would add some valuable recovery capacity to the streamcosystems.

Adding a function for minimum buffer strip width to the Runoffotential (RP) model improved its power to predict summed Toxicnits in the study streams from 46% to 64% without changing the

lope or intercept of the regression line. This underlines the impor-ance of considering buffer strip dimensions in the near upstreamrea within the risk assessment procedure. Using high-resolutionata (including buffer strip dimensions) the RP model was found toe a useful screening tool for the identification of stream sectionst risk for pesticide contamination. However, we suggest thatesticide transport from agricultural catchments to streams viaile drain flow would further improve the predictive power of the

odel. Future research should address these shortcomings of theodel.

cknowledgements

This study was financed by the Danish Research Council (granto. 2104-07-0035) and is part of the RISKPOINT program. Theuthors greatly acknowledge Rikke Juul Monberg, Uffe Mensberg,enrik Stenholt and Marlene Venø Skjærbæk for their valu-ble support during the field campaign. Furthermore, we thankwo anonymous reviewers for constructive comments on the

anuscript.

ppendix A. Supplementary data

Supplementary data associated with this article can be found,n the online version, at doi:10.1016/j.ecoleng.2011.08.016.

eferences

nbumozhi, V., Radhakrishnan, J., Yamaji, E., 2005. Impact of buffer zones on waterquality. Ecol. Eng. 24, 517–523.

anish Environmental Protection Agency, 2008. Pesticide Statistics 2009. Orienter-ing fra Miljøstyrelsen nr. 5, Copenhagen (in Danish).

anish Environmental Protection Agency, 2009. Pesticide statistics 2010. Orienter-ing fra Miljøstyrelsen nr. 6, Copenhagen (in Danish).

uropean Commission, 22.12.2000. Directive 2000/60/EC of the European Parlia-ment and of the Council of 23 October 2000 Establishing a Framework forCommunity Action in the Field of Water Policy. Off. J. Eur. Commun. L327, p. 77.

riberg, N., Lindstrom, M., Kronvang, B., Larsen, S.E., 2003. Macroinverte-brate/sediment relationships along a pesticide gradient in Danish streams.Hydrobiologia 494, 103–110.

urse, M., Hering, D., Moog, O., Verdonschot, P., Johnson, R.K., Brabec, K., Gritzalis, K.,Buffagni, A., Pinto, P., Friberg, N., Murray-Bligh, J., Kokes, J., Alber, R., Usseglio-Polatera, P., Haase, P., Sweeting, R., Bis, B., Szoszkiewicz, K., Soszka, H., Springe, G.,Sporka, F., Krno, I., 2006. The STAR project: context, objectives and approaches.Hydrobiologia 566, 3–29.

rant, R., Blicher-Mathiesen, G., Pedersen, L.E., Jensen, P.G., Madsen, I., Hansen, B.,Brüsch, W., Thorling, L., 2006. Survailance catchments for the National Moni-toring Program. Faglig rapport nr. 640. Danmarks Miljøundersøgelser, AarhusUniversitet, Denmark, p. 121, (In Danish).

reve, M.H., Greve, M.B., Bøcher, P.K., Balstrøm, T., Madsen, H.B., Krogh, L., 2007.Generating a Danish raster-based topsoil property map combining choroplethmaps and point information. Geografisk Tidsskrift 107 (2).

offmann, C.C., Kjaergaard, C., Uusi-Kämppä, J., Hansen, H.C.B., Kronvang, B., 2009.
Phosphorus retention in riparian buffers: review of their efficiency. J. Environ.Qual. 38, 1942–1955.
ohnson, T.E., McNair, J.N., Srivastave, P., Hart, D.D., 2007. Stream ecosystemresponses to spatially variable land cover: an empirically based model for devel-oping riparian restoration strategies. Freshw. Biol. 52, 680–695.

W

W

eering 37 (2011) 1990– 1997 1997

löppel, H., Kördel, W., Stein, B., 1997. Herbicide transport by surface runoff andherbicide retention in a filter strip: rainfall and runoff simulation studies.Chemosphere 35, 129–141.

reuger, J., 1998. Pesticides in stream water within an agricultural catchment insouthern Sweden, 1990–1996. Sci. Total Environ. 216, 227–251.

reuger, J., Brink, N., 1988. Losses of pesticides from agriculture. In: Pesticides:Food and Environmental Implications. Int. Atomic Energy Agency, Vienna, pp.101–112.

ronvang, B., Strøm, H.L., Hoffmann, C.C., Laubel, A., Friberg, N., 2004. Subsurface tiledrainage loss of modern pesticides: field experiment results. Water Sci. Technol.49, 139–148.

acas, J.G., Voltz, M., Gouy, V., Carluer, N., Gril, J.J., 2005. Using grassed buffer stripsto limit pesticide transfer to surface water: a review. Agron. Sustain. Dev. 25,253–266.

ewan, E., Kreuger, J., Jarvis, N., 2009. Implications of precipitation patterns andantecedent soil water content for leaching of pesticides from arable land. Agric.Water Manage. 96, 1633–1640.

iess, M., Schulz, R., Neumann, M., 1996. A method for monitoring pesticides boundto suspended particles in small streams. Chemosphere 32, 1963–1969.

iess, M., von der Ohe, C., 2005. Analyzing effects of pesticides on invertebratecommunities in streams. Environ. Toxicol. Chem. 24, 954–965.

iess, M., Brown, C., Dohmen, P., Duquesne, S., Hart, A., Heimbach, F., Kreuger, J.,Lagadic, L., Maund, S., Reinert, W., Streloke, M., Tarazona, J.V., 2005. Effects ofpesticides in the field – EPIF. Setac Press, Brussels, pp. 136.

inders, J., Mensink, H., Stephenson, G., Wauchope, D., Rache, K., 2000. Foliarinterception and retention values after pesticide application: a proposal forstandardised values for environmental risk assessment. J. Appl. Chem. 72,2199–2218.

ander, U., 2005. Purification processes, ecological functions, planning and designof riparian buffer zones in agricultural watersheds. Ecol. Eng. 24, 421–432.

eumann, M., Schulz, R., Schäfer, K., Müller, W., Mannheller, W., Liess, M., 2002. Thesignificance of entry routes as point and non-point sources of pesticides in smallstreams. Water Res. 36, 835–842.

orum, U., Friberg, N., Jensen, M.R., Pedersen, J.M., Bjerregaard, P., 2010. Behav-ioral changes in three species of freshwater macroinvertebrates exposed tothe pyrethroid lambda-cyhalothrin: laboratory and stream microcosm studies.Aquat. Toxicol. 98, 328–335.

ot, V., Simunek, J., Benoit, P., Coquet, Y., Yra, A., Martinez-Cordon, M.J., 2005. Impactof rainfall intensity on the transport of two herbicides in undisturbed grassedfilter strip soil cores. J. Contam. Hydrol. 81, 63–88.

robst, M., Berenzen, N., Lentzen-Godding, A., Schulz, R., Liess, M., 2005. Linkingland use variables and invertebrate taxon richness in small and medium-sizedagricultural streams on a landscape level. Ecotoxicol. Environ. Saf. 60, 140–146.

asmussen, J.J., Friberg, N., Larsen, S.E., 2008. Impact of Lambda-cyhalothrin on amacroinvertebrate assemblage in outdoor experimental channels: implicationsfor ecoystem functioning. Aquat. Toxicol. 90, 228–234.

asmussen, J.J., Baattrup-Pedersen, A., Larsen, S.E., Kronvang, B., 2011. Local physicalhabitat quality clouds the effect of predicted pesticide runoff from agriculturalland in Danish streams. J. Environ. Monit. 13, 943–950.

enaud, F.G., Brown, C.D., 2008. Simulating pesticides in ditches to assess ecologicalrisk (SPIDER): II. Benchmarking for the drainage model. Sci. Total Environ. 394,124–133.

chäfer, R.B., Caquet, T., Siimes, K., Mueller, R., Lagadic, L., Liess, M., 2007. Effectsof pesticides on community structure and ecosystem functions in agriculturalstreams of three biogeographical regions in Europe. Sci. Total Environ. 382,272–285.

chäfer, R.B., Pettigrove, V., Rose, G., Allinson, G., Withtwick, A., von der Ohe, P.C.,Shimeta, J., Kühne, R., Kefford, B.J., 2011a. Effects of pesticides monitored withthree sampling methods in 24 sites on macroinvertebrates and microorganisms.Environ. Sci. Technol. 45, 1665–1672.

chäfer, R.B., van den Brink, P.J., Liess, M., 2011b. Impacts of pesticides on freshwaterecosystems. In: Sanchez-Bayo, F., van den Brink, P.J., Mann, R.M. (Eds.), EcologicalImpacts of Toxic Chemicals. Bentham:Bussum, The Netherlands.

chulz, R., 2004. Field studies on exposure, effects, and risk mitigation ofaquatic nonpoint-source insecticide pollution: a review. J. Environ. Qual. 33,419–448.

chriever, C.A., von der Ohe, C., Liess, M., 2007a. Estimating pesticide runoff in smallstreams. Chemosphere 68, 2161–2171.

chriever, C.A., Ball, M.H., Holmes, C., Maund, S., Liess, M., 2007b. Agricultural inten-sity and landscape structure: influence on the macroinvertebrate assemblagesof small streams in Germany. Environ. Toxicol. Chem. 26, 346–357.

homas, D.S.G., Goudie, A., 2000. The Dictionary of Physical Geography, 3rd edn.Blackwell Publishing Ltd.

omlin, C.D.S., 2001. The Pesticide Manual, A World Compendium. Crop ProtectionPublications, Farnham, Surrey, UK.

usi-Kämppä, J., 2005. Phosphorus purification in buffer zones in cold climates. Ecol.Eng. 24, 491–502.

idon, P., Allan, C., Burns, D., Duval, T.P., Gurwick, N., Inamdar, S., Lowrance, R., Scott,D., Sebestyen, S., 2010. Hot spots and hot moments in riparian zones: potentialfor improved water quality management. J. Am. Water Res. Assoc. 46, 278–298.

arne, M.S.J., Hawker, D.W., 1995. The number of components in a mixture deter-mines whether synergistic and antagonistic or additive toxicity predominate:the Funnel hypothesis. Ecotoxicol. Environ. Saf. 31, 23–28.

auchope, R.D., 1978. The pesticide content of surface water draining from agri-cultural fields – a review. J. Environ. Qual. 7, 459–472.

http://dx.doi.org/10.1016/j.ecoleng.2011.08.016

Paper II

Dynamic Article LinksC<Journal ofEnvironmentalMonitoringCite this: J. Environ. Monit., 2011, 13, 943

www.rsc.org/jem PAPER

Dow

nloa

ded

by S

tats

bibl

iote

ket o

n 14

Feb

ruar

y 20

12Pu

blis

hed

on 0

9 M

arch

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0EM

0074

5E

View Online

Local physical habitat quality cloud the effect of predicted pesticide runofffrom agricultural land in Danish streams

Jes Jessen Rasmussen,* Annette Baattrup-Pedersen, Søren Erik Larsen and Brian Kronvang

Received 16th December 2010, Accepted 9th February 2011

DOI: 10.1039/c0em00745e

The combination of intensive agricultural activities and the close connectivity between land and stream

emphasise the potential risk of pesticide exposure in Danish streams. Benthic macroinvertebrates are

applied in the assessment of stream ecological status, and some sensitive species have been shown to

respond strongly to brief pulses of pesticide contamination. In this study we investigate the impact of

agriculturally derived pesticides on stream macroinvertebrate communities in Denmark. As a measure

of toxic pressure we apply the Runoff Potential. We investigated a total of 212 streams. These were

grouped into distinct classes according to the magnitude of pesticide contamination in the period from

2003–2006. A total of 24 different macroinvertebrate indices were applied to detect effects of pesticide

runoff (e.g. the SPEAR-index and the number of EPT taxa). We found high predicted pesticide runoff

in 39% of the streams, but we found no significant effect of predicted pesticide exposure on stream

macroinvertebrate indices. We, additionally, examined the influence of a series of environmental

parameters ranging from site scale to catchment scale on the macroinvertebrate community. Relative

proportions of gravel, sand and silt in bed sediments explained most of the variation in

macroinvertebrate indices as well as the upstream riparian habitat quality. We suggest that the Runoff

Potential model overestimate pesticide runoff contamination in Danish streams due the presence of

buffer strips enforced by Danish legislation. When pesticide runoff contamination is low to moderate,

poor physical properties (indirectly related to agricultural activity) are the main impediment for the

ecological quality of Danish streams.

1. Introduction

The brief occurrence of pesticides in streams resulting from

surface runoff,1,2 tile drains3,4 and spray drift5 from agricultural

fields has been documented in several studies. Furthermore,

surface runoff and tile drains have often been emphasised as

important entry routes for pesticides to streams having a poten-

tial impact on benthic macroinvertebrates.6,7 However, several

macroinvertebrate species declining in abundance during pesti-

cide exposure have been shown to recover within a year.6 The

National Environmental Research Institute (NERI), Aarhus University,Vejlsøvej 25, 8600 Silkeborg, Denmark. E-mail: [email protected]

Environmental impact

Understanding how pesticides impact stream ecosystems is importa

provide the best information about system dynamics during and afte

time consuming, expensive and very difficult. Thus there is a scarc

streams and use macroinvertebrate communities as a measure of s

predicted pesticide exposure. We applied the novel approach o

contamination consistently recurring every year within a four year

This journal is ª The Royal Society of Chemistry 2011

presence of forest on both sides of the stream corridor in the near

upstream area has, furthermore, been shown to promote the

process of recolonisation by adding some recovering capacity to

the system.8,9 Pesticide concentrations though rarely reach the

level of acute mortality for macroinvertebrates emphasising the

potential importance of considering sublethal effects such as

feeding behaviour,10 emergence success and offspring produc-

tion.11,12 Sublethal effects potentially change inter-specific inter-

actions which in time could translate into the displacement of

sensitive species. Considering long-term effects is, consequently,

important when evaluating effects of pesticides on macro-

invertebrates. Recently the SPEAR index was introduced by

Liess and von der Ohe8 aiming to tackle the challenge of

nt in order to conduct proper risk assessment. Actual field data

r pesticide exposure. Collecting sufficient field data is, however,

ity of large scale field studies. In this study we investigate 212

tream ecosystem health. These parameters are correlated with

f grouping streams according to predicted pesticide runoff

period. We hereby expect to enhance the strength of the signal.

J. Environ. Monit., 2011, 13, 943–950 | 943

http://dx.doi.org/10.1039/c0em00745e






Fig. 1 Geographical distribution of sampling sites (n ¼ 212).

Dow

nloa

ded

by S

tats

bibl

iote

ket o

n 14

Feb

ruar

y 20

12Pu

blis

hed

on 0

9 M

arch

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0EM

0074

5E

View Online

detecting effects of pesticides on macroinvertebrate community

structure in the field. The SPEAR index classifies species as at

risk (SPEAR) or not at risk (SPEnotAR) considering both

species sensitivity to pesticides and their ecological traits and

thus provides a promising tool for evaluating long-term effects of

pesticides on macroinvertebrates in streams.

In Denmark 62% of the total area (43 000 km2) is used for

agriculture. The total length of the stream network is 64 000 km,

the majority of which are small (<2 metres wide). The combi-

nation of intensive agricultural activities and the close connec-

tivity between land and stream enhanced artificially by drainage

and narrow buffer strips (2 metres) emphasise the potential risk

of pesticide exposure. Intensive agricultural activities, however,

have other additional consequences including comprehensive

habitat degradation (facilitated by channelisation, dredging

activities and siltation) and eutrophication.13,14 90% of all Danish

streams are to some extent physically modified which, conse-

quently, translates into low habitat stability and, in more extreme

cases, high mud cover. Both habitat degradation and eutrophi-

cation have been shown to significantly impact sensitive macro-

invertebrate species and the community structure in general.14–16

When conducting field studies it is therefore a major challenge to

disentangle the effects of one environmental stressor from those

of other influencing environmental stressors.

In this study we aim to detect the potential impact of agri-

culturally derived pesticide runoff on stream macroinvertebrate

communities in Denmark. As a measure for toxic pressure from

pesticide runoff we applied the RP model (Runoff Potential

model) which has been validated in German lowland streams

both in terms of predictive power for actual pesticide concen-

trations17 and its close correlation to the SPEAR index.9 We

grouped the streams according to the predicted pesticide

contamination consistently recurring every year in the period

2003–2006, which we expect to enhance the strength of the signal.

2. Methods

2.1 Study area

Sampling sites were evenly distributed across Denmark (Fig. 1).

Denmark is generally characterised by low elevation averaging

31 metres above sea level. The western part of Denmark is

characterised by sandy soils with high infiltration capacity

whereas the eastern part is characterised by moraine lands with

low infiltration capacity. Climatic conditions are temperate and

average annual precipitation varies from 500 mm year�1 in the

north eastern part of Zealand to 800 mm year�1 in the central

parts of Jutland. 62% of the total area (43 000 km2) is used for

agricultural activity. Dominating crop types are cereals

(70%) followed by grasslands (9%), oilseed rape (5%) and maize

(5%).18–21

2.2 Stream characteristics and macroinvertebrates

In total 212 Danish stream sites were included in this study. The

distance to source was minimum 1500 metres for all sites. At all

sites fauna was sampled using standardised kick-sampling22 in

March or April in each of the years 2004–2007. All macro-

invertebrates were identified to species level except individuals

belonging to Oligochaeta, Tipuloidea, Ceratopogonidae and

944 | J. Environ. Monit., 2011, 13, 943–950

Chironomidae. Applying the publicly available SPEAR data-

base23 %SPEAR abundance was calculated as the proportion of

species characterised as at risk (SPEAR). DSFI score was

calculated after Pedersen et al.22 Applying Asterisks 3.1.1 we,

furthermore, calculated German Saprobic index value, life index

score, total abundance, species richness, separate and summed

species numbers and abundance of Ephemeroptera, Plecoptera

and Trichoptera, Shannon–Wiener diversity, Simpson diversity,

Evenness and r/K relationship. Parameter ranges of selected

community descriptive parameters are presented in Table 1.

Site specific physical and chemical parameters in terms of

qualitative and quantitative substrate assessments, water chem-

istry and other physical variation (including species specific

coverage of emergent and submergent macrophytes, stream

width and distance to source) were extracted from the ODA

database (https://oda.dk) (Table 1). Substrate composition and

macrophyte species specific coverage were surveyed on a repre-

sentative 100 metre reach according to Pedersen et al.22 Based on

the detailed substrate data we calculated substrate heterogeneity

(SH) after Baattrup-Pedersen and Riis.24

Land cover types (relative proportion of forest, agriculture,

permanent grasslands, undisturbed dry areas (e.g. heath),

undisturbed wet areas (e.g. meadows and marsh) and surface

water bodies) were scored at three different scales: (i) the total

watershed of the stream reach; (ii) a two-sided 50 m buffer along

all upstream reaches in the watershed; and (iii) an approximately

100 m � 100 m area located immediately upstream of the

sampling site. Land cover types at the different spatial scales

were determined using the Analysis Tools in ESRI ArcGIS 9.2.

2.3 Predicted pesticide exposure

We calculated predicted pesticide exposure applying the Runoff

Potential (RP) model. The RP is a generic indicator which was



Table 1 Average, std. error, minimum and maximum values for physical and chemical properties for sampling sites, catchment characteristics andselected site specific macroinvertebrate community descriptive parameters. The number of sites for which data were available is indicated

Parameter # Sites Average Std. Error Minimum Maximum

PhysicalCatchment area/km�2 212 112.7 1.9 0.4 1223.1Distance to source/m�1 212 9188 651.6 1500 208 000Cobble (%) 164 5.7 0.4 0 69Gravel (%) 164 19.2 0.8 0 92Sand (%) 164 41.1 0.9 0 85Silt (%) 164 26.2 1 0 100Substrate heterogeneity 164 0.37 0.007 0 0.68Macrophyte coveragea (%) 160 65.6 1.1 0.5 100# Macrophyte speciesa 160 12.4 0.3 1 50ChemicalOrtho-phosphateb/mg L�1 212 0.05 0.004 0 1.67Nitrateb/mg L�1 212 3.73 0.08 0 14Ammoniab/mg L�1 212 0.12 0.005 0 1.81BODb/mg L�1 212 1.35 0.04 0 9.94pHb 212 7.6 0.03 6.7 8.8Catchment characteristicsAgriculture in the near upstream

areac,d (%)212 62.9 4.1 0 100

Undisturbed area in the nearupstream areac,d (%)

212 23.3 1.1 0 100

Undisturbed area in catchmentc

(%)212 27.3 0.9 0 88.9

Macroinvertebrate community descriptors# Macroinvertebrate taxae 212 36.4 0.4 15 77%SPEARe 212 30.0 0.3 5.0 52.9DSFIe 212 5.0 0.04 2 7German saprobic indexe 212 2.1 0.008 1.5 2.9# EPT taxae 212 10.5 0.2 0 28EPT abundancee (%) 212 22.7 0.5 0 61.5

a Measured in months June–September. b Measured 1–12 times per year. c Analysis based on digital spatial data. d Represents a 100 m � 100 m areaextending upstream from the sampling site. e Fauna samples collected in March–April in each of the years 2004–2007.

Dow

nloa

ded

by S

tats

bibl

iote

ket o

n 14

Feb

ruar

y 20

12Pu

blis

hed

on 0

9 M

arch

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0EM

0074

5E

View Online

developed to quantify risk of pesticide runoff contamination to

streams from agricultural land.9 The runoff model (eqn (1))

underlying the RP (eqn (2)) predicts runoff contamination of

a generic compound instead of predicting actual runoff losses for

a specific compound. Due to this simplistic model approach

outcome uncertainty is, consequently, increased. However, due

to limited data availability, we could not meet data requirements

for more detailed models which consider e.g. detailed compound

application patterns or buffer strip width or quality. For further

details on the RP model consult Schriever et al.9 We calculated

RP for all sites for each of the years 2003–2006 applying the

100 m � 100 m area. For convenience this is referred to as the

RP-square.

gLOAD¼Xn

i¼1

Xm

j¼1

Ai; jDgeneric

�1� Ij

100

�1

1þ KocgenericOCi

100

f ðsiÞf�Pi;T

�i

Pi

(1)

where Ai,j is the size of agricultural land (ha), index i refers to the

respective fields, index j refers to different crop types present on

the fields. Dgeneric is the application rate of the generic substance,

Ij is the crop- and growth phase specific plant interception of the

substance at the time of the precipitation event (%), Kocgenericis the

organic carbon sorption coefficient of the generic compound,

OCi is the soil organic carbon content of a field patch (%), si is the

mean slope of a field (%), f(si) describes the influence of the field

slope. Pi is precipitation depth (mm) of the considered event, Ti


refers to soil texture of the field (sandy/loamy), f(Pi, Ti) is

a function describing the surface runoff volume for vegetated

soils in the middle or late vegetation period. RP is calculated as:

RP ¼ log maxn

i¼1ðgLOADiÞ

� �(2)

We parameterised the runoff model as follows. The major

crops grown in Denmark were assumed to be uniformly

distributed in the country, and agricultural land within the RP-

square was assigned values for national relative crop distribution

(Danish Environmental Protection Agency, 2003–2006). The soil

slope in the RP-square was estimated using DEM (Digitalised

Elevation Map) with 1.6 metre resolution in ArcGis 9.2. The soil

texture composition (including humus content) within the RP-

square was extracted from the Hair database.25 According to

Thomas and Goudie26 sandy soil was defined as soils containing

<10% clay and >85% sand. The relative organic carbon content

of soils was calculated as 57% of the humus content.26 We did not

have detailed information concerning the application rate of the

generic compound. Consequently, it was set to a constant value

of 1 g ha�1 for all crop types. Daily precipitation data were

provided by the Danish Meteorological Institute (http://

www.dmi.dk) (10 km2 resolution). Daily recorded precipitation

depth was assumed to result from a single precipitation event.

Only precipitation events occurring in May or June were

considered as these months represent the main pesticide appli-

cation season. Precipitation depth values <10 mm were assumed

J. Environ. Monit., 2011, 13, 943–950 | 945


Fig. 2 Relationship between predicted pesticide runoff (RP) and frac-

tion of species at risk (%SPEAR abundance, n ¼ 175). RP classes

represent sites with RP scores consistently remaining within the RP class

during a four year period (2003–2006) and %SPEAR abundance repre-

sents averages from the subsequent years (2004–2007) (n ¼ 4). Error bars

indicate SE.

Dow

nloa

ded

by S

tats

bibl

iote

ket o

n 14

Feb

ruar

y 20

12Pu

blis

hed

on 0

9 M

arch

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0EM

0074

5E

View Online

not to offset a runoff event.27 According to Linders et al.28 plant

interception values (Ij) were assigned to all crop types which were

present during the considered precipitation event. No pesticide

runoff episodes were predicted if either no agricultural land was

present in the RP-square or no precipitation event exceeded

10 mm during May and June. In this case a minimum RP value

was assigned indicating potential but minimal pesticide runoff.

The assigned value was equal to 50% of the maximum RP

calculated for the site with the smallest amount of agricultural

land in the RP-square.

2.4 Data analysis

Upstream riparian habitat quality was assessed as the summed

relative proportion of all types of undisturbed areas (forest,

undisturbed dry areas and undisturbed wet areas) in the two-

sided 50 m buffer extending from the sampling site and upstream.

We applied catchment size as a measure of stream size, because

stream width is influenced by human activities (channelisation

and dredging).

Predicted pesticide exposure ranged from �9.96 (indicating

potential but minimal risk of exposure) to �0.23. RP scores were

further subdivided into four categories (RP > �1 to 0; RP > �2

to �1; RP > �3 to �2 and RP # �3). Sites were grouped into

those consistently being in the same RP class and those changing

RP class during the period 2003–2006. 175 sites were consistently

assigned to the same RP class in the years 2003–2006 whereas 37

sites changed RP class. At sites consistently remaining within the

same RP class we correlated macroinvertebrate community

structure in March or April (2004–2007) with predicted pesticide

exposure in May or June the preceding year. The effect of pre-

dicted pesticide exposure on %SPEAR abundance was tested

with one-way ANOVA (P < 0.05). The effect of stream size and

upstream riparian habitat quality on %SPEAR abundance was

tested with Pearson correlation, and, subsequently we tested the

influence of RP on the correlation between %SPEAR abundance

and upstream habitat quality with ANCOVA (p < 0.05).

Furthermore, we correlated all physical/chemical properties with

all macroinvertebrate community descriptors with Pearson

correlation. ANOVA’s, ANCOVA’s and Pearson correlations

were conducted using SAS Enterprise Guide 4.2. Performing

DCA analyses we down weighed rare species by disregarding all

species present in fewer than five samples. Furthermore, samples

with five or fewer species were disregarded in order to enhance

the signal from the most common trends. DCA and PCA

analyses were performed applying PC-ORD 4.25.

3. Results

3.1 Correlation between runoff potential and %SPEAR

abundance

We found that %SPEAR abundance was not significantly

different among RP classes (P¼ 0.433) (Fig. 2). Furthermore, we

found no significant correlation between RP and %SPEAR

abundance (n ¼ 175, P ¼ 0.428). %SPEAR abundance increased

significantly with increasing proportions of undisturbed area in

the two sided 50 metres buffer extending from the sampling point

and upstream (P ¼ 0.016) (data not shown). There was no

significant effect of catchment area (representing stream size) on

946 | J. Environ. Monit., 2011, 13, 943–950

%SPEAR abundance (P ¼ 0.24). We, additionally, tested the

potential interaction of RP class on the correlation between

proportion of undisturbed area in the two sided 50 metre buffer

and %SPEAR abundance and found no significant difference

among RP classes (P ¼ 0.366) (data not shown).

3.2 Macroinvertebrate communities and environmental

parameters

We plotted samples (n ¼ 656) in a DCA ordination (considering

species qualitatively and quantitatively) with DCA axes one and

two in combination explaining 11% of the variation in the data

(data not shown). We found no significant differences among

correlations between DCA values and RP class (P > 0.05). We

correlated DCA axis scores with site and catchment based

environmental parameters applying Pearson correlation analysis

and found that relative proportions of gravel, silt and the

upstream riparian habitat quality explained most of the variation

in macroinvertebrate community composition (horizontal DCA

axis one) (r ¼ 0.398, r ¼ 0.351 and r ¼ 0.310, respectively, P <

0.001) (Table 2). The catchment area explained most variation in

the macroinvertebrate community along the vertical DCA axis

two (r ¼ 0.425, P < 0.001) followed by the relative proportion of

silt and substrate heterogeneity (r ¼ 0.22 and r ¼ 0.215, respec-

tively, P < 0.001).

The PCA analysis of calculated macroinvertebrate community

descriptors (n ¼ 175) explained 62% of the total variation in the

dataset (Fig. 3). Vectors depicting the six macroinvertebrate

community descriptors explaining most of the variation in the

dataset are added (SPEAR, The Saprobic index, DSFI, #EPT

taxa, EPT abundance and total #taxa). Selective criterion was |r|

> 0.2. Vectors representing Shannon–Wiener diversity and

Evenness are, furthermore, added to the plot despite weaker

explanatory power. The vector length represents explanatory

power. The Saprobic index and %SPEAR abundance demon-

strated strongest explanatory power and together with DSFI and

EPT abundance primarily explained variation along the plane



Table 2 Pearson correlations with DCA ordination axes. Only thestrongest correlations of 29 variables analysed are included in the table.All environmental variables not included here but described in Methodsdid not correlate significantly (p < 0.01) with any of the DCA axes. Valuesindicate |r| and asterisks indicate significant correlations (**p < 0.01 and***p< 0.001). In the case of autocorrelation characterised by |r| > 0.5, theenvironmental parameter with the strongest correlation coefficient ispresented

Parameter DCA 1 DCA 2

Catchment size — 0.425***Substrate heterogeneity — 0.215***Silt (%) 0.351*** 0.220***Gravel (%) 0.398*** —Sand (%) 0.215*** —Agriculture in the 100 m � 100 m

quadrant (%)0.112** 0.150***

Upstream riparian habitat quality 0.310*** —

ver

teb

rate

com

mu

nit

yd

escr

ipti

ve

para

met

ers

an

den

vir

on

men

tal

vari

ab

les.

Ast

eris

ks

ind

icate

corr

ela

tio

nic

ate

wh

eth

erco

rrel

ati

on

sare

po

siti

ve

or

neg

ati

ve,

resp

ecti

vel

y

ind

ex#

Ma

cro

inv

.ta

xa

#E

PT

tax

aE

PT

ab

un

da

nce

Sh

an

no

n–

Wie

ner

div

ersi

ty

*0

.16

**

*0

.17

*0

.24

**

*0

.09

*�

0.1

8*

**

�0

.31

**

*�

0.4

4*

**

�0

.05

*0

.06

0.2

7*

**

0.2

1*

**

0.1

2*

*�

0.0

5�

0.1

6*

**

�0

.13

**

�0

.1*

*�

0.1

1*

�0

.21

**

*�

0.0

6�

0.0

7*

0.3

1*

**

�0

.13

**

�0

.23

**

*0

.22

**

*0

.37

**

*0

.32

**

*0

.03

0.2

4*

**

0.0

80

.10

0.0

70

.02

Dow

nloa

ded

by S

tats

bibl

iote

ket o

n 14

Feb

ruar

y 20

12Pu

blis

hed

on 0

9 M

arch

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0EM

0074

5E

View Online

from top left to down right of the ordination space. Community

descriptors based on the presence/absence of species explained

variation along the plane from top right to down left.

Considering all environmental parameters the local physical

characteristics in terms of substrate types were most strongly

correlated to macroinvertebrate community descriptors

(Table 3). Especially the relative proportion of silt, sand and

gravel was strong governing factors applying a negative (silt) or

positive impact (sand or gravel) on macroinvertebrate commu-

nity structure. Increasing concentrations of ortho-phosphate and

BOD, additionally, had a negative impact on the macro-

invertebrate community structure (especially %SPEAR abun-

dance) whereas predicted pesticide runoff (RP) only was weakly

correlated with %SPEAR abundance (r ¼ 0.08, P > 0.05). RP

Fig. 3 Scatter plot of streams with consistent predicted pesticide expo-

sure during pesticide spraying season from 2003–2006 in a PCA ordi-

nation (n ¼ 175). PCA axes one and two together explained 62% of the

total variation in the dataset. Vectors representing selected macro-

invertebrate community descriptors (%SPEAR abundance, the Saprobic

index, EPT abundance, the number of EPT taxa, the total number of

taxa, DSFI, Shannon–Wiener diversity and evenness) are depicted. The

vector length is congruent with explanatory power of the community

descriptor. Ta

ble

3P

ears

on

corr

elati

on

coef

fici

ents

are

pre

sen

ted

for

ase

lect

edsu

bse

to

fm

acr

oin

sig

nifi

can

cele

vel

(*p

<0

.05

,*

*p

<0

.01

an

d*

**p

<0.0

01).

Po

siti

ve

or

neg

ati

ve

valu

esin

d

Ph

ysi

cal/

chem

ica

lp

rop

erti

es%

SP

EA

Ra

bu

nd

an

ceD

SF

IS

ap

rob

ic

Sa

nd

0.2

9*

**

0.3

0**

*�

0.3

4**

Sil

t�

0.4

2*

**

�0

.41

**

*0

.45

**

Gra

vel

0.3

8*

**

0.4

5**

*�

0.2

6**

ort

ho-P

�0

.25

**

*�

0.1

1*

0.1

5**

BO

D�

0.2

2*

**

�0

.2*

**

0.2

1**

Ma

cro

ph

yte

cov

era

ge

�0

.26

**

*�

0.2

***

0.2

8**

#M

acr

op

hyte

spec

ies

0.0

80

.22

**

*0

.02

RP

0.0

80

.07

0.0

6

This journal is ª The Royal Society of Chemistry 2011 J. Environ. Monit., 2011, 13, 943–950 | 947


Dow

nloa

ded

by S

tats

bibl

iote

ket o

n 14

Feb

ruar

y 20

12Pu

blis

hed

on 0

9 M

arch

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0EM

0074

5E

View Online

was, furthermore, correlated with both ortho-phosphate (r ¼0.313, P ¼ 0.034) and BOD (r ¼ 0.197, P ¼ 0.044). DSFI score,

saprobic value, %SPEAR abundance, the number of EPT taxa

and EPT abundance were negatively correlated to macrophyte

coverage, whereas the number of macrophyte species was posi-

tively correlated to DSFI scores, the number of EPT taxa and

Shannon–Wiener diversity. The macrophyte coverage was,

furthermore, correlated to both relative proportions of silt (r ¼0.374, P < 0.001) and sand (r ¼ 0.223, P < 0.001).

4. Discussion

4.1 Predicted pesticide runoff and SPEAR abundance

Predicted pesticide runoff (RP) ranged approximately ten orders

of magnitude from �9.96 to �0.23. Schriever et al.9 applied the

runoff potential model for predicting pesticide runoff in German

lowland streams generating RP values ranging from �9.67 to

�0.36. Worst case scenario and range of predicted pesticide

runoff are therefore comparable between these two studies.

We did not find any significant effect of consistent predicted

pesticide runoff on %SPEAR abundance. One main explaining

factor is probably that fauna samples were collected in early

spring prior to the main pesticide application season. The

potential impact from agriculturally derived pesticides was,

consequently, related to exposure scenarios the preceding year

leaving ten months for the macroinvertebrate community to

recover. Several studies confirm that the signal of pesticide

impact on the macroinvertebrate community structure in streams

is strongest during exposure (pesticide spraying season) and in

the following post-exposure months.6,29,30

There was no significant effect of stream size on %SPEAR

abundance, indicating that the increasing connectivity between

stream and land with decreasing stream size did not translate into

increasing impact of predicted pesticide runoff from agricultural

fields. However, other studies confirmed that the footprint of

pesticides is more distinct in small streams where the interaction

between stream and land is more intense.8,31 Furthermore, the

dilution effect decreases with decreasing stream size due to

decreasing discharge.

The proportion of undisturbed area in the two sided 50 metre

buffer extending from the sampling site and upstream had

a significant positive effect on %SPEAR abundance. The pres-

ence of undisturbed areas in the stream corridor potentially acts

as large scale spatial refugia for macroinvertebrates promoting

the process of recolonisation and system recovery. This is facil-

itated by increased net downstream transport of especially

Ephemeroptera, Plecoptera and Trichoptera (EPT) as these

insect orders often maintain high densities at such habitats due to

higher habitat quality and complexity of both stream and

riparian zones.32,33 Our findings are supported by Liess and von

der Ohe8 and Schriever et al.9 who, additionally, found increased

%SPEAR abundance at stream sites with forest on both sides of

the stream corridor in the near upstream environment.

When comparing %SPEAR abundance grouped according to

consistent predicted pesticide runoff to those found by Schriever

et al.,9 there is a difference between sites with highest predicted

pesticide runoff. At sites where RP > �2 average %SPEAR

abundance was 31% in Danish streams whereas %SPEAR

948 | J. Environ. Monit., 2011, 13, 943–950

abundance was 30% and 16% in German streams with upstream

forested reaches being present or absent, respectively. Assuming

that the near stream environment is not significantly different

among streams from these studies, %SPEAR abundance was

generally lower in the German streams. One possible explanation

could be that the most extreme runoff incidents in reality are

characterised by higher pesticide concentrations in Germany

than in Denmark. Consequently, the maximum impact on

benthic macroinvertebrates would be smaller in Danish streams

as the maximum concentration is the primary factor determining

severity of effects.34 Such a difference could be mediated by

a higher pesticide application rate, which generally increases with

decreasing latitude.35 This is not considered in the RP model as

application rate was set to a constant value of 1 for a generic

compound. Additionally, quantitative assessments of buffer

strips are not considered in this model but have major impact on

the amount of pesticide reaching surface water recipients via

runoff.36,37 In Denmark, streams are often surrounded by small

buffer strips, partly due to legislative requirements. In the

Braunschweig area, Germany, where Schriever et al.9 conducted

their study, streams were not surrounded by buffer strips

(Matthias Liess, personal communication). Due to the presence

of buffer strips and lower pesticide application rates, we suggest

that Danish streams in general are not impacted by heavy

pesticide runoff contamination. Because these factors are not

included in the runoff model, predicted pesticide contamination

is probably overestimated, which could explain the impaired

correlation between RP and %SPEAR abundance found in this

study. However, relatively high pesticide concentrations have

been detected previously in both water and sediments in Danish

streams,4,38 and in more extreme cases pesticide concentrations

have been significantly correlated to variation in macro-

invertebrate community structure.39 We therefore infer that

agriculturally derived pesticide contamination at least locally can

pose a threat to stream macroinvertebrate communities.

We found that Danish streams in general were characterised

by high macrophyte coverage (65%). Previous studies confirm

that macrophytes are very abundant in small lowland streams in

Denmark.15,40 Other studies indicated lower macrophyte abun-

dance in European streams; e.g. in France (19%), Finland (14%)

and Germany (4–13%).8,9,41 Within macrophyte patches flow is

reduced,42 and the vast majority of a pesticide pulse, conse-

quently, passes in the main flow channel outside/between

patches. Beketov and Liess43 showed that thiacloprid concen-

tration was reduced up to 60% within macrophyte patches during

a 10 hour pulse exposure compared to the non-vegetated main

flow channel. Macrophyte patches therefore potentially provide

important small scale spatial refugia sheltering resident macro-

invertebrates from pesticide exposure. In consequence, benthic

macroinvertebrate community resistance and recovery potential

after pesticide exposure could be improved in macrophyte rich

streams. This could be an additional explaining factor for the

observed high %SPEAR abundance in Danish streams despite

high predicted pesticide runoff exposure. However, as macro-

phyte data were collected from June to September it probably

overestimates macrophyte abundance during the main insecti-

cide application season in May and June. Further studies are

needed to clarify the role of macrophytes as small scale physical

refugia for macroinvertebrates during pesticide exposure.



Dow

nloa

ded

by S

tats

bibl

iote

ket o

n 14

Feb

ruar

y 20

12Pu

blis

hed

on 0

9 M

arch

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0EM

0074

5E

View Online

4.2 Importance of local physical stream properties

Applying all supplementary environmental variables we found

that local physical habitat properties most strongly explained

variation in macroinvertebrate community structure compared

to chemical properties and land-use characteristics. Stream bed

coverage of silt was the most important environmental para-

meter having a negative impact on abundance and species rich-

ness of EPT taxa. Stream bed coverage of gravel and sand were

the two most important parameters having a positive effect on

abundance and species richness of EPT taxa. Most EPT species

are associated with hard substratum or macrophytes and are

depending on high oxygen concentrations resulting from the

turbulent flow prevailing around these structures.15,44 Our find-

ings are supported by Pedersen et al.45 and Pedersen and

Friberg15 who conducted comprehensive studies linking envi-

ronmental variables at different scales to ecological stream

quality in Danish streams. They found that local physical stream

properties primarily were explaining variation in macro-

invertebrate communities in Danish streams. We therefore

consent to the fact that habitat degradation is a major problem in

Danish streams.

The applied macroinvertebrate indices (SPEAR, DSFI,

Saprobic index) responded congruently with abundance and

species richness of EPT taxa to physical habitat property char-

acteristics. As many EPT species are sensitive to various types of

pollution they are often applied as indicators for high stream

ecosystem quality in various macroinvertebrate based biological

indices.14,44

Overall, a multitude of environmental parameters at different

scales co-explain variation in macroinvertebrate communities as

many environmental parameters at different scales are directly or

indirectly inter-correlated. The stream habitat quality (substrate

composition and heterogeneity) is tightly linked to land use in the

catchment with streams draining agricultural land being char-

acterised by increasing relative coverage of silt and decreasing

relative coverage of coarse substrate probably due to stream

channelisation, dredging and weed cutting.13 Depending on

spatial and temporal location of the stream, present biota is

influenced by both small and large scale environmental para-

meters46 as the biota has to pass through several environmental

filters at different spatial scales to potentially persist at any site.

Additionally, local scale filters are influenced by changes in large

scale filters (i.e. at the catchment or regional scale).16,47

5. Conclusion

In this study the negative impact of poor physical habitat

properties overclouded the impact of pesticides as predicted from

the runoff potential model. The runoff potential model, however,

probably overestimated pesticide runoff contamination due to

the presence of buffer strips surrounding Danish streams and due

to the moderate pesticide application rate in Denmark. We

therefore suggest that heavy pesticide contamination from agri-

cultural non-point sources in Danish streams is only a potential

problem locally. Streams influenced by agricultural activities are

often severely impacted by other agriculturally related activities

(e.g. channelisation, dredging and weed cutting) resulting in

physical habitat degradation which, eventually, translates into


the disappearance of sensitive macroinvertebrate species. At least

in the absence of heavy pesticide runoff contamination, physical

habitat properties are primarily governing stream macro-

invertebrate community structure in Danish streams. When

pesticide runoff contamination is low to moderate, it is therefore

important to consider physical habitat properties in risk assess-

ment in order to avoid overestimating toxic impact of pesticides

in streams. This study emphasises the importance of considering

the hierarchical contribution of all potential environmental

stressors when conducting risk assessment of agricultural

streams.

Acknowledgements

This study was financed by the Danish Research Council (grant

no. 2104-07-0035) and is part of the RISKPOINT program. The

authors greatly acknowledge the work of Marianne Pedersen and

Peter Mejlhede who contributed to analysing data. Furthermore,

we thank Morten Lauge-Pedersen and two anonymous reviewers

for constructive comments on earlier drafts of the manuscript.

References

1 R. Schulz, M. Haushild, M. Ebeling, J. Nanko-Drees, J. Wogram andM. Liess, Chemosphere, 1998, 36, 3071.

2 M. Liess, R. Schulz, M. H. D. Liess, B. Rother and R. Kreuzig, WaterRes., 1999, 33, 239.

3 L. E. Gentry, M. B. David, K. M. Smith-Starks and D. A. Kovacic, J.Environ. Qual., 2000, 29, 232.

4 B. Kronvang, H. L. Iversen, H. Løkke, K. Vejrup, B. B. Mogensen,A. Hansen and L. B. Hansen, Pesticide Research No. 69,Copenhagen, 2003.

5 R. Schulz, Chemosphere, 2001, 45, 543.6 M. Liess and R. Schulz, Environ. Toxicol. Chem., 1999, 18, 1948.7 B. Kronvang, H. L. Strom, C. C. Hoffmann, A. Laubel and

N. Friberg, Water Sci. Technol., 2002, 49, 139.8 M. Liess and P. C. von der Ohe, Environ. Toxicol. Chem., 2005, 24,

954.9 C. A. Schriever, M. H. Ball, C. Holmes, S. Maund and M. Liess,

Environ. Toxicol. Chem., 2007, 26, 346.10 J. J. Rasmussen, N. Friberg and S. E. Larsen, Aquat. Toxicol., 2008,

90, 228.11 M. A. Beketov and M. Liess, Environ. Toxicol. Chem., 2005, 24, 1281.12 R. Schulz and M. Liess, Hydrobiologia, 1995, 299, 103.13 M. L. Pedersen, Fundam. Appl. Limnol., 2009, 174, 89.14 N. Friberg, J. Skriver, S. E. Larsen, M. L. Pedersen and A. Buffagni,

Freshwater Biol., 2010, 55, 1405.15 M. L. Pedersen and N. Friberg, Fundam. Appl. Limnol., 2009, 174, 27.16 L. Sandin, Fundam. Appl. Limnol., 2009, 174, 75.17 C. A. Schriever, P. C. von der Ohe and M. Liess, Chemosphere, 2007,

68, 2161.18 Danish Environmental Protection Agency, Pesticide statistics,

Orientering fra Miljøstyrelsen nr. 5, Copenhagen, 2003 (in Danish).19 Danish Environmental Protection Agency, Pesticide statistics,



Orientering fra Miljøstyrelsen nr. 5, Copenhagen, 2006 (in Danish).22 M. L. Pedersen, A. Baattrup-Pedersen and P. Wiberg-Larsen, Teknisk

anvisning fra DMU nr. 21., Aarhus, 2008.23 M. Liess, P. C. von der Ohe, C. A. Schriever, R. Sch€afer and

M. A. Beketov, Online Database of Species at Risk (SPEARdatabase), 2006, http://www.ufz.de/index.php?en ¼ 2138.

24 A. Baattrup-Pedersen and T. Riis, Freshwater Biol., 1999, 42, 375.25 M. H. Greve, M. B. Greve, P. K. Bøcher, T. Balstrøm and L. Krogh,

Geografisk Tidsskrift, 2007, 107, 2.26 D. S. G. Thomas and A. Goudie, The Dictionary of Physical

Geography, Blackwell Publishing Ltd., 3rd edn, 2000.

J. Environ. Monit., 2011, 13, 943–950 | 949


Dow

nloa

ded

by S

tats

bibl

iote

ket o

n 14

Feb

ruar

y 20

12Pu

blis

hed

on 0

9 M

arch

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0EM

0074

5E

View Online

27 R. D. Wauchope, J. Environ. Qual., 1978, 7, 459.28 J. Linders, H. Mensink, G. Stephenson, D. Wauchope and K. Racke,

Pure Appl. Chem., 2000, 72, 2199.29 P. K. Sibley, N. K. Kaushik and D. P. Kreutzweiser, Environ. Pollut.,

1991, 70, 35.30 L. H. Heckmann and N. Friberg, Environ. Toxicol. Chem., 2005, 24, 582.31 M. Probst, N. Berenzen, A. Lentzen-Godding, R. Schulz and

M. Liess, Ecotoxicol. Environ. Saf., 2005, 60, 140.32 R. E. Lester and A. J. Boulton, Environ. Manage., 2008, 42, 310.33 T. Muotka and J. Syrjanen, Freshwater Biol., 2007, 52, 724.34 R. Schulz and M. Liess, Chemosphere, 2000, 41, 1511.35 C. A. Schriever and M. Liess, Sci. Total Environ., 2007, 384, 264.36 H. Kl€oppel, W. K€ordel and B. Stein, Chemosphere, 1997, 35, 129.37 M. Probst, N. Berenzen, A. Lentzen-Godding and R. Schulz,

Ecotoxicol. Environ. Saf., 2005, 62, 145.

950 | J. Environ. Monit., 2011, 13, 943–950

38 B. Kronvang, A. Laubel, S. E. Larsen and N. Friberg, Hydrobiologia,2003, 494, 93.

39 N. Friberg, M. Lindstrom, B. Kronvang and S. E. Larsen,Hydrobiologia, 2003, 494, 103.

40 A. Baattrup-Pedersen and S. E. Larsen, Hydrobiologia, 2003, 495,171.

41 R. B. Sch€afer, T. Caquet, K. Siimes, R. Mueller, L. Lagadic andM. Liess, Sci. Total Environ., 2007, 382, 272.

42 K. Sand-Jensen, Freshwater Biol., 1998, 39, 663.43 M. A. Beketov and M. Liess, Environ. Pollut., 2008, 156, 1364.44 R. G. Young and K. J. Collier, Freshwater Biol., 2009, 54, 2155.45 M. L. Pedersen, N. Friberg and S. E. Larsen, River Res. Appl., 2004,

20, 653.46 L. Sandin, Fundam. Appl. Limnol., 2009, 174, 1.47 N. L. Poff, J. North Am. Benthol. Soc., 1997, 16, 391.



Paper III

at SciVerse ScienceDirect

Environmental Pollution 164 (2012) 142e149

Contents lists available

Environmental Pollution

journal homepage: www.elsevier .com/locate/envpol

Stream habitat structure influences macroinvertebrate response to pesticides

Jes Jessen Rasmussen*, Peter Wiberg-Larsen, Annette Baattrup-Pedersen, Nikolai Friberg, Brian KronvangAarhus University, Department of Bioscience, Vejlsøvej 25, 8600 Silkeborg, Denmark

a r t i c l e i n f o

Article history:Received 9 November 2011Received in revised form4 January 2012Accepted 8 January 2012

Keywords:MacroinvertebratesSPEARPesticide stressPhysical habitatsMultiple stressorsRisk assessment

* Corresponding author.E-mail address: [email protected] (J.J. Rasmussen).

0269-7491/$ e see front matter � 2012 Elsevier Ltd.doi:10.1016/j.envpol.2012.01.007

a b s t r a c t

Agricultural pesticides continue to impair surface water ecosystems, although there are few assessmentsof interactions with other modifications such as fine sediment and physical alteration for flood drainage.We, therefore, surveyed pesticide contamination and macroinvertebrates in 14 streams along a gradientof expected pesticide exposure using a paired-reach approach to differentiate effects between physicallymodified and less modified sites. Apparent pesticides effects on the relative abundance of SPEcies At Risk(SPEAR) were increased at sites with degraded habitats primarily due to the absence of species withspecific preferences for hard substrates. Our findings highlight the importance of physical habitatdegradation in the assessment and mitigation of pesticide risk in agricultural streams.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Ecology is being increasingly integrated into ecotoxicology toimprove the understanding of how contaminants impact naturalecosystems. This interest has been particularly strong in aquaticecosystems with a special emphasis on the non-target effects ofpesticides (Liess et al., 2005; Relyea and Hoverman, 2006). In thesesame systems, the use of biological traits to appraise the effects ofnatural disturbances on macroinvertebrate communities is prob-ably more advanced than in any other system (Paillex et al., 2009;Townsend and Hildrew, 1994; Townsend et al., 2008), and thisapproach is considered promising for the detection and interpre-tation of the impact of pesticides (Baird et al., 2011).

The distribution and abundance of specific traits of lotic mac-roinvertebrates are remarkably constant among similar types ofecosystems e even across biogeoregions e whereas the speciescomposition is muchmore variable (Baird and Van den Brink, 2007;Rubach et al., 2011). Moreover, biological traits respond quitespecifically to certain sources of anthropogenic stress such as landuse (Doledec et al., 2006), sewage effluents (Statzner et al., 2001)and the periodic exposure of agricultural pesticides (Liess and Vonder Ohe, 2005; Rubach et al., 2010) making the trait approachpotentially extremely valuable in assessing the effects of multiplestressors that frequently co-occur in stream ecosystems.

All rights reserved.

One example of the assessment of ecological risk using biolog-ical traits is the SPEcies At Risk index (SPEAR) which is an ecologicalindicator system that is based on the biological traits of lotic andlentic macroinvertebrates (Liess and Von der Ohe, 2005). TheSPEAR index applies physiological (sensitivity to pesticides andother organic toxicants) and biological (generation time, timing andduration of terrestrial life stages and dispersal capacity) traits todetermine the relative abundance of the species that are at risk ofbeing impacted by the periodic exposure of pesticides. The SPEARindicator has been demonstrated to remain relatively constantamong reference stream ecosystems and, more importantly itresponds selectively to pesticide stress in small streams acrossdifferent biogeoregions (Beketov and Liess, 2008; Liess and Von derOhe, 2005; Schäfer et al., 2007, 2011).

However, in spite of the clear strength in using biological traitsfor risk assessment such as SPEAR, the potential interaction ofdifferent anthropogenic stressors may prompt altered effects(Matthaei et al., 2010; Rasmussen et al., 2011a; Wagenhoff et al.,2011). As the effect of one stressor may reduce the fitness of anorganism, the effect of another stressor may increase. Conversely,the effect of one stressor may result in competitive release for oneorganism (e.g. elimination of predators) increasing its fitness andpotentially reducing its sensitivity to other stressors (Pedersen andFriberg, 2009). One such potential stressor that often interacts withthe presence of pesticides in agricultural areas is the intensivemaintenance of streams by channelisation and dredging, usedto maximise flow conveyance into the stream channel and along it.One side-effect of the maintenance activities is a severe


www.sciencedirect.com/science/journal/02697491

http://www.elsevier.com/locate/envpol

http://dx.doi.org/10.1016/j.envpol.2012.01.007



J.J. Rasmussen et al. / Environmental Pollution 164 (2012) 142e149 143

degradation of the diversity and quality of physical habitats in thestreams. The physical structure of heavily maintained streams isoften characterised by higher proportions of soft substrates and lowvariability in depth and width of the streams (Pedersen, 2009). Inaddition, the degradation of physical habitats results in morelaminar and in general slower flow which reduces the reaerationcapacity and periodically the total oxygen content in the streamwater (Friberg et al., 2009; Pedersen, 2009).

Significant variation in the community structure of benthicmacroinvertebrates has been ascribed to substrate characteristics(e.g. particle size) and habitat heterogeneity (Brunke et al., 2001;Pedersen and Friberg, 2009). Many lotic Ephemeroptera, Plecopteraand Trichoptera (EPT) require high current velocities, welloxygenated water and coarse substrate (Giller and Malmqvist,1998), and moreover many EPT taxa have narrower nicherequirements than taxa related to physically uniform streams withsoft sediments (Dunbar et al., 2010). Therefore, high habitatheterogeneity is essential if the site is to support a larger proportionof the EPT taxa. The majority of EPT species are additionally char-acterised as at risk in the SPEAR index (Liess and Von der Ohe,2005). In consequence, the species pool of potential macro-invertebrate colonisers is severely reduced for reaches withdegraded physical conditions e and/or resident species may addi-tionally experience increasing sensitivity to pollutants due to sub-optimal physical conditions at the level of micro-habitats. Inconsequence, both of these factors could influence the outcome ofthe SPEAR index as previously insinuated by Rasmussen et al.(2011a) and Schletterer et al. (2010). Riffle sections of streamshave been prioritised for macroinvertebrate sampling in previousstudies using the SPEAR index (Liess and Von der Ohe, 2005;Schletterer et al., 2010; Schäfer et al., 2007, 2011; von der Ohe et al.,

Fig. 1. Geographic overview of the study sites (to the left) and a schematic presentation of this used as a surrogate measure for physical habitat characteristics. The figure is modified a

2007). To our knowledge studies that address pesticide effects onconjunction with channel modifications have not been conductedbefore.

Theprincipal aimof this studywas to investigate the relationshipbetween pesticide contamination and macroinvertebrate commu-nity composition in streamswith differing structure and diversity ofphysical micro-habitats. We sampled 14 Danish 1st and 2nd orderstreams in each ofwhich two reacheswith differing physical habitatcharacteristics were selected. We used this set-up to test thehypotheses that 1) the fraction of SPEcies At Risk (SPEAR) is lower atreaches that are characterised by degraded physical habitatscompared to reaches with more unmodified habitats due to espe-cially the preference for hard substrate formany EPT taxa, and 2) thereduction of the fraction of SPEAR during the main pesticide appli-cation season is higher at sites that are characterised by degradedhabitats as a consequence of increased multiplicative stress.

2. Methods

2.1. Study design

The field campaign was conducted in 2009 in 14 1st and 2nd order streamslocated on Funen, Denmark. The region is dominated by agriculture and the studycatchments are characterised by low elevation and loamy/clayey soils (Fig. 1). Rye,wheat, barley, grass and oilseed rape were the dominating crop types (see alsoRasmussen et al., 2011b). Of all pesticides that were expected to be applied onto thefields, 48% were herbicides, 28% were fungicides (primarily boscalid and triazolecompounds) and 24% were insecticides (mainly pyrethroids) (Danish EPA, 2010).

The study streams were selected based on the following criteria: 1) year-roundwater flow, 2) no physical maintenance activities during the sampling period, and 3)no sources of pollution other than those originating from normal agriculturalpractices. The streams were selected to exhibit a gradient of agricultural pressureand potential exposure to agricultural pesticides (predicted from the proportion ofadjacent agricultural land).

e neighbouring reaches of 50 m in each of study streams (to the right). Current velocityfter Rasmussen et al. (2012).

J.J. Rasmussen et al. / Environmental Pollution 164 (2012) 142e149144

In each study stream, two reaches of 50 m were selected based on the physicaland hydromorphological properties of the reach (Fig. 1). The selection criteria for thereaches were 1) presence of an upstream reach characterised by low currentvelocity, low physical diversity and low reaeration capacity; 2) a downstream reachcharacterised by fast flow, high physical diversity and high reaeration capacity, and3) the distance between reaches should not exceed 250 m in order to maximise thesimilarity of water chemistry between reaches. For convenience reaches are referredto as sampling sites in the succeeding parts of the article.

2.2. Pesticide monitoring and water chemistry

A total of 19 herbicides, 6 fungicides and 6 insecticides were monitored repre-senting some of the most frequently applied compounds in Danish agriculture alongwith presently banned compounds that are commonly found in groundwatermagazines (Supplementary material A). Pesticides were sampled with event-triggered samplers (Liess and Von der Ohe, 2005) during the spraying season(May and June) and additional water samples were collected manually in Augustduring low flow to characterise the pesticide contamination that originates fromgroundwater inflow. Two major precipitation events triggered some (the 28th ofMay) or all (the 12th of June) event-triggered samplers. A detailed overview of thesampling methods, chemical analysis methods and detected pesticide compoundsare described elsewhere (Rasmussen et al., 2011b, 2012).

To compare the potential impact of pesticides on macroinvertebrates amongstreams, the measured pesticide concentrations were used to calculate the logmaximum Toxic Unit (mTU) of all compounds detected in each sample. According toTomlin (2000) the calculated log TU for each compound was based on the standard48-h acute LC50 value for Daphnia magna:

log TUD: magna ¼ logðCi=LC50iÞ (1)

where log TUD. magna is the toxic unit for pesticide i, Ci is the measured concentrationof the pesticide i and LC50i is the corresponding 48-h LC50 value for D. magnaexposed to the pesticide i.

Water samples for general chemistry were collected manually in April, June andAugust (Table 1). The BOD5, ortho-phosphate and ammonia-N were analysedaccording to European Standards (DS/EN 1899 1999, DS/EN 1189-1997 and DS 117322005, respectively). Nitrate-N was analysed using Lachat-methods (Lachat Instru-ments, USA, Quickchem. No. 10-107-06-33-A (Salycate method)). Concentrations oftotal N and total P were measured (unfiltered samples) applying the Kjeldahl-Nmethod (Kjeldahl, 1883) and Danish standard (DS-291), respectively.

2.3. Physical properties

For each of the sampling sites twenty-five transects were established with twometer intervals. Results from the physical surveys are presented in Table 1 and infurther detail in Rasmussen et al. (2012). Stream width, (W), depth (D) and water

Table 1Physical and chemical characteristics for two sites in each of 14 streams. Theupstream site was characterised by homogenous and heavily modified physicalproperties, and the downstream site was characterised by more diverse andunmodified physical properties. Values represent the average (�SE) of three surveysconducted in April, June and August, 2009.

Parameter Reaches withheterogeneousphysical conditions

Reaches withhomogenousphysical conditions

Discharge (L s�1) 4.7 � 0.3 2.8 � 0.2Shear stress 6.5 � 0.3 5.5 � 0.2Width (m) 1.27 � 0.02 1.4 � 0.01Depth (m) 0.05 � 0.01 0.10 � 0.01Cobble (%) 14.2 � 0.6 4.0 � 0.4Gravel (%) 42.8 � 1.2 10.7 � 1.9Sand (%) 18.1 � 0.6 24.5 � 1.4Silt (%) 16.3 � 0.4 50.1 � 1.6Debris (%) 6.1 � 0.4 8.4 � 0.7# Submergent

macrophyte species0.7 � 0.0 0.9 � 0.1

Total macrophytecoverage (%)

13.4 � 1.3 22.3 � 0.8

Temperature (�C) 11.3 � 0.4 14.2 � 0.8BOD5 (mg L�1) 1.90 � 0.1 1.98 � 0.1Oxygen concentration (mg L�1) 8.67 � 0.4 8.32 � 0.7Ammonia (mg L�1) 0.07 � 0.01 0.06 � 0.01Nitrate (mg L�1) 2.21 � 0.12 2.29 � 0.11Total N (mg L�1) 2.91 � 0.12 2.79 � 0.12Ortho-Phosphate (mg L�1) 0.07 � 0.00 0.06 � 0.00Total P (mg L�1) 0.13 � 0.00 0.16 � 0.01

velocity (0.4 � depth from the stream bed) (U) were measured at four points alongeach transect corresponding to 25, 50, 75 and 100% of the wetted width. Currentvelocity was measured using a flow-meter (Höntzsch mP-TAD). The discharge wascalculated from average stream dimensions and average current velocity. Further-more, four rectangular plots were established between each pair of transects(2 m � 25 % of the stream width). Substrate types and macrophyte coverage werequantified in each plot, and submergent and emergent macrophytes were identifiedto the lowest possible taxonomical level. The proportional coverage for each taxonwas also quantified. Stream water temperature, conductivity and oxygen concen-tration were registered using a multi-meter (WTW multi-350i) and pH wasmeasured with a (YSI-60) pH-meter.

2.4. Macroinvertebrates

Macroinvertebrates were sampled in April, June and August, 2009 using a surbersampler (200 cm2) (Supplementary material B). Five surber samples were collectedfrom each site according to a predefined randomised sampling approach andsamples were preserved in 70% ethanol in the field. All macroinvertebrates werecounted and identified to the level of species or genus, except Chironomidae (sub-family), Simuliidae (family), Oligochaeta (order). In addition, %SPEAR abundancewas computed for each of the sites and sampling months using the freely availableonline SPEAR calculator (http://www.systemecology.eu/SPEAR/index.php).

2.5. Data analysis

Applying a principal component analysis (PCA) the twomost important stressorswere identified in a companion paper (Rasmussen et al., 2012) with one beingcharacterised by land-use intensity and log mTUD.magna and the other being charac-terised by substrate characteristics (relative coverage of boulders, gravel and silt).

Separate linear regressions were performed for the two site categories (homoge-neous and heterogeneous physical conditions) on the relationship between log mTUD.

magna and %SPEAR abundance and the relationship between the log mTUD.magna and%SPEARmod abundance. We used Analysis of Covariance (ANCOVA) to test for differ-ences in slopes and intercepts between regression lines (P < 0.05) applying thesoftware SAS 9.1.

We used the method from Dufrêne and Legendre (1997) to calculate indicatorvalues for all taxa that were detected in June samples in order to characterise theprobability for a taxon to occur in a certain habitat type. A significance threshold of0.1 was applied to promote the most overall trends. The series of permutations wereperformed distinguishing between a) sites with homogeneous and heterogeneousphysical conditions. The indicator values were tested for statistical significance usinga Monte Carlo technique, and the analyses were performed using PCOrd 4. The levelof genus was applied for the caddisflies belonging to the genus Silo, since the twospecies that were found in the study streams (Silo palipes and Silo nigricornis) havesimilar preferences for habitats with coarse substrate (Wallace et al., 2003). Weadditionally calculated a modified %SPEAR abundance (SPEARmod) excluding thespecies that were found to be characteristic for reaches with heterogeneous physicalconditions in order to evaluate if these taxa were the primary reason for lower %SPEAR abundance at sites with homogenous physical conditions. In addition, wecalculated the %SPEAR abundance only using taxa belonging to Ephemeroptera,Plecoptera and Trichoptera (EPT) or only using the taxa that did not belong to EPT(SPEAREPT and SPEARnotEPT, respectively) to evaluate whether potential differencesin %SPEAR abundance between site types was governed by SPEAR taxa belonging ornot belonging to EPT. The SPEARmod, SPEAREPT and SPEARnotEPT were calculatedusing the SPEAR classification from Liess and Von der Ohe (2005).

Lastly, we grouped the sites according to their physical conditions (homoge-neous and heterogeneous) and the measured toxicity of pesticides during stormflow (TU � �5, �5 < TU � �3 and TU > �3). The temporal development from Aprilto August in average %SPEAR abundance, %SPEAREPT abundance and %SPEARnotEPT

abundance were compared for each of the six site groups (P < 0.05). Moreover, wecompared the average %SPEAR abundance between the site groups for each of themonths April, June and August.

Lastly, we compared the average difference in %SPEAR abundance between thesites with heterogeneous and homogeneous physical properties among the sitegroups (TU��5,�5< TU��3 and TU>�3, representing unimpacted, moderatelyand heavily impacted streams, respectively) for each of the months April, June andAugust using one way analysis of variance. In the case of significant differencesamong groups (P < 0.05, ANOVA) a Bonferroni t-test was used as post hoc test toevaluate pairwise differences. The ANOVAS were performed using SAS EnterpriseGuide 4.2.

3. Results

3.1. Physical habitat properties and SPEAR

The log mTUD. magna was consistently higher for runoff-relatedwater samples compared to base-flow water samples. The

http://www.systemecology.eu/SPEAR/index.php

A

B

C


pesticide compounds causing the log mTUD. magna in storm flowwater samples differed along the gradient of agricultural intensityin the catchment (Table 2).

The log mTUD. magna explained a significant proportion of thevariation in %SPEAR abundance in April, June and August for siteswith homogeneous (R2 ¼ 0.60, P < 0.001; R2 ¼ 0.60, P < 0.001;R2 ¼ 0.61, P < 0.001) and heterogeneous physical properties(R2 ¼ 0.41, P < 0.05; R2 ¼ 0.68, P < 0.001; R2 ¼ 0.40, P < 0.05)(Fig. 2). Slopes of the regression lines for heterogeneous andhomogenous sites in April, June and August were not significantlydifferent, whereas the intercepts were significantly different inJune (P < 0.05) but not in April and August (Fig. 2).

Since the regression lines characterising the correlationbetween log mTUD. magna and %SPEAR abundance were onlysignificantly different between site groups in June, the subsequentanalyses for indicator taxa were only conducted for June data. InJune three taxa were significant indicators for sites with homoge-neous physical conditions and seven taxa were significant indica-tors for sites with heterogeneous physical conditions (P < 0.1)(Table 3). Fully detailed results of the analyses for indicator taxa canbe found in the Supplementary material C. Three of the indicatortaxa for heterogeneous physical conditions are additionally rankedas SPEcies At Risk (Dicranota sp., Leuctra fusca and Silo spp.) in theSPEAR classification (Liess and Von der Ohe, 2005). Removing thesethree taxa from the June taxa list, we re-calculated the %SPEARabundance for the remaining taxa (%SPEARmod abundance) (Fig. 3).The %SPEARmod abundance was significantly correlated with logmTUD. magna for sites with heterogeneous (R2 ¼ 0.61, P < 0.001) andhomogeneous physical conditions (R2¼ 0.55, P< 0.001). The slopesand intercepts of the regression lines representing sites withheterogeneous and homogeneous physical conditions, respectively,were not significantly different (P > 0.05).

3.2. Temporal analysis of macroinvertebrate community dynamics

Grouping sites according to their physical conditions (hetero-geneous and homogeneous) and the measured log mTUD. magna

(unimpacted (log mTUD. magna < �5), moderately impacted(�5 � log mTUD. magna < �3) and heavily impacted (log mTUD.

magna � �3)) we found that the average %SPEAR abundancedecreased significantly from April to June in unimpacted streamswith homogeneous physical conditions (P < 0.05) (Fig. 4A). Nosignificant temporal changes of %SPEAR abundance were observedfor other site groups. The %SPEAREPT significantly decreased from

Table 2The proportion of agriculture in the catchment, log mTUD. magna measured in streamwater during two heavy precipitation episodes in May and June, 2009 and thecompound causing the log mTUD. magna is presented for each of 14 Danish streams.Samples with no pesticide content are indicated as n.d.

Site %Agriculture

log mTUD. magna

in runoffevent #1

log mTUD. magna

in runoffevent #2

Compound

R1 2.2 �6.32 �6.32 Desethylterbutylazine (H)R2 3.0 n.d. �6.63 Desethylterbutylazine (H)R3 2.2 �6.32 �6.15 Desethylterbutylazine (H)R4 2.8 n.d. �6.23 Desethylterbutylazine (H)R5 15.0 n.d. �6.45 Desethylterbutylazine (H)A1 58.0 n.d. �4.95 Boscalid (F)A2 58.8 �5.14 �4.15 Pendimethaline (H)A3 80.6 �2.79 �2.79 Azoxystrobin (F)A4 94.5 �3.23 �2.92 Azoxystrobin (F)A5 83.7 �1.87 �1.72 Pirimicarb (I)A6 98.8 �3.41 �2.73 Pirimicarb (I)A7 87.4 �3.38 �3.32 Azoxystrobin (F)A8 54.4 �3.41 �3.18 Azoxystrobin (F)A9 84.1 �3.13 �2.45 Pirimicarb (I)

Fig. 2. %SPEAR abundance as a function of the log maximum TUD. magna. SPEARcalculations are based on macroinvertebrate samples collected before (A), during (B)and after (C) the main spraying season of agricultural insecticides. The log maximumTUD. magna is based on measured pesticide concentrations during one of two heavyprecipitation events in May and June. Regression lines indicate the linear correlationsof data points that represent sites with heterogeneous (solid line) and homogeneous(dashed line) physical properties.

April to June in unimpacted stream sites with homogeneousphysical conditions (P < 0.05) (Fig. 4B). No significant temporalchanges of %SPEAREPT abundance or %SPEARnotEPT abundance wereobserved for other site groups between either of the months(P > 0.05) (Fig. 4B, C).

In April the average %SPEAR abundance was significantly higherfor unimpacted streams than for heavily contaminated streamsand for the intermediately contaminated stream sites with

Table 3List of taxon groups that were found to be significant indicators for sites withhomogeneous or heterogeneous physical conditions. Silo sp. represents the twoDanish species of the genus: Silo pallipes and Silo nigricornis. P-values indicate thelevel of significance for the MONTE CARLO test of the observed maximum indicatorvalue.

Taxa Physical indicator group P-value

Asellus aquaticus Homogeneous 0.048Baetis rhodani Heterogeneous 0.063Dicranota sp. Heterogeneous 0.034Dugesia gonocephala Heterogeneous 0.070Leuctra fusca Heterogeneous 0.083Limnius volckmari Heterogeneous 0.070Pisidium sp. Homogeneous 0.077Silo spp. Heterogeneous 0.010Simuliidae Heterogeneous 0.013Tanytarsini Homogeneous 0.087

A

B

C


homogeneous physical conditions (P < 0.05) (Fig. 4A). In June theaverage %SPEAR abundance for intermediately and heavilycontaminated streams were significantly lower than unimpactedstream sites with heterogeneous physical conditions. Furthermore,the average %SPEAR abundance for unimpacted stream sites withhomogeneous physical conditions and intermediately contami-nated stream sites with heterogeneous physical conditions weresignificantly higher than the average %SPEAR abundance for inter-mediately contaminated stream sites with homogeneous physicalconditions and heavily contaminated streams. In addition, the %SPEAREPT abundance for intermediately contaminated stream siteswith heterogeneous physical conditions was significantly higherthan intermediately contaminated siteswithhomogeneousphysicalconditions. Other pairwise comparisons (between site groups) ofthe average %SPEAR, %SPEAREPTand %SPEARnotEPTabundancewithinseparate time intervals were not significantly different (P > 0.05).

In June the difference in %SPEAR abundance between sites withheterogeneous and homogeneous physical conditions was signifi-cantly greater for intermediately contaminated streams comparedto heavily contaminated streams (P < 0.05, One-way ANOVAP¼ 0.033). The difference in %SPEAR abundance between sites withheterogeneous and homogeneous physical conditions for uncon-taminated streams was not significantly different from the

Fig. 3. %SPEARmod abundance as a function of the log maximum TUD. magna. SPEARcalculations are based on macroinvertebrate data collected in June during the maininsecticide spraying season. The log maximum TUD. magna is based on measuredpesticide concentrations during heavy precipitation events in May and June. Regres-sion lines indicate the linear correlations of data points that represent sites withheterogeneous (solid line) or homogeneous (dashed line) physical properties.

Fig. 4. Average %SPEAR abundance (A), relative abundance of the SPEAR taxa thatbelong to EPT (B) and relative abundance of the SPEAR taxa that do not belong to EPT(C) before (April), during (June) and after (August) the main spraying season. Datarepresents fourteen pairs of stream sites that were characterised by heterogeneous andhomogeneous physical conditions, respectively. The stream sites are grouped accord-ing to the physical conditions and measured log mTUD. magna. Error bars indicate SE.

difference in %SPEAR abundance between sites with heterogeneousand homogeneous physical conditions for the intermediately orheavily contaminated streams (Supplementary material D). Thedifference in %SPEAR abundance between sites with heterogeneousand homogeneous physical conditions was not significantlydifferent among site categories in April or August (P > 0.05)(Supplementary material D).


4. Discussion

The relative abundance of SPEcies At Risk (SPEAR) during themain pesticide application season (June) was significantly lower atstream sites characterised by uniform and degraded physical habi-tats compared with sites characterised by more undisturbed andheterogeneous physical habitats. This may reflect that macro-invertebrates residing in micro-habitats with soft substrate andslow flow are subject to increased pressure from multiplicativestressors (e.g. decreased boundary layers, increased amplitude indissolved oxygen concentrations and low substrate stability)potentially increasing their susceptibility to pesticide contamina-tion (Maltby, 1999; Pedersen, 2009; Thrush et al., 2008). The factthat themacroinvertebrate organisms couldmigrate freely betweenthe paired set of reaches highlights the strength of these results.Increased effects of pesticides on freshwater macroinvertebrateshave been observed with e.g. decreasing oxygen concentrations(Ferreira et al., 2010), increasing temperature (Singh and Singh,2009) and the presence of predators (Beketov and Liess, 2006;Montoya et al., 2009). The effects of pesticides on stream biota issuggested to increase in more complex systems with several co-occurring stressors (Guy et al., 2011), and Schletterer et al. (2010)additionally indicated that the fraction of the abundance of SPEARsignificantly differed between sub-samples that were collected onsoft and hard substrates. Moreover, hydromorphological degrada-tion is a key pressure on agricultural stream ecosystems that aloneare likely to be amain reason thatmany streamsdonot achieve goodecological status (Friberg et al., 2009; Sandin, 2009; Wagenhoffet al., 2011). Considering the proposed threshold for good ecolog-ical quality (required in the EU Water Framework Directive) in theonline SPEAR calculator (33 %SPEAR abundance) all stream sites(even uncontaminated) with degraded physical conditions wereunable to fulfil the required ecological objectives in June andAugust.

We found that the reduced %SPEAR abundance at stream sitescharacterised by homogeneous physical conditions compared withsites with heterogeneous physical conditions was primarilyexplained by the reduction or absence of the caddis flies Silo pallipesand Silo nigricornis (Goeridae), the stonefly Leuctra fusca (Leuc-tridae) and the crane fly Dicranota sp. (Limoniidae). These taxa arestrongly associated with coarse substrates, turbulent flow and highoxygen concentrations (Schmera, 2004; Urbanic et al., 2005). Thismay suggest that the primary stressor acting on these taxa wasrelated to physical habitat properties rather than pesticidecontamination. Re-computing the %SPEAR abundance withoutthese taxa (%SPEARmod abundance) removed the site-specificdifferences in the correlation with log mTUD. magna. This further-more indicates that the absence of hard substrate rather thanpesticide contamination restricted these three taxa from occupyingthe sampling sites with homogenous and degraded physicalconditions. Interactions between macroinvertebrates and substrateare well documented, and many studies congruently report thatmacroinvertebrate species richness, abundance and diversityincrease with increasing particle size primarily facilitated by EPTtaxa (Brunke et al., 2001; Miyake and Nakano, 2002). Moreover,taxa that require fast flow and coarse substrates tend to have morespecific requirements for physical habitats emphasising that highhabitat heterogeneity is essential if a stream site is to support a highnumber of macroinvertebrate species. In consequence, a consider-able proportion of the macroinvertebrate taxa (primarily EPT taxa)that are classified as SPEAR may be partly or completely restrictedfrom colonising agricultural streams due to physical habitatdegradation. Furthermore, SPEAR taxa that require fast flow andcoarse substrates may become increasingly sensitive to pesticideexposure in physically degraded agricultural streams due to sub-optimal growth conditions (Maltby, 1999).

In June the average %SPEAR abundance was higher for sites withheterogeneous physical properties compared to sites with homo-geneous physical properties within three different contaminationgroups (uncontaminated sites (mTUD. magna < �5), intermediately(�5 � mTUD. magna < �3) and heavily contaminated sites (mTUD.

magna � �3)). However, the difference was only significant for thegroup of intermediately contaminated streams. For these inter-mediately contaminated streams, the significantly lower %SPEARabundance at sites with homogeneous physical conditionscompared to sites with heterogeneous physical conditions wasprimarily governed by the abundance of SPEAR taxa belonging toEphemeroptera, Plecoptera and Trichoptera (EPT). This suggeststhat the abundance of SPEAR taxa is additionally reduced beyondthe effects of pesticides probably due to the additional stressoreffects that are related to physical habitat degradation. However, inJune the difference in %SPEAR abundance between sites withheterogeneous and homogeneous physical conditions was signifi-cantly reduced in heavily contaminated streams compared tointermediately contaminated streams. This insinuates that thepotential positive effects of heterogeneous physical stream condi-tions could become less important or even overruled by pesticidecontamination when the toxic pressure becomes sufficiently high.

The %SPEAR abundance before the main pesticide applicationseason (April) was significantly correlated to the log maximumToxic Unit for 48 h acute exposure to Daphnia magna (log mTUD.

magna) obtained subsequently during the main pesticide applicationseason (May and June) suggesting that sites characterised by highlog mTUD. magna during this field campaign have probably beensubject to similar levels of pesticide contamination in the previousyears. Similar findings have been provided by Liess and Von derOhe (2005) and Liess and Beketov (2011). Indeed, several studieshave shown that sublethal effects of environmentally realisticduration and concentration of pesticide exposure can lead toaltered behaviour and reduced competitive abilities for somestream macroinvertebrates (Fleeger et al., 2003; Liess, 2002;Rasmussen et al., 2008), and the history of intensive agriculturalactivities in the near-stream areas can be a better predictor forbiodiversity than present land-use characteristics (Harding et al.,1998).

5. Conclusions

We found that the %SPEAR abundance during themain pesticideapplication seasonwas significantly lower at stream sites that werecharacterised by uniform and degraded physical habitats. Duringand immediately after the main pesticide application season evenstreams that were not contaminated with diffuse source pesticidese but were characterised by homogeneous and degraded physicalhabitats e were unable to achieve good ecological status (>33%SPEAR abundance). These differences were found to be primarilydue to the presence or absence of macroinvertebrate taxa withspecific requirements for stream habitats that are characterised byhard substrate, turbulent flow and high oxygen concentrations.Moreover, we found that the interaction between the two stressors(physical habitat degradation and diffuse source pesticidecontamination) may change along a gradient in pesticidecontamination.

Our results underline that the summed macroinvertebratecommunity response to the total amount of stress is a result ofdirect and indirect effects and feedback mechanisms that aredefined by species biology and habitat requirements. Consequently,simple interpretation of guidelines and threshold values mayunderestimate responses that are triggered in the natural envi-ronment due to the interaction of multiple stressors such aspesticide contamination and physical habitat degradation. This


study has important implications for the area of multiplicativestress in future research, risk assessment and stream management.

In addition, we suggest creating a formal macroinvertebratesampling strategy in terms of timing of sampling and sampledhabitat types to further improve the applicability and comparabilityof SPEAR. Moreover, the formalisation of a sampling strategy wouldalso improve the usefulness and consistency of results if SPEARwasto be formally implemented as indicator tool for the detection andevaluation of effects of pesticide contamination within the EU.

Acknowledgements

This study was financed by the Danish Research Council (grantno. 2104-07-0035) and is part of the RISKPOINT program. We wishto thank Johnny Nielsen for assisting in the counting and identifi-cation of macroinvertebrates. Additionally, we wish to thank Hen-rik Stenholt, Uffe Mensberg, Dorte Nedergaard and Marlene VenøSkjærbæk and Rikke Juul Monberg for invaluable field assistance.Lastly, we wish to thank two anonymous reviewers for thoroughwork and helpful comments on the original manuscript.

Appendix. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.envpol.2012.01.007.

References

Baird, D.J., Baker, C.J.O., Brua, R.B., Hajibabaei, M., McNicol, K., Pascoe, T., DeZwart, D., 2011. Toward a knowledge infrastructure for traits-based ecologicalrisk assessment. Integrated Environmental Assessment and Management 7,209e215.

Baird, D.J., Van den Brink, P.J., 2007. Using biological traits to predict speciessensitivity to toxic substances. Ecotoxicology and Environmental Safety 67,296e301.

Beketov, M.A., Liess, M., 2006. The influence of predation on the chronic response ofArtemia sp. populations to a toxicant. Journal of Applied Ecology 43, 1069e1074.

Beketov, M.A., Liess, M., 2008. An indicator for effects of organic toxicants on loticinvertebrate communities: independence of confounding environmentalfactors over an extensive river continuum. Environmental Pollution 156,980e987.

Brunke, M., Hoffmann, A., Pusch, M., 2001. Use of mesohabitat-specific relationshipsbetween flow velocity and river discharge to assess invertebrate minimum flowrequirements. Regulated Rivers-Research & Management 17, 667e676.

Danish Environmental Protection Agency, 2010. Pesticide Statistics, Orientering fraMiljøstyrelsen nr. 6 (In Danish), Copenhagen, Denmark.

Doledec, S., Phillips, N., Scarsbrook, M., Riley, R.H., Townsend, C.R., 2006. Compar-ison of structural and functional approaches to determining landuse effects ongrassland stream invertebrate communities. Journal of the North AmericanBenthological Society 25, 44e60.

Dufrêne, M., Legendre, P., 1997. Species assemblages and indicator species: the needfor a flexible asymmetrical approach. Ecological Monographs 67.

Dunbar, M.J., Warren, M., Extence, C., Baker, L., Cadman, D., Mould, D.J., Hall, J.,Chadd, R., 2010. Interaction between macroinvertebrates, discharge and phys-ical habitat in upland rivers. Aquatic Conservation-Marine and FreshwaterEcosystems 20, S31eS44.

Ferreira, A.L.G., Serra, P., Soares, A., Loureiro, S., 2010. The influence of naturalstressors on the toxicity of nickel to Daphnia magna. Environmental Science andPollution Research 17, 1217e1229.

Fleeger, J.W., Carman, K.R., Nisbet, R.M., 2003. Indirect effects of contaminants inaquatic ecosystems. Science of the Total Environment 317, 207e233.

Friberg, N., Sandin, L., Pedersen, M.L., 2009. Assessing the effects of hydro-morphological degradation on macroinvertebrate indicators in rivers: exam-ples, constraints and outlook. Integrated Environmental Assessment andManagement 5, 86e96.

Giller, P.S., Malmqvist, B., 1998. The Biology of Streams and Rivers, first ed. OxfordUniversity Press, Oxford, UK.

Guy, M., Singh, L., Mineau, P., 2011. Using field data to assess the effects of pesticideson Crustacea in freshwater aquatic ecosystems and verifying the level ofprotection provided by water quality guidelines. Integrated EnvironmentalAssessment and Management.

Harding, J.S., Benfield, E.F., Bolstad, P.V., Helfman, G.S., Jones, E.B.D., 1998. Streambiodiversity: the ghost of land use past. Proceedings of the National Academy ofSciences of the United States of America 95, 14843e14847.

Liess, M., 2002. Population response to toxicants is altered by intraspecific inter-action. Environmental Toxicology and Chemistry 21, 138e142.

Liess, M., Beketov, M., 2011. Traits and stress: keys to identify community effects oflow levels of toxicants in test systems. Ecotoxicology 20, 1328e1340.

Liess, M., Brown, C.D., Dohmen, G.P., Duquesne, S., Hart, A., Heimbach, F., Kreuger, J.,Lagadic, L., Maund, S., Reinert, W., Streloke, M., Tarazona, J.V., 2005. Effects ofPesticides in the Field. SETAC Press, Berlin, Germany.

Liess, M., Von der Ohe, P.C., 2005. Analyzing effects of pesticides on invertebratecommunities in streams. Environmental Toxicology and Chemistry 24,954e965.

Maltby, L., 1999. Studying stress: the importance of organism-level responses.Ecological Applications 9, 431e440.

Matthaei, C.D., Piggott, J.J., Townsend, C.R., 2010. Multiple stressors in agriculturalstreams: interactions among sediment addition, nutrient enrichment and waterabstraction. Journal of Applied Ecology 47, 639e649.

Miyake, Y., Nakano, S., 2002. Effects of substratum stability on diversity of streaminvertebrates during baseflow at two spatial scales. Freshwater Biology 47,219e230.

Montoya, J.M., Woodward, G., Emmerson, M.C., Sole, R.V., 2009. Press perturbationsand indirect effects in real food webs. Ecology 90, 2426e2433.

Paillex, A., Doledec, S., Castella, E., Merigoux, S., 2009. Large river floodplainrestoration: predicting species richness and trait responses to the restoration ofhydrological connectivity. Journal of Applied Ecology 46, 250e258.

Pedersen, M.L., 2009. Effects of channelisation, riparian structure and catchmentarea on physical habitats in small lowland streams. Fundamental and AppliedLimnology 174, 89e99.

Pedersen, M.L., Friberg, N., 2009. Influence of disturbance on habitats and biologicalcommunities in lowland streams. Fundamental and Applied Limnology 174,27e41.

Rasmussen, J.J., Baattrup-Pedersen, A., Larsen, S.E., Kronvang, B., 2011a. Localphysical habitat quality cloud the effect of predicted pesticide runoff fromagricultural land in Danish streams. Journal of Environmental Monitoring 13,943e950.

Rasmussen, J.J., Baattrup-Pedersen, A., Wiberg-Larsen, P., Kronvang, B., 2011b. Bufferstrip width and agricultural pesticide contamination in Danish lowlandstreams: implications for stream and riparian management. Ecological Engi-neering 37, 1990e1997.

Rasmussen, J.J., Friberg, N., Larsen, S.E., 2008. Impact of lambda-cyhalothrin ona macroinvertebrate assemblage in outdoor experimental channels: implica-tions for ecosystem functioning. Aquatic Toxicology 90, 228e234.

Rasmussen, J.J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Kronvang, B., 2012.Impacts of pesticides and natural stressors on leaf litter decomposition inagricultural streams. Science of the Total Environment 16, 148e155.

Relyea, R., Hoverman, J., 2006. Assessing the ecology in ecotoxicology: a review andsynthesis in freshwater systems. Ecology Letters 9, 1157e1171.

Rubach, M.N., Ashauer, R., Buchwalter, D.B., De Lange, H.J., Hamer, M.J.,Preuss, T.G., Töpke, K., Maund, S., 2011. Framework for traits-based assess-ment in ecotoxicology. Integrated Environmental Assessment and Manage-ment 7, 172e186.

Rubach, M.N., Baird, D.J., Van den Brink, P.J., 2010. A new method for ranking mode-specific sensitivity of freshwater arthropods to insecticides and its relationshipto biological traits. Environmental Toxicology and Chemistry 29, 476e487.

Sandin, L., 2009. The effects of catchment land-use, near-stream vegetation, andriver hydromorphology on benthic macro invertebrate communities in a south-Swedish catchment. Fundamental and Applied Limnology 174, 75e87.

Schletterer, M., Fureder, L., Kuzovlev, V.V., Beketov, M.A., 2010. Testing the coher-ence of several macroinvertebrate indices and environmental factors in a largelowland river system (Volga River, Russia). Ecological Indicators 10, 1083e1092.

Schmera, D., 2004. Spatial distribution and coexistence patterns of caddisfly larvae(Trichoptera) in a Hungarian stream. International Review of Hydrobiology 89,51e57.

Schäfer, R.B., Caquet, T., Siimes, K., Mueller, R., Lagadic, L., Liess, M., 2007. Effects ofpesticides on community structure and ecosystem functions in agriculturalstreams of three biogeographical regions in Europe. Science of the Total Envi-ronment 382, 272e285.

Schäfer, R.B., Pettigrove, V., Rose, G., Allinson, G., Wightwick, A., von der Ohe, P.C.,Shimeta, J., Kuhne, R., Kefford, B., 2011. Effects of pesticides monitored withthree sampling methods in 24 sites on macroinvertebrates and microorgan-isms. Environmental Science & Technology 45, 1665e1672.

Singh, V., Singh, D.K., 2009. The effect of abiotic factors on the toxicity of cyper-methrin against the snail Lymnaea acuminata in the control of fascioliasis.Journal of Helminthology 83, 39e45.

Statzner, B., Bis, B., Doledec, S., Usseglio-Polatera, P., 2001. Perspectives for bio-monitoring at large spatial scales: a unified measure for the functionalcomposition on invertebrate communities in European running waters. Basicand Applied Ecology 2, 73e85.

Thrush, S.F., Hewitt, J.E., Hickey, C.W., Kelly, S., 2008. Multiple stressor effectsidentified from species abundance distributions: interactions between urbancontaminants and species habitat relationships. Journal of Experimental MarineBiology and Ecology 366, 160e168.

Tomlin, C.D.S., 2000. The Pesticide Manual, A World Compendium. Crop ProtectionPublications, Farnham, Surrey, UK.

Townsend, C.R., Hildrew, A.G., 1994. Species traits in relation to a habitat templet forriver systems. Freshwater Biology 31, 265e275.

Townsend, C.R., Uhlmann, S.S., Matthaei, C.D., 2008. Individual and combinedresponses of stream ecosystems to multiple stressors. Journal of AppliedEcology 45, 1810e1819.



Urbanic, G., Toman, M.J., Krusnik, C., 2005. Microhabitat type selection of caddisflylarvae (Insecta: Trichoptera) in a shallow lowland stream. Hydrobiologia 541,1e12.

von der Ohe, P.C., Prub, A., Schafer, R.B., Liess, M., de Deckere, E., Brack, W., 2007.Water quality indices across Europe e a comparison of the good ecologicalstatus of five river basins. Journal of Environmental Monitoring 9, 970e978.

Wagenhoff, A., Townsend, C.R., Phillips, N., Matthaei, C.D., 2011. Subsidy-stress andmultiple-stressor effects along gradients of deposited fine sediment and dis-solved nutrients in a regional set of streams and rivers. Freshwater Biology 56,1916e1936.

Wallace, I.D., Wallace, B., Phillipson, G.N., 2003. Keys to the Case-Bearing CaddisLarvae of Britain and Ireland. Freshwater Biological Association.

Paper IV

Science of the Total Environment 416 (2012) 148–155

Contents lists available at SciVerse ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Impacts of pesticides and natural stressors on leaf litter decomposition inagricultural streams

Jes Jesssen Rasmussen a,⁎, Peter Wiberg-Larsen a, Annette Baattrup-Pedersen a,Rikke Juul Monberg b, Brian Kronvang a

a Department of Bioscience, Aarhus University, Vejlsøvej 25, 8600 Silkeborg, Denmarkb Department for Forest and Landscape, University of Copenhagen, Rolighedsvej 23, 1958 Frederiksberg C, Denmark

⁎ Corresponding autho rat: Department of Bioscienc25, 8600 Silkeborg, Denmark. Tel.: +45 87158895.

E-mail address: [email protected] (J.J. Rasmussen).

0048-9697/$ – see front matter © 2011 Elsevier B.V. Alldoi:10.1016/j.scitotenv.2011.11.057

a b s t r a c t
a r t i c l e i n f o
Article history:Received 4 October 2011Received in revised form 11 November 2011Accepted 20 November 2011Available online 15 December 2011

Keywords:MacroinvertebratesMicroorganismsPesticide contaminationLeaf litter decompositionStreamsMultiple stressors

Agricultural pesticides are known to significantly impact the composition of communities in stream ecosys-tems. Moreover, agricultural streams are often characterised by loss of physical habitat diversity which mayimpose additional stress resulting from suboptimal environmental conditions. We surveyed pesticide con-tamination and rates of leaf litter decomposition in 14 1st and 2nd order Danish streams using litter bagswith coarse and fine mesh sizes. Two sites differing in physical habitat complexity were sampled in eachstream, and we used this approach to differentiate the effects of pesticides between sites with uniform (siltand sand) and more heterogeneous physical properties.Microbial litter decomposition was reduced by a factor two to four in agricultural streams compared to for-ested streams, and we found that the rate of microbial litter decomposition responded most strongly to pes-ticide toxicity for microorganisms and not to eutrophication. Moreover, the rate of microbial litterdecomposition was generally 50% lower at sites with uniform physical habitats dominated by soft substratecompared to the sites with more heterogeneous physical habitats. The rate of macroinvertebrate shreddingactivity was governed by the density of shredders, and the density of shredders was not correlated to pesti-cide contamination mainly due to high abundances of the amphipod Gammarus pulex at all sites. Our studyprovides the first field based results on the importance of multiple stressors and their potential to increasethe effect of agricultural pesticides on important ecosystem processes.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Stream ecosystems are impacted by numerous anthropogenicstressors, and they remain some of the most impaired ecosystems onearth in terms of species extinction rates (MEA, 2005). However, one– yet unresolved question – is whether impacted ecosystems are ableto maintain their ecological functions (Dudgeon et al., 2006). This em-phasises the importance of addressing and characterising relationshipsbetween ecosystem function and anthropogenic stress.

Contamination of streams with pesticides applied in normal agri-cultural production is widely acknowledged as one of the greatest an-thropogenic stressors to stream ecosystems (Meybeck, 2003; Schulz,2004). Agricultural pesticides are primarily transported to stream re-cipients via runoff and flow through tile drains which highly coincidewith heavy precipitation events (Kronvang et al., 2002; Neumannet al., 2002; Wauchope, 1978), and agricultural pesticides are pro-posed to pose a threat to all living organisms in stream ecosystems(Liess et al., 2005). Furthermore, streams draining agricultural land

e, Aarhus University, Vejlsøvej

rights reserved.

are often characterised by loss of hydromorphological and habitatheterogeneity due to dredging and channelisation, increased concen-trations of macronutrients, and increased transport of suspended ma-terial (Friberg et al., 2010; Pedersen, 2009). These effects mayadditionally have profound effects on ecosystem structure and func-tion (Pedersen and Friberg, 2009; Piscart et al., 2009; Rasmussenet al., 2011a). In consequence, studies that investigate the combinedeffects of pesticides and other agriculturally related stressors are im-portant in order to sufficiently protect organisms that reside in multi-stressor environments and to improve the environmental realism inthe risk assessment of agricultural streams.

In streams, fungi and bacteria are key organisms in the decompo-sition and conversion of riparian plant litter into more palatable foodresources for macroinvertebrate shredders and collectors/gatherers(Barlocher, 2005; Gessner et al., 2007). Aquatic fungi, mainly hypho-mycetes, are dominant over bacteria during the earlier stages of leaflitter decomposition — even though the relative contribution of thetwo groups may change in severely impacted streams (Duarte et al.,2008a). Increased concentrations of nitrate and phosphate in streamsdue to loadings from agricultural diffuse sources have been shown toincrease fungal biomass and rates of leaf decomposition (Gulis et al.,2006). However, several environmental parameters (e.g. water

http://dx.doi.org/10.1016/j.scitotenv.2011.11.057


http://dx.doi.org/10.1016/j.scitotenv.2011.11.057


149J.J. Rasmussen et al. / Science of the Total Environment 416 (2012) 148–155

temperature) that are related to agricultural production probably actin concert to increase decomposition rates of leaf litter (Friberg et al.,2009a). Conversely, these effects might be counteracted by additionalside-effects of the agricultural practices, i.e. the presence of pesticidesand increased erosion and sediment transport (Young et al., 2008).Microbial organisms are not considered in the risk assessment proce-dure in spite of their important role in stream ecosystems. However,recent studies by Maltby et al. (2009) and Dijksterhuis et al. (2011)conclude that agricultural pesticides do pose a threat to aquaticfungi and that effect concentrations might be below the minimumallowed threshold concentrations for the standard test organisms.Moreover, Schäfer et al. (2011a) documented effects of pesticideson microbial leaf decomposition and fungal biomass in Australianstreams and the authors highlighted the necessity for further re-search to sufficiently protect the microbial decomposing organisms.However, the potentially increased effect of pesticides in the pres-ence of the stress that is related to loss of physical habitat heteroge-neity and quality (e.g. increased siltation, increased amplitude ofoxygen concentrations) on leaf decomposition has not beenstudied.

Macroinvertebrate shredders constitute the other major group oforganisms that facilitate the conversion of leaf litter into secondaryproduction and fine particulate organic matter, and the density ofshredders in streams is tightly coupled to the temporal and spatial ac-cumulation of coarse particulate organic matter (Graca, 2001). Fur-thermore, the food preference of shredders is strongly controlled bythe nutritional quality of the food source, which is governed by the in-ternal nutrient concentrations and the biomass of microbial organ-isms (Graca, 2001). Microcosm studies have indicated thatshredding macroinvertebrates may reduce their consumption of litterafter the exposing of shredder assemblages (Rasmussen et al., 2008)or leaves (Lauridsen et al., 2006) to pyrethroid insecticides. In addi-tion, Schäfer et al. (2007) found that shredder induced leaf massloss in agricultural streams in one of the studied biogeoregions wasprimarily explained by the fraction of species that are sensitive to pe-riodic pesticide pollution (the SPEcies At Risk concept) reflecting thatseveral shredders in the studied streams in addition were charac-terised as SPEAR (Liess and Von der Ohe, 2005). However, these find-ings are contrasted by those of other authors reporting that shredderdensities and shredding activity did not respond to anthropogenicstress related to agriculture (Hagen et al., 2006). Shredder densitiesprobably also respond strongly to physical habitat characteristics asmany species of e.g. Plecoptera and Trichoptera are shredders withnarrow habitat preferences for different types of substrate or vegeta-tion (Dunbar et al., 2010). Thus reduced heterogeneity of physicalhabitats may have the potential to indirectly reduce rates of litter de-composition beyond the effect of pesticides alone.

We conducted field investigations in 14 streams that representeda gradient in agricultural activity. Two sites with differing physicalcomplexity were sampled in each stream to differentiate the effectsof pesticides on microbial decomposition and macroinvertebrateshredding activity between sites with high or low physical habitatheterogeneity. More specifically we tested the hypotheses that1) the rate of microbial litter decomposition decreases with increas-ing pesticide toxicity to microbial leaf decomposers, 2) the rate ofmicrobial leaf decomposition is lower at sites that are characterisedby uniform physical conditions (due to side-effects of agriculture assiltation and previous dredging activity) compared to sites withmore diverse physical conditions 3) shredding activity decreaseswith increasing pesticide toxicity to macroinvertebrates (direct ef-fects) and to microorganisms (indirect effects through decreasingnutritional value of leaves — using microbial activity as a proxy formicrobial biomass and nutritional value), and 4) macroinvertebrateshredding activity is further reduced at sites with homogeneousphysical conditions due to lower density of macroinvertebrateshredders.

2. Materials and methods

2.1. Study area

The field campaign was conducted in the months April to August,2009 in two catchments on the island of Funen (Denmark) that arecharacterised by low elevation and loamy soils with medium tolow infiltration capacity. The climatic conditions are temperateand yearly precipitation averages 700 mm. Agriculture is the domi-nant land use (67 %) with small forest patches and urban areas em-bedded in the landscape. The dominant defoliating tree is beech(Fagus sylvatica). For further details on the study area, consultRasmussen et al. (2011b). In total, fourteen streams of 1st and 2ndorder were surveyed and catchment areas ranged from 18 to 73 ha(Fig. 1).

The stream selection was based on a series of criteria: 1) year-round water flow, 2) no dredging and weed cutting during the sam-pling period, and 3) no sources of pollution other than from agricul-ture. The streams represent a gradient of potential pesticidecontamination predicted from the proportion of adjacent agriculturalland (Rasmussen et al., 2011b).

Two stream reaches (50 m) were selected in each of the studystreams based on the physical properties of the reach (Fig. 2). The cri-teria for selecting the stream reaches were: i) presence of an up-stream reach characterised by low current velocity, low physicaldiversity, low reaeration capacity; ii) a downstream reach charac-terised by fast flow, high physical diversity, and high reaeration ca-pacity; iii) that the distance between paired reaches should notexceed 250 m. The consequent upstream positioning of the reachesthat were characterised by low physical diversity was supposed tominimise the potential bias from accidental downstream drift ofmacroinvertebrates that are strongly associated with habitat typesthat are not present at the reaches with low physical diversity. Forconvenience, reaches are referred to as sampling sites in the followingparts of the paper.

2.2. Physicochemical variables

Three litre water samples were collected at the downstream site ofeach stream (n=14) in each of the months April, June and August forgeneral description of the water chemistry (Table 1). Stream specificwater chemistry is presented in Appendix A. The following parame-ters were analysed according to European standards: BOD5 (DS/EN1899 1999), ortho-phosphate (DS/EN 1189–1997) and ammonia-N(DS 11732 2005). Nitrate-N was analysed using Lachat-methods(Lachat Instruments, USA, Quickchem. no. 10-107-06-33-A (Salycatemethod)). Chloride concentration was measured using silver nitrate(AgNO3) (APHA, 1998). Concentrations of total N and total P weremeasured (unfiltered samples) applying the Kjeldahl-N method(Kjeldahl, 1883) and Danish standard (DS-291), respectively.

At each sampling site, twenty-five transects were establishedwith two meter intervals. In each transect, wetted width (W),depth (D) and water velocity (at 0.6×depth — from the surface)(U) were measured at four points corresponding to 25, 50, 75 and100% of the wetted width using a flow-meter (Höntzsch μP-TAD).Four rectangular plots were established between each pair of tran-sects (2 m×25% of wetted width). In each plot, substrate type andtotal macrophyte coverage were estimated. Submergent and emer-gent macrophytes were identified to the lowest possible taxonomi-cal level. Furthermore, proportional coverage was estimated foreach taxon. Stream water temperature, conductivity and oxygenconcentration were registered using a multi-meter (WTW multi-350i) and pH was measured applying a (YSI-60) pH-meter(Table 1). Sampling site specific physical characteristics are pre-sented in Appendix B.

Fig. 1. Geographical distribution of the study streams.

Table 1

150 J.J. Rasmussen et al. / Science of the Total Environment 416 (2012) 148–155

2.3. Quantification of pesticide contamination

In total, 19 herbicides, 6 fungicides and 6 insecticides were includ-ed in the sampling program. The selected list of pesticides repre-sented the most frequently applied compounds in Danishagriculture in 2008 supplemented by a series of banned pesticidesthat are commonly found in groundwater. The methods used for

Fig. 2. Schematic presentation of the paired sites representing uniform (upstream) anddiverse (downstream) physical habitats. Each sampling site represents a 50 m reach.Current velocity is used as a surrogate measure for the spatial distribution of physicalhabitat types as current velocity has a direct effect on substrate types, turbulence andreaeration.

pesticide sampling and quantification have been published elsewhere(Rasmussen et al., 2011b).

Pesticide contamination during heavy precipitation events (repre-senting runoff and flow through tile-drains) was characterised usingevent triggered water samplers during the months May and June

Average, standard deviation, minimum and maximum values for physical and chemicalparameters measured in 14 Danish streams. Dashes indicate that values were belowdetection limit.

Parameter Average Standarddeviation

Minimum Maximum

Agriculture in catchment (%)b 51.7 38.8 2.2 100Minimum buffer strip width (m)b 7.9 7.8 1 25PhysicalDischarge (L s−1)a 3.8 3.7 – 13.1Current velocity (m s−1)a 0.04 0.06 – 0.16Shear stressa 5.9 2.9 – 11Width (m)a 1.3 0.5 0.55 2.5Depth (m)a 0.06 0.05 0.02 0.29Boulder (%)a 9.1 9.2 0 33.5Gravel (%)a 26.8 21.5 0 66.7Sand (%)a 21.3 14.6 0 58.0Silt (%)a 33.2 24.7 3.3 89.3Debris (%)a 7.3 7.7 0.6 35.3# Macrophyte speciesa 2.6 1.5 0 6Total macrophyte coverage (%)a 17.8 15.2 0 67Temperature (°C)a 13.3 1.7 10.9 16.7ChemicalBOD5 (mg L−1)a 1.91 0.62 0.96 3.05Oxygen concentration (mg L−1)a 8.25 1.04 5.94 10.21Ammonia (mg L−1)a 0.07 0.06 0.01 0.19Nitrate (mg L−1)a 2.27 1.57 0.59 6.66Total N (mg L−1)a 2.9 1.61 1.19 7.48Ortho-phosphate (mg L−1)a 0.07 0.04 0.02 0.14Total P (mg L−1)a 0.13 0.05 0.06 0.24Chloride (mg L−1)a 30.73 4.53 22.58 38.68Sulphate (mg L−1)a 49.77 19.93 28.63 75.25Suspended matter (mg L−1)a 9.63 4.40 3.80 19.57

a Parameter measured on a reach level (n=28).b Parameter measured on a stream level (n=14).

image of Fig.�2


(Liess and Von der Ohe, 2005). Event triggered samplers weredeployed just downstream of the downstream sampling site. Thus,sample represents both sampling sites as it was assumed that the pes-ticide contamination was not significantly different between sites.The sampling system consisted of two 1 L glass bottles that weredeployed in the flowing part of the stream channel. Bottles were filledpassively through small (0.5 cm in diameter) glass tubes when thewater level increased above the glass tube opening. The two bottleswere positioned 5 cm and 10 cm above base flow water level, respec-tively. During the sampling period, two precipitation episodes triggeredthe sampling system of which the last was the only episode triggeringthe samplers in all streams. In addition, we collected water samples(1 L) in August during low flow conditions to characterise pesticideinput that originated from groundwater inflow. However, these resultsare disregarded in this paper since the groundwater input of pesticideswas characterised by only few herbicide compounds in low concentra-tions. See also Rasmussen et al. (2011b) for further details.

Retrieved samples were all sent to the OMEGAM laboratories inAmsterdam for pesticide analyses (including sample filtering andsolid phase extraction). The final extract of each sample was pro-cessed using gas-chromatography mass-spectrometry (GC–MS) orliquid-chromatography mass-spectrometry (LC–MS). Results werecorrected for recovery, which was determined by Omegam laborato-ries. For further details consult Rasmussen et al. (2011b).

2.4. Macroinvertebrates

Macroinvertebrates were sampled in April, June and August, 2009using randomised surber sampling (each sample covering 200 cm2).Five surber samples were collected from each sampling site (n=28)in each sampling month and samples were preserved in 70% ethanolin the field. All macroinvertebrates were counted and identified togenus or species level, except Chironomidae (sub-family), Simuliidae(family) and Oligochaeta (order) (Appendix C). Feeding functionalgroups were assigned to each taxon according to Tachet et al.(2002), and the density of shredders was calculated. In addition, theabundance of SPEcies At Risk for being impacted by periodic pesticidecontamination (%SPEAR abundance) was computed for each site andsampling using the freely available online SPEAR calculator (http://www.systemecology.eu/SPEAR/index.php) (Appendix C).

2.5. Leaf decomposition

Newly fallen beech leaves (F. sylvatica) were used for studying leafdecomposition. The leaves were retrieved from the forest floor in Vellingforest, Denmark in the autumn of 2008. The forest catchment is entirelycharacterised by old beech forest, and collected leaves were not sup-posed to be exposed to significant environmental pollutants prior tothe study. The collected leaves were air-dried in a climate chamber at15 °C and subsequently stored at 4 °C.

OnMay 1st wemounted 15 leaf bags with finemesh size (500 μm)and 15 leaf bags with coarse mesh size (9 mm) to the stream bed ateach sampling site. Riffle sections were preferred when possible tominimise siltation and because litter material primarily accumulatesat these sites. However, none of the sites that were characterised byuniform physical conditions contained distinct riffle sections, andthe leaf bags were deployed on the hardest substrate that was avail-able at the sampling site. Leaf bags with fine mesh size contained0.38±0.02 g DW, and leaf bags with coarse mesh size contained0.75±0.02 g DW. Only unbroken leaves were used. However, thepetiole was removed from all leaves to increase the signal of biologi-cal decomposition processes. Three leaf bags with fine and coarsemesh size, respectively, were sampled after 14, 21, 35, 49 and80 days. Sediment material and macroinvertebrates were removedin the laboratory by gently rinsing the leaf packs in demineralisedwater in white trays and subsequently the leaves were dried at

60 °C to constant mass (48–72 h). The dry weight of the remainingleaf material in each leaf bag was measured (three decimals accuracy)using a Mettler Toledo XP-204. Leaf pack weight loss reflects microbi-al activity and leaching (fine mesh bags) whereas the weight loss incoarse mesh bags additionally reflects shredding activity andabrasion.

2.6. Data analysis

The remaining dry mass of beech leaves was fit to the exponentialdecay model mt=m0∗e−kt where mt is the leaf dry mass remainingat time t, m0 is the initial dry mass, and k (day−1) is the rate of leafdecomposition. The physical leaching of soluble compounds may ac-count for 4 to 42% of the total leaf biomass, and this mass loss willbe effectuated within 2 to 7 days after leaf submergence (Abelho,2001). In consequence, we used the leaf bags that were retrievedafter 14 days as initial dry mass to minimise the noise from physicalcompound leaching. Shredder decomposition rates (kshreddersday−1) (coarse meshed bags) were corrected for microbial decompo-sition (kmicrobial day−1).

The toxicity of the measured pesticide concentrations for macroin-vertebrates and microbial organisms was quantified using Toxic Units(TU)

log TU ¼ logðCi=EC50iÞ ð1Þ

where Ci is the concentration of the compound i and EC50i is themedianacute effect concentration of a standard test species. Daphnia magnawas selected as standard test species for macroinvertebrates. Sincethere is no standard test species available for microorganisms, weused species of aquatic hyphomycetes when available and Selenastrumcapricornutum for the remaining compounds as surrogate microorgan-isms to estimate the potential toxic effect on microbial leaf decom-posers (log TUmicrobial) (Schäfer et al., in press). The summed and thelog maximum TU were calculated for all samples for both D. magnaand microorganisms. We only report log maximum TUs (log mTU) asthe log sum TUs and log mTUs were highly intercorrelated forD. magna and microorganisms (r>0.85, Pb0.001). Furthermore, apply-ing the log sum TUs did not significantly improve relationships with bi-otic endpoints compared to the log mTUs which are consistent withprevious findings (Schäfer et al., 2007; Schäfer et al., in press).

In order to identify if the experimental set-up was successful inseparating the two agriculturally related stressors of pesticide con-tamination and loss of physical habitat heterogeneity, a PrincipalComponent Analysis (PCA) of all physical and chemical variableswas performed using PC-ORD 4.25. Four groups were identified inthe PCA, and the groups were identified as significantly differentusing the site scores in a one-way analysis of variance.

We tested for differences between average kmicrobial and kshreddersfor each of the four groups of sampling sites that were identifiedwith the PCA using a two-way analysis of variance. We used Pearson'scorrelation analyses to evaluate the effect of all measured environ-mental parameters and shredder metrics on site specific kmicrobial

and kshredders in SAS Enterprise Guide 4.2. Analysis of covariance wasused to identify potential differences in slopes and intercepts of theregression lines that characterised log mTU (for microorganismsand D. magna) as a function of k (for microorganisms and shredders,respectively). In addition to the kshredders we applied kshredders thatwas normalised according to the population density of shredders toevaluate if there was an effect at the level of individuals. Analyses ofcovariance were conducted in Sigma-Plot 11.0.

The potential indirect effect of agricultural pesticides on shreddingactivity (through a reduction in nutritional value of the litter) wastested using a Spearman Rank correlation analysis on the relation be-tween kmicrobial and kshredders. Moreover, we tested if the relative pro-portion of agriculture in the catchment had an effect on the ratio




(kmicrobial/kshredders) using Spearman Rank correlation. All correlationanalyses were performed in SAS Enterprise Guide 4.2.

3. Results

3.1. Pesticides and TU

In total 13 herbicides, 5 fungicides and 2 insecticides were detectedin the streamwater samples (Table 2). Sumconcentrations ranged from0.01 to 3.17 μg L−1, and the number of detected pesticides per watersample ranged from 1 to 13. The log mTUD.magna ranged from −6.63to−1.72, and logmTUmicrobial ranged from−4.03 to−0.63. The carba-mate insecticide pirimicarb was clearly the most important contributorto high values of log mTUD.magna followed by the strubilurine fungicideazoxystrobin. The fungicides dimethomorph (morpholine) and propi-conazole (triazole) and the dinitroaniline herbicide pendimethalinwere the most important compounds governing high values of logmTUmicrobial.

3.2. Physico-chemical parameters

Several environmental parameters were significantly intercorre-lated (P|>0.05), and in these cases we selected the environmentalparameter that wasmost strongly correlated to kmicrobial and kshredders.We used log mTUmicrobial as measure for pesticide contamination, be-cause this parameter was clearly more strongly correlated with envi-ronmental parameters than mTUD.magna and we pooled the differenttypes of hard substrate (gravel and boulder) in one category to in-crease overall trends. The Principal Component Analysis (PCA) ofphysical and chemical variables was conducted using a strep-wiseforward selection excluding insignificant environmental parameters.The final PCA was conducted using the proportion of silt, chlorideconcentration, log mTUmicrobial, discharge, the proportion of debris,discharge and BOD5 as environmental parameters. Axis 1, explaining50.5% of the variation in the data set, separated the sites with forestedcatchments (>50% forest) from those with agricultural catchments

Table 2Detected pesticide compounds in stream water of 14 Danish streams. Three sampleswere collected from each stream of which two were samples with event-triggeredsamplers during May and June and one was sampled manually during base-flow inAugust.

Compound Maxconcentration(ug/L)

Detectionfrequency(%)

MaximumTUD.magna

aMaximumTUmicrobial

Aclinofen (H) 0.14 7 −3.93 −1.54c

Atrazine (H) 0.02 7 −6.63 −3.81c

Desethylterbutylazine (H) 0.11 100 −4.65 −2.99c

Diflufenican (H) 0.15 29 −3.20 −2.08c

Hexazinone (H) 0.06 7 −6.15 −2.4c

Metachlor (H) 0.05 57 −5.82 −2.2c

Metamitron (H) 0.12 7 −4.68 −2.52c

Pendimethaline (H) 0.97 14 −2.46 −0.92c

Propyzamide (H) 0.43 21 −4.11 −3.23c

Prosulfocarb (H) 0.07 21 −3.86 −3.21c

Simazine (H) 0.03 7 −4.56 −3.12c

Terbutylazine (H) 0.6 57 −4.55 −2.26c

Azoxystrobin (F) 0.51 43 −2.77 −1.92b

Boscalid (F) 0.72 36 −3.87 −2.82c

2,4 Dichlorophenoxyaceticacid (H)

0.05 7 −6.18 −3.71c

Dimethomorph (F) 0.08 14 −5.12 −0.63b

Propiconazole (F) 0.27 43 −4.58 −0.88c

Tebuconazole (F) 0.24 50 −4.24 −2.99b

Dimethoate (I) 0.18 14 −4.05 −6.11c

Pirimicarb (I) 0.32 21 −1.72 −5.64c

a Based on the LC50 derived from 48 h acute toxicity tests.b Based on species of aquatic hyphomycetes.c Based on Selenastrum capricornosum.

(>50% agriculture). Axis 2, additionally explaining 28.5% of the vari-ation in the data set, separated the upstream sites with uniform phys-ical conditions from the downstream sites with more diverse physicalconditions. Positive values on axis 2 are characterised by higher pro-portions of boulder and gravel whereas negative values are charac-terised by higher proportions of silt. The PCA separated thesampling sites into four significantly different groups according toland-use characteristics and physical habitat properties (Pb0.05)(Fig. 3).

3.3. Leaf decomposition

Regression coefficients for the fitted exponential decay functionsaveraged 0.71±0.22 and 0.54±0.21 for leaf bags with fine andcoarse mesh sizes, respectively (data not shown). Relative weightloss ranged from 46 to 100% in leaf bags with coarse mesh size andfrom 0 to 52% in leaf bags with fine mesh size.

The average rate of microbial leaf decomposition (kmicrobial) was sig-nificantly different amongst the site groups that were identified usingPCA on environmental variables (Fig. 4) (Pb0.05). The average kmicrobial

for forested sites was significantly higher than the average kmicrobial foragricultural sites (Pb0.001) and the average kmicrobial for forested siteswith diverse physical conditions (kmicrobial=0.0084 day−1, n=5) wassignificantly higher than the average kmicrobial for forested sites withuniform physical conditions (kmicrobial=0.0047 day−1, n=5)(Pb0.05). However, the average kmicrobial for agricultural sites with di-verse physical conditions (kmicrobial=0.0026 day−1, n=9)was not sig-nificantly different from the average kmicrobial for agricultural sites withuniform physical conditions (kmicrobial=0.0016 day−1, n=9,P=0.061). The average rate of shredder decomposition (kshredders) forthe respective stream sites was not significantly different amongst thesite groups that were identified using PCA on environmental variables(P>0.05, Appendix D). Normalising the average rate of shredder de-composition to the level of individuals did not provide significantly dif-ferent results (data not shown).

We found that the most important environmental parametergoverning the kmicrobial was log mTUmicrobial (r=−0.72, Pb0.001)followed by the log mTUD.magna (r=−0.63, Pb0.001), the relative pro-portion of agriculture in the catchment (r=−0.58, Pb0.001) and thelogarithm to the total concentrations of pesticides (r=0.56,P=0.025) (Appendix E). On average, the concentrations of ammonia–N and nitrate-N were two times higher in agricultural streamswhereasthere were no significant differences in concentrations of ortho-phosphate. However, no macro- or micronutrient concentrations weresignificantly correlated with rates of microbial or shredder decomposi-tion. Subdividing the sites according to their physical properties furtherincreased the correlation coefficient for the regression between kmicrobial

80

60

40

20

0

0

Principal Component Axis 1

Forest, uniform phys. cond.

Forest, diverse phys. cond.

Agriculture, uniform phys. cond.

Agriculture, diverse phys. cond.

-20

-40

-60

Pri

nci

pal

Co

mp

on

ent

Axi

s 2

80604020-20-40-60-80

Fig. 3. Principal Component Analysis (PCA) of physico-chemical parameters for 28reaches (50 m). Loading of PCA axis 1=50.5% and loading of PCA axis 2=28.5%.

image of Fig.�3

0,012

A

B

C

C

Diverse physical conditionsUniform physical conditions

0,010

0,008

k mic

robi

al (

day

-1)

0,006

0,004

0,002

0,000Forest Agriculture

Fig. 4. Rates of microbial leaf decomposition (kmicrobial) in 14 streams. Data representstwo sampling sites in each of the streams and were characterised by diverse or uniformphysical conditions, respectively. Grouping of sampling sites was based on a PCA on en-vironmental variables (see Fig. 3). Error bars indicate SE of the mean, and capital lettersindicate significantly different groups (Pb0.05).

Table 3Shredder densities for 14 paired stream sites that differ with respect to diversity of thephysical habitats (±SE). Shredder densities are based on 5 surber samples collectedfrom each site in the months April, June and August, 2009. Furthermore, the Pearsonr is presented for the correlations between the rate of leaf decomposition ascribed toshredders (kshredders) and shredder densities for all sites in each of the samplingmonths. Asterisks indicate the level of significance.

All sites Sites withuniformphysicalconditions

Sites withdiversephysicalconditions

Pearson r forkshredders(all sites)

Shredder density (April)(ind. m−2)

2794±402 2918±571 2669±598 0.47⁎

Shredder density (June)(ind. m−2)

5693±1429 7071±2696 4313±960 0.52⁎⁎

Shredder density(August) (ind. m−2)

4515±694 4413±1150 4616±827 0.52⁎⁎

Average shredderdensity (ind. m−2)

4226±716 4696±1335 3756±559 0.61⁎⁎

⁎ Pb0.05.⁎⁎ Pb0.01


and log mTUmicrobial (r=−0.76, Pb0.001 and r=−0.71, Pb0.005, forsites with diverse and uniform physical conditions, respectively)(Fig. 5). There was no significant difference between the slopes of theregression lines (P>0.05) representing kmicrobial as a function of logmTUmicrobial, but the intercepts were significantly different (Pb0.05)(Fig. 5).

The rate of leaf decomposition in the leaf bags with shredders(kshredders) was significantly correlated to the site-based individualdensity of shredders and the proportional coverage of silt (Table 3).However, shredder densities for April, June and August were not corre-lated with the proportional coverage of silt (r=0.15, P>0.05, n=28).The log mTUD.magna was not significantly correlated with kshredders(P>0.05, Appendix E, F). Normalising rates of shredder decompositionto the level of individuals did not provide significant correlations to pes-ticide parameters. Similarly, subdividing the sites according to theirphysical properties and/or land-use characteristics did not provide sig-nificant correlations with pesticide parameters (data not shown). Thefreshwater amphipod Gammarus pulexwas clearly the dominating spe-cies of shredding macroinvertebrates in terms of individual density onaverage accounting formore than 75%of the total shredder density (Ap-pendix C).

The ratio between kmicrobial and kshredders decreased with increasedagricultural activity in the catchment (r=0.72, Pb0.001) (Fig. 6). Theratio between kmicrobial and kshredders additionally decreased

0,0025

0,0020

0,0015R2 = 0.58

Uniform physical conditionsDiverse physical conditions

R2 = 0.50

k mic

robi

al (

day

-1)

0,0010

0,0005

0,0000

-4 -3 -2

mTUmicrobial

-1 0

Fig. 5. The rate of microbial leaf decomposition (kmicrobial) as a function of the log mTU-microbial measured in 14 Danish streams with event triggered samplers. Data representstwo neighbouring sites from each streamwith the upstream site being characterised byuniform physical habitats and the downstream site being characterised by diversephysical habitats.

significantly with increasing log mTUD.magna (r=0.61, Pb0.01) andlog mTUmicrobial (r=0.45, Pb0.05) (data not shown).

4. Discussion

4.1. Impacts of pesticides on microbial leaf decomposition

The rate of microbial decomposition of beech leaves (F. sylvatica)was two to four times higher at forested sites compared to agriculturalsites, and the most important environmental parameter that governedthe variation in rates of microbial decomposition was pesticide toxicityfor microorganisms (log mTUmicrobial) followed by pesticide toxicity formacroinvertebrates (log mTUD.magna) and the proportion of agriculturein the catchment. The microbial decomposition of leaf litter in agricul-tural streams has been investigated in several studies (Gessner andChauvet, 1994; Hagen et al., 2006; Magbanua et al., 2010; Piscartet al., 2009). However, only two of these quantified the pesticide expo-sure (Schäfer et al., 2011, 2007) reporting a 2.5–4 fold reduction in leafdecomposition at agricultural sites compared to forested sites which isvery comparable to our results. In contrast, the increasing eutrophica-tion that is associated with agricultural production has been shown toincrease microbial diversity, biomass and rates of decomposition(Hagen et al., 2006; Magbanua et al., 2010; Pascoal et al., 2003). Our re-sults therefore accentuate that pesticides have an important impact onthe microbial communities clearly over-ruling the well-documentedand stimulating effects of eutrophication, and our results are supported

0.6Uniform physical conditionsDiverse physical conditions

0.5

0.4

0.3

0.2

0.1

0.00 20 40 60

% Agriculture80 100

k mic

robi

al (

day

-1)

k shr

edde

rs (

day

-1)

R2 = 0.54

/

Fig. 6. The ratio between rates of microbial leaf decomposition (kmicrobial) and shredderbased leaf decomposition (kshredders) as a function of the proportion of agriculture inthe catchment. The data represents14 paired sites differing in the diversity of physicalconditions.


by a recent study by Schäfer et al. (2011a). Estimating pesticide toxicityto microbial decomposers is still somewhat ambiguous as none of themicrobial decomposers are included as standard test organisms in therisk assessment program and thus there is only very limited toxicitydata available (Dijksterhuis et al., 2011). Since the microbial decompo-sition is dominated by aquatic fungi (Duarte et al., 2008a, 2008b), thesupplementation of toxicity data for aquatic fungi is highly importantto assess the risk of pesticide impacts on this ecosystem process.

4.2. Physical habitat properties influence the effect of pesticides onmicrobialleaf decomposition

The response of microbial decomposition to increasing pesticidestress (slopes of regression lines) was similar for the two groups ofsites differing in physical habitat characteristics, but the rate of microbialleaf decomposition was consistently approximately 50% lower at siteswith degraded and uniform physical properties compared to sites withmore undisturbed and heterogeneous physical properties. Sites with de-graded and uniform physical conditions were characterised by low cur-rent velocity and high proportions of sand and silt which probablyreflect higher rates of sedimentation. Rates of microbial decompositionmay decrease as a direct consequence to sedimentation, because themi-croorganisms become curtailed in their ability to use external nutrientsources and oxygen (Barlocher, 2005; Imberger et al., 2010; Medeiroset al., 2009). The effects of reduced oxygen availability are probably fur-ther promoted by increased thickness of the boundary layer surroundingthe leaf litter (Friberg et al., 2009b).

4.3. Macroinvertebrate shredding activity

Rates of leaf decomposition in leaf bags with macroinvertebrateswere primarily correlated to the density of shredding macroinverte-brates. Surprisingly however, the density of shredding macroinverte-brates did not change significantly along the gradient of agriculturalintensity in the catchment and, moreover there were no significantdifferences in shredder densities between sites of differing physicalproperties. This was mainly due to very high abundances of the fresh-water amphipod, G. pulex that is generally the dominating taxon inFunen streams (Wiberg-Larsen et al., 1997). This species is very sen-sitive to insecticide contamination and has been shown to respond toshort pulses of pyrethroids down to 10 ng L−1 (corresponding toTUD.magna=−1.6) in micro- and mesocosms (Nørum et al., 2010;Rasmussen et al., 2008). However, G. pulex also has a relativelyhigh migration and reproduction potential (Liess and Von der Ohe,2005), which are ecological traits that may compensate for the highsensitivity to pesticides. On the level of populations this species maytherefore be rather insensitive to recurring pesticide exposure (Liessand Von der Ohe, 2005). Furthermore, G. pulex is an omnivorous spe-cies reflecting its additional ability to act as a predator and even as agrazer (Marchant, 1981; Tachet et al., 2002). The combination of di-verse food preferences and ecological traits entails that G. pulex hassome important competitive advantages in frequently disturbed agri-cultural streams making it an essential organism for the processing oforganic matter in streams in Denmark. However, since pulse exposureof insecticides has been shown to change the size structure of D.magna through several generations (Liess and Foit, 2010; Liess et al.,2006), similar effects are likely to occur for populations of gammaridsin the field. Applying shredder biomass as supplement to shredderdensity would have provided useful information on the size structureof the populations of gammarids in this study.

Normalising shredder decomposition to the level of individualsdid not provide significant correlations with pesticide parameterssuggesting that indirect effects in terms of reduced feeding rateswere not detectable. These findings are contrasted by results fromstudies using simpler set-ups showing clear indirect effects of insecti-cides on Gammaridae feeding rates induced by directly exposing

macroinvertebrates (Rasmussen et al., 2008) or exposing food items(Bundschuh et al., 2011). However, our study may underestimate the in-direct effects of pesticides on shreddingmacroinvertebrates since leaf lit-ter was not very abundant in the agricultural streams. Thus, submergedleaves could be subject to amassive feeding pressure if present gammar-ids preferred the submerged litter over their normal food sources.

Numerous field studies have been conducted investigating relation-ships between shredding activity and various sources of anthropogenicstress, but in terms of conventional agriculture the results are contrasting.A series of studies report that macroinvertebrate shredding activity wasprimarily explained by the density of shredders (Ferreira et al., 2006;Gessner and Chauvet, 2002; Hagen et al., 2006)whereas other studies re-port that rates of macroinvertebrate induced leaf decomposition congru-ently responded to the intensity of anthropogenic stress (Pascoal et al.,2003; Young and Collier, 2009). Piscart et al. (2009) reported thatmacro-invertebrate induced leaf decomposition increasedwith increasing densi-ty of Gammaridae and decreased with increasing agriculture in thecatchment. Possibly, the density of Gammaridae decreased in responseto high pesticide concentrations, but unfortunately pesticide concentra-tions were not measured. In contrast, Schäfer et al. (Schäfer et al., 2007)found that shredding activity was significantly correlated to the relativeabundance of SPEcies At Risk (indicating direct effects of pesticides onmacroinvertebrates) in French streams inwhichGammaridae additional-ly occurred in high densities. However, the French streams represented alarger gradient of toxic pesticide pressure with measured mTUD.magna upto −0.5, which might have suppressed the ecological function of Gam-maridae in contaminated streams (Rasmussen et al., 2008).

4.4. Macroinvertebrate decomposers and nutritional food quality of leaves

The relationship between rates of microbial and shredder leaf de-composition decreased with increasing proportions of agriculture inthe catchment. However, the change was primarily facilitated by re-duced microbial decomposition at agricultural sites as shredding ac-tivity was not significantly different amongst sites. Applying rates ofmicrobial decomposition as a proxy for microbial biomass(Barlocher, 2005) our results indicate that the reduced palatabilityand nutritional value of leaves that is a consequence of reduced mi-crobial biomass and activity had no effect on the total shredderbased leaf mass loss. In existing literature it is consistently reportedthat shredding macroinvertebrates selectively choose highly palat-able leaves with high nutritional value (Abelho, 2001; Graca, 2001).Shredding activity has been shown to decrease when either leaves(Bundschuh et al., 2011; Lauridsen et al., 2006) or macroinvertebrates(Rasmussen et al., 2008) are exposed to insecticides or fungicides.However, applied exposure concentrations in the studies ofBundschuh et al. (2011) and Lauridsen et al. (2006) were above envi-ronmental realistic concentrations and are not necessarily represen-tative for actual field scenarios. We therefore suggest that morecontrolled studies using environmentally realistic exposure concen-trations and durations are needed to provide the mechanistic knowl-edge about howmicrobial communities are affected and whether thischange has potential to affect macroinvertebrate shredders.

5. Conclusions

We showed that diffuse source agricultural pesticides significantlyreduced the rate of microbial leaf decomposition in agriculturalstreams and that pesticide toxicity for aquatic microorganisms wasthe primary reason for this. Based on our results we therefore proposethat it is essential to consider freshwater microorganisms in risk as-sessment and that there is an urgent need for generating relevanttoxicity data for the organisms that are at risk in the field.

An important result for the management of streams was that thequality and heterogeneity of physical habitats influence the actual effectof pesticides on microbial leaf processing emphasising the importance


of considering the physical conditions of streams in risk assessment andmitigation.

Despite some evidence from single-organism and assemblage basedmicro- and mesocosm studies showing that pesticide exposure cancause indirect effects in terms of reduced feeding rates formacroinverte-brates, our results confirm that such effects are difficult to identify incomplex and natural ecosystems as a result of functional redundancyand local variation in shredder abundances. However, we suggest apply-ing other and probably more sensitive endpoints like size structure andtotal biomass of shredder populations to evaluate the potential effectsof pesticides on shredding macroinvertebrates in the field.

Supplementary data related to this article can be found online atdoi:10.1016/j.scitotenv.2011.11.057.

Acknowledgements

This study was financed by the Danish Research Council (grantno. 2104-07-0035) and is part of the RISKPOINT program. We wish tothank Johnny Nielsen for assisting in the counting and identification ofmacroinvertebrates. Additionally, we wish to thank Henrik Stenholt,Uffe Mensberg, Dorte Nedergaard and Marlene Venø Skjærbæk for in-valuable field assistance. Søren Erik Larsen provided help with the SASstatistical software package. In addition, we thank three anonymous re-viewers for very constructive comments that significantly improved themanuscript.

References

Abelho M. From litterfall to breakdown in streams: a review. Sci World 2001;1:656–80.APHA. Standard methods for the examination of water and wastewater. Washington,

DC: American Public Health Association; 1998.Barlocher F. Freshwater fungal communities. In: Deighton J, Oudemans P,White J, editors.

The fungal communitiy: its organization and role in the ecosystem. Florida, UnitedStates: CRC Press; 2005. p. 39–59.

BundschuhM, Zubrod JP, Kosol S, Maltby L, Stang C, Duester L, et al. Fungal composition onleaves explains pollutant-mediated indirect effects on amphipod feeding. Aquat Toxicol2011;104:32–7.

Dijksterhuis J, van Doorn T, Samson R, Postma J. Effects of seven fungicides on non-targetaquatic fungi. Water Air Soil Pollut 2011;222:421–5.

Duarte S, Pascoal C, Alves A, Correia A, Cassio F. Copper and zinc mixtures induce shiftsin microbial communities and reduce leaf litter decomposition in streams. FreshwBiol 2008a;53:91-101.

Duarte S, Pascoal C, Cassio F. High diversity of fungi may mitigate the impact of pollutionon plant litter decomposition in streams. Microb Ecol 2008b;56:688–95.

Dudgeon D, Arthington AH, Gessner MO, Kawabata ZI, Knowler DJ, Leveque C, et al.Freshwater biodiversity: importance, threats, status and conservation challenges.Biol Rev 2006;81:163–82.

Dunbar MJ, Warren M, Extence C, Baker L, Cadman D, MouldDJ, et al. Interaction betweenmacroinvertebrates, discharge and physical habitat in upland rivers. Aquat Conserv-MarFreshw Ecosyst 2010;20:S31–44.

Ferreira V, Gulis V, GracaMAS.Whole-stream nitrate addition affects litter decompositionand associated fungi but not invertebrates. Oecologia 2006;149:718–29.

Friberg N, Dybkjaer JB, Olafsson JS, Gislason GM, Larsen SE, Lauridsen TL. Relationshipsbetween structure and function in streams contrasting in temperature. Freshw Biol2009a;54:2051–68.

Friberg N, Sandin L, PedersenML. Assessing the effects of hydromorphological degradationon macroinvertebrate indicators in rivers: examples, constraints and outlook. IntegrEnviron Assess Manag 2009b;5:86–96.

Friberg N, Skriver J, Larsen SE, Pedersen ML, Buffagni A. Stream macroninvertebrateoccurrence along gradients in organic pollution and eutrophication. Freshw Biol2010;55:1405–19.

Gessner MO, Chauvet E. Importance of stream microfungi in controlling breakdownrates of leaf-litter. Ecology 1994;75:1807–17.

Gessner MO, Chauvet E. A case for using litter breakdown to assess functional streamintegrity. Ecol Appl 2002;12:498–510.

GessnerMO,Gulis V,KuehnKA,Chauvet E, SuberkroppK. Fungal decomposers of plant litterin aquatic ecosystems. In: Kubicek CPaD IS, editor. Environmental and microbialrelationships. 2nd edition½. Berlin: Springer Verlag; 2007.

Graca MAS. The role of invertebrates on leaf litter decomposition in streams — a review.Int Rev Hydrobiology 2001;86:383–93.

Gulis V, Ferreira V, GracaMAS. Stimulation of leaf litter decomposition and associated fungiand invertebrates by moderate eutrophication: implications for stream assessment.Freshw Biol 2006;51:1655–69.

Hagen EM, Webster JR, Benfield EF. Are leaf breakdown rates a useful measure ofstream integrity along an agricultural landuse gradient? J North Am BenthologicalSoc 2006;25:330–43.

Imberger SJ, Thompson RM, Grace MR. Searching for effective indicators of ecosystemfunction in urban streams: assessing cellulose decomposition potential. FreshwBiol 2010;55:2089–106.

Kjeldahl J. A new method for the determination of nitrogen in organic matter. Zeits-chreft fur Analytische Chemie 1883;22:366.

Kronvang B, Strom HL, Hoffmann CC, Laubel A, Friberg N. Subsurface tile drainage lossof modern pesticides: field experiment results. Water Sci Technol 2002;49:139–47.

LauridsenRB, Kronvang B, FribergN. Occurrence of sediment-bound pyrethroids in danishstreams and their impact on ecosystem function.Water Air Soil Pollut 2006;6:423–32.

Liess M, Foit K. Intraspecific competition delays recovery of population structure. AquatToxicol 2010;97:15–22.

Liess M, Von der Ohe PC. Analyzing effects of pesticides on invertebrate communities instreams. Environ Toxicol Chem 2005;24:954–65.

LiessM, Brown CD, DohmenGP, Duquesne S, Hart A, Heimbach F, et al. Effects of pesticidesin the field. Berlin, Germany: SETAC press; 2005.

Liess M, Pieters BJ, Duquesne S. Long-term signal of population disturbance after pulseexposure to an insecticide: rapid recovery of abundance, persistent alteration ofstructure. Environ Toxicol Chem 2006;25:1326–31.

Magbanua FS, Townsend CR, Blackwell GL, Phillips N, Matthaei CD. Responses of streammacroinvertebrates and ecosystem function to conventional, integrated and organicfarming. J Appl Ecol 2010;47:1014–25.

Maltby L, Brock TCM, van den Brink PJ. Fungicide risk assessment for aquatic ecosystems:importance of interspecific variation, toxic mode of action, and exposure regime.Environ Sci Technol 2009;43:7556–63.

Marchant R. The ecology of Gammarus in runningwater. In: LockMA,Williams DD, editors.Perspectives in running water ecology. New York, USA: Plenum Press; 1981. p. 225–50.

Medeiros AO, Pascoal C, Graca MAS. Diversity and activity of aquatic fungi under lowoxygen conditions. Freshw Biol 2009;54:142–9.

MeybeckM. Global analysis of river systems: from Earth system controls to Anthropocenesyndromes. Philos Trans R Soc B 2003;358:1935–55.

Millenium Ecosystem Assessment. Ecosystems and human well-being: synthesis.Washington, DC: Island Press, 2005.

Neumann M, Schutz R, Schafer K, Muller W, Mannheller W, Liess M. The significance ofentry routes as point and non-point sources of pesticides in small streams. WaterRes 2002;36:835–42.

NørumU, Friberg N, JensenM, Pedersen J, Bjerregaard P. Behavioural changes in three spe-cies of freshwatermacroinvertebrates exposed to the pyrethroid lambda-cyhalothrin:laboratory and stream microcosm studies. Aquat Toxicol 2010;98:328–35.

Pascoal C, Pinho M, Cassio F, Gomes P. Assessing structural and functional ecosystemcondition using leaf breakdown: studies on a polluted river. Freshw Biol2003;48:2033–44.

PedersenML. Effects of channelisation, riparian structure and catchment area on physicalhabitats in small lowland streams. Fundam Appl Limnol 2009;174:89–99.

Pedersen ML, Friberg N. Influence of disturbance on habitats and biological communitiesin lowland streams. Fundam Appl Limnol 2009;174:27–41.

Piscart C, Genoel R, Doledec S, Chauvet E, Marmonier P. Effects of intense agriculturalpractices on heterotrophic processes in streams. Environ Pollut 2009;157:1011–8.

Rasmussen JJ, Friberg N, Larsen SE. Impact of lambda-cyhalothrin on a macroinvertebrateassemblage in outdoor experimental channels: implications for ecosystem function-ing. Aquat Toxicol 2008;90:228–34.

Rasmussen JJ, Baattrup-Pedersen A, Larsen SE, Kronvang B. Local physical habitat qualitycloud the effect of predicted pesticide runoff from agricultural land in Danish streams.J Environ Monit 2011a;13:943–50.

Rasmussen JJ, Baattrup-Pedersen A, Wiberg-Larsen P, Kronvang B. Buffer strip widthand agricultural pesticide contamination in Danish lowland streams: implicationsfor stream and riparian management. Ecol Eng 2011b;37:1990–7.

Schäfer RB, Caquet T, Siimes K,Mueller R, Lagadic L, LiessM. Effects of pesticides on commu-nity structure and ecosystem functions in agricultural streams of three biogeographicalregions in Europe. Sci Total Environ 2007;382:272–85.

Schäfer RB, Pettigrove V, Rose G, Allinson G, Wightwick A, von der Ohe PC, et al. Effects ofpesticidesmonitoredwith three samplingmethods in 24 sites onmacroinvertebratesand microorganisms. Environ Sci Technol 2011;45:1665–72.

Schäfer RB, BundschuhM, RouchDA, E.S., Von derOhe PC, PettigroveV, et al. Effects of pes-ticide toxicity, salinity and other environmental variables on selected ecosystem func-tions in streams and the relevance for ecosystem services. Sci Total Environ in press,doi:10.1016/j.scitotenv.2011.05.063.

Schulz R. Field studies on exposure, effects, and riskmitigation of aquatic nonpoint-sourceinsecticide pollution: a review. J Environ Qual 2004;33:419–48.

Tachet H, Richoux P, Bournard M, Usseglio-Polatera P. Invertébrés D'eau Douce: Systema-tique, biologie, écologie. Paris: CRNS Editions; 2002.

WauchopeRD. Pesticide content of surface-water draining fromagriculturalfields— review. JEnviron Qual 1978;7:459–72.

Wiberg-Larsen P, Pedersen SE, TornbjergNH, Sode A,Muus K,WehrsM. The Funen streams.Aquatic environmental monitoring: theme: freshwater. Funen CountyOdense: Dept. ofNature and Aquatic Environment; 1997. p. 209.

Young RG, Collier KJ. Contrasting responses to catchment modification among a rangeof functional and structural indicators of river ecosystem health. Freshw Biol2009;54:2155–70.

Young RG, Matthaei CD, Townsend CR. Organic matter breakdown and ecosystem me-tabolism: functional indicators for assessing river ecosystem health. J North AmBenthological Soc 2008;27:605–25.

http://dx.doi.org/doi:10.1016/j.scitotenv.2011.05.063

Paper V

1

Effects of a triazole fungicide and a pyrethroid insecticide on the decomposition

of leaves in the presence or absence of macroinvertebrate shredders

Jes Jessen Rasmussena*, Rikke Juul Monberga,b, Annette Baattrup-Pedersena, Nina Cedergreenc,

Peter Wiberg-Larsena, Bjarne Strobelc and Brian Kronvanga

a Aarhus University, Department of Bioscience, Vejlsøvej 25, 8600 Silkeborg, Denmark

b Department for Forest and Landscape, University of Copenhagen, Rolighedsvej 23, 1958

Frederiksberg C, Denmark

c Department of Basic Sciences and Environment, Faculty of Life Sciences, University of

Copenhagen, Frederiksberg, Denmark

* Corresponding author:

Jes Jessen Rasmussen

Aarhus University

Department of Bioscience

Vejlsøvej 25

8600 Silkeborg

Denmark

e-mail: [email protected]

phone: (+45)87158779

2

Abstract

Laboratory experiments have revealed that freely diluted azole fungicides potentiate the direct toxic

effect of pyrethroid insecticides on Daphnia magna. More ecologically relevant exposure scenarios

where pesticides are adsorbed have not been addressed. In this study we exposed beech leaves

(Fagus sylvatica) to the azole fungicide propiconazole (50 or 500 μg L-1), the pyrethroid insecticide

alpha-cypermethrin (0.1 or 1 μg L-1) or any combination of the two for 3 hours. Exposed leaves

were transferred to aquaria with or without an assemblage of macroinvertebrate shredders, and we

studied treatment effects on rates of microbial leaf decomposition, microbial biomass (using C:N

ratio as a surrogate measure) and macroinvertebrate shredding activity during 26 days post-

exposure. Microbial leaf decomposition rates were significantly reduced in the propiconazole

treatments, and the reduction in microbial activity was significantly correlated with loss of

microbial biomass (increased C:N ratio). Macroinvertebrate shredding activity was significantly

reduced in the alpha-cypermethrin treatments. In addition, the macroinvertebrate assemblage

responded to the propiconazole treatments by increasing their consumption of leaf litter with lower

microbial biomass, probably to compensate for the reduced nutritional quality of this leaf litter. We

found no interaction between the two pesticides on macroinvertebrate shredding activity, using

Independent Action as a reference model. In terms of microbial leaf decomposition rates, however,

alpha-cypermethrin acted as an antagonist on propiconazole. Based on these results we emphasise

the importance of considering indirect effects of pesticides in the risk assessment of surface water

ecosystems.

Keywords: Leaf litter decomposition, pesticides, macroinvertebrates, microorganisms, mixture,

pulse exposure

3

1. Introduction

The conversion of coarse particulate organic matter (CPOM) into fine particulate organic

matter (FPOM) is a fundamental stream ecosystem process that is primarily mediated by

microorganisms and shredding macroinvertebrates. Microorganisms, especially hyphomycetes, are

important for the conversion of leaf litter into more palatable food resources for macroinvertebrate

shredders and collectors/gatherers as they degrade indigestible polysaccharides and increase the

nutritional value of leaves (Bärlocher and Kendrick, 1975b; Gessner et al., 2007). Shredding

macroinvertebrates convert CPOM into minor leaf fragments and faecal pellets (Graca, 2001), and

several species of macroinvertebrates have been shown to consistently select or reject leaf

fragments dependent on which species of aquatic fungi that colonized the leaf material (Arsuffi and

Suberkropp, 1989). However, total microbial biomass on leaves is recognized as an overall robust

indicator for preferred macroinvertebrate shredding activity (Graca, 2001; Gulis et al., 2006).

Periodic contamination with agricultural pesticides potentially impairs the structure and/or the

function of stream biota, and agricultural streams are recognized as some of the most impacted on

earth (MEA, 2005). Indeed the co-occurrence of numerous pesticides in agricultural streams has

been reported frequently in scientific studies and monitoring programs (Bøgestrand, 2007; Martin et

al., 2003; Rasmussen et al., 2011; Schäfer et al., 2011b).

Research reviews of the interaction between chemical toxicants in mixtures show that the vast

majority of effects can be predicted with the model of concentration addition (Belden et al., 2007;

Cedergreen et al., 2008; Deneer, 2000). However, in approximately 5% of the studies of binary

mixtures the effect was significantly potentiated, and the majority of these studies involved azole

fungicides (Cedergreen et al., 2008; Cedergreen et al., 2006; Nørgaard and Cedergreen, 2010).

Azole fungicides (including triazoles and imidazoles) are C14α-demethylase inhibitors that obstruct

the biosynthesis of ergosterol (a significant part of fungi cell walls) by inhibiting the activity of an

4

enzyme belonging to the group of P450 monooxygenases (Copping and Hewitt, 1998). Their

synergising potential is hypothesised to be due to inhibition of P450 monooxygenases responsible

for the oxygenation and thereby degradation of other xenobiotics in target organisms (Walker,

2009). Azole fungicides are often applied to agricultural fields in tank mixtures with pyrethroid

insecticides, and recent studies have confirmed that some azole fungicides synergise the effect of

pyrethroid insecticides both under laboratory and field conditions (Bjergager et al., 2011; Nørgaard

and Cedergreen, 2010). Field screenings of stream water have disclosed concentrations up to 175 μg

L-1 and 6.2 μg L-1 of azole fungicides and pyrethroid insecticides, respectively (Elsaesser and

Schulz, 2008; Liess et al., 1999). However, due to their physicochemical properties pyrethroids and

some azole fungicides probably primarily occur in surface waters as sorption-complexes with

organic particles (Ding et al., 2010; Ensminger et al., 2011).

Short term effects of pyrethroid insecticides on benthic macroinvertebrates have been reported

to include reduced feeding rate of some shredders and grazers (Lauridsen et al., 2006; Rasmussen et

al., 2008) and altered locomotor behaviour (Nørum et al., 2010), whereas long term effects include

reduced growth, fecundity and emergence success (Liess and Schulz, 1996; Schulz and Liess,

2001a, b). Recently, research on effects of fungicides on non-target microorganisms has received

increasing attention, and changes in structure or function of microbial communities have been

documented in laboratory studies and in the field (Dijksterhuis et al., 2011; Rasmussen et al., 2012;

Schäfer et al., 2011a). Moreover, Bundschuh et al. (2011) showed that Gammarus fossarum

preferred unexposed leaves over those that were previously exposed to an azole fungicide, and this

selective behaviour was ascribed to pesticide-induced changes in the microbial community structure

associated with the leaf fragments. In the field however, these effects may be counteracted by

functional redundancy and high abundances of shredders such as Gammaridae (Piscart et al., 2009;

5

Rasmussen et al., 2012). Moreover, in agricultural streams shredding macroinvertebrates may not

have the option to select between exposed and unexposed sources of food.

In the present study we applied a classic cross factorial design to study the decomposition

rates of leaf litter that was pulse-exposed to the triazole fungicide propiconazole and/or the

pyrethroid insecticide alpha-cypermethrin. The decomposition rates were studied in the presence or

absence of an assemblage of macroinvertebrate shredders. In addition, we used the C:N ratio of

leaves as a proxy for microbial biomass and nutritional quality for macroinvertebrate shredders (as

C:N ratio decreases with increasing biomass of fungi). More specifically we tested the hypotheses

that 1) propiconazole exposure reduces the biomass of leaf-associated fungi (increasing the C:N

ratio) which is additionally reflected by reduced rates of microbial leaf decomposition, 2) alpha-

cypermethrin exposure decreases the rate of macroinvertebrate shredding due to either direct toxic

effects or a repelling effect, 3) the toxicity of alpha-cypermethrin to macroinvertebrates is

potentiated by the presence of propiconazole, and 4) the average macroinvertebrate shredding

activity will increase with decreased nutritional quality of leaves (higher C:N ratio), as the

shredding macroinvertebrates must compensate for low nutritional quality of ingested food by

increasing the feeding rate in order to maintain basic body functions. This picture may, however, be

obscured by the direct toxic effect of alpha-cypermethrin on macroinvertebrates in the insecticide

and combination treatments. The joint effects of alpha-cypermethrin and propiconazole were

predicted using the model of independent action (IA) (Bliss, 1939). This model was chosen as the

two pesticides are expected to have different modes of action on the two measured endpoints,

microbial and shredder mediated leaf decomposition, with propiconazole having a direct fungicidal

effect on the fungal community colonising the leaves, while alpha-cypermethrin will most likely

only exhibit a narcotic toxicity towards microorganisms which lack the nervous system of higher

organisms. The shredders are, however, likely to be directly affected by alpha-cypermethrin through

6

its effect on the nervous system (Copping and Hewitt, 1998), whereas propiconazole should not

have a direct toxic effect apart from the mentioned potential effects on the P450 mediated

metabolism of the macroinvertebrates (Walker, 2009).

2. Materials and methods

2.1 Leaf packs

Beech leaves (Fagus sylvatica) were collected in April from Søndervinge Brook – an

unpolluted first order stream in a catchment that is dominated by old beech forest. Thus, the

collected leaves were supposed to be uncontaminated by pesticides prior to the experiment and were

further assumed to be colonised with microorganisms. The leaves were stored in lightly aerated

stream water at 10 °C for three weeks prior to the experiment to allow for the microorganism

community to adapt to laboratory conditions using a diurnal light/darkness cycle of 14h/10h

respectively. Water chemistry of the stream water used at all stages of the experiment is presented

in table 1.

Leaves used in the experiment were carefully selected with highest possible similarity in terms

of texture and colour. Based on 50 randomly selected leaves the relation between fresh weight (FW)

and dry weight (DW) for the leaves was established prior to the experiment. The leaves were

Chemical parameters Concentration (mg L-1) NH4-N 0.006 (NO2 + NO3)-N 2.57 Total N 2.79 PO4-P 0.002 Total P 0.003 Total Fe 0.004 pH 7.0

Table 1. Chemical parameters for stream water used in the laboratory experiment

7

blotted and the petiole was removed from all leaves before establishing the relationship between

FW and DW. The DW was obtained by drying the leaves to constant mass at 60 °C. The leaves

were subsequently weighed using a Mettler Toledo XP-204 (10 μg accuracy).

Leaf packs for the experiment were produced the day before exposure by stacking 4-6 leaf

discs (2 cm diameter, petioles removed if present) with a total weight of 0.30 ± 0.01 g FW. The leaf

discs were threaded in a bundle using a polyester string.

2.2 Macroinvertebrates

The macroinvertebrates that was used in this study included two species: the amphipod

Gammarus pulex (L.) and the caddisfly Halesus radiatus (Curtis). They were collected in early May

in uncontaminated streams (Hagenstrup Millbrook, eastern Jutland, for G. pulex and Silke Stream,

South Funen, for H. radiatus). All macroinvertebrates were stored in lightly aerated stream water at

10 °C with sufficient leaf material to avoid biased rates of leaf decomposition in the early phase of

the experiment due to starved individuals. To harmonise the size of animals for the study, G. pulex

smaller than 10 mm in length were discarded, whereas only instars IV and V of H. radiatus were

selected. However, in order to relate the size of animals to results from previous studies, we

calculated the average DW of G. pulex based on a previously published relation between the length

of the first thoracic segment and their DW (n = 50) (Iversen and Jessen, 1977). The DW of the

incubated H. radiatus was measured as the average DW (n = 50) drying the animals to constant

mass at 60 °C. The average DW was 4.37 ± 0.14 mg for G. pulex and 4.56 ± 0.10 g for H. radiatus.

2.3 Pesticides and exposure

Propiconazole and alpha-cypermethrin were applied as the analytical standards PESTANAL®

(99.8% purity) and were purchased from Sigma-Aldrich (Selze, Germany). Dilutions of the

pesticides were produced immediately before use using a dilution series based on 96% ethanol. The

stock solutions were additionally produced in 96% ethanol. Nominal exposure concentrations were

8

50 μg L-1 and 500 μg L-1 for propiconazole and 0.1 μg L-1 and 1 μg L-1 for alpha-cypermethrin

including all combinations of propiconazole and alpha-cypermethrin. All pesticide exposure doses

were applied in 5 ml 96% ethanol, and control aquaria received an equivalent dose of ethanol in

order to compensate for potentially unintended effects of ethanol.

The leaves were exposed to pesticides in 2 L glass tanks in lightly aerated and moderately

stirred stream water at 10 °C for three hours reflecting one runoff episode from agricultural fields

(Wauchope, 1978). One water sample was collected from the exposure tank before the leaves were

submerged in order to verify actual exposure concentrations. Following the pesticide exposure, the

leaves were transferred to plastic aquaria with lightly aerated stream water for 24 hours to remove

potential pesticides that were not adsorbed to the leaves. Lastly, the leaf packs were transferred to

the aquaria microcosms with or without the macroinvertebrates.

Actual pesticide concentrations were measured using a method that is based on solid phase

extraction of propiconazole and alpha-cypermethrin followed by quantification by reverse-phase

HPLC-MS. The method is described in full detail in Nørum et al. (2010), and the detection limit

was 10 ng L-1 for both compounds. The actual concentrations of alpha-cypermethrin were 60-90%

of the nominal concentration which is acceptable given the physicochemical properties of the

compound. The actual concentrations of propiconazole were 80-90% of nominal concentrations.

For clarity the nominal concentrations are used in the succeeding parts of the article. We attempted

to measure pesticide content on leaves using acetone to extract alpha-cypermethrin and

propiconazole from the leaves and subsequently analyse the extract by gas chromatography and

mass spectrometry (GC-MS/NCI (Negative Chemical Ionisation). However, we were unable to

obtain sufficiently low detection limits to measure the actual pesticide contents on leaves.

2.4 Experimental set-up

9

The study was conducted at 10 °C using a diurnal light/dark cycle of 14/10 hours. Triplicate

plastic aquaria (25 L) were used to study rates of leaf decomposition facilitated by microbial

decomposers and/or macroinvertebrate shredders. A plastic net (500 μm mesh size) was mounted to

the smooth aquarium floor to provide a substrate suitable for the crawling of H. radiatus. The water

was lightly aerated in all aquaria. A classic cross factorial design was used to study the rates of

decomposition (with or without macroinvertebrates) of leaves that were pre-exposed to

propiconazole (50 μg L-1 and 500 μg L-1), alpha-cypermethrin (0.1 μg L-1 and 1 μg L-1) or any

combination of propiconazole and alpha-cypermethrin, and a control group was added (Table 2).

Triplicate aquaria were applied for each treatment totally amounting to two sets of 27 aquaria (one

set with and one set without macroinvertebrates). Twenty-one leaf packs were designated with ID

(on small plastic labels) and transferred to each of the aquaria. The average sum weight of leaf

packs was 1.38 ± 0.01 g DW per aquarium. Furthermore, three H. radiatus and 33 G. pulex were

introduced to each of 27 aquaria. The rationale behind the shredder ratio was that G. pulex in most

streams would be significantly more abundant, but at the same time significantly less efficient in the

shredding process than H. radiatus.

On experimental days 3, 6, 11, 18 and 26 three leaf packs were collected from each aquarium

using the leaf pack ID to randomise the sampling. Moreover, the macroinvertebrates in each

aquarium were registered as alive or dead. Sampled leaf packs were stored at -18 °C. In order to

Propiconazole (µgL-1)

0 50 500

0 Control + + 0.1 + + + Alpha-cypermethrin (µgL-1) 1.0 + + +

Table 2. Treatments with the insecticide alpha-cypermethrin and the fungicide propiconazole in the laboratory experiment with leaf packs (with and without macroinvertebrate shredders). Each of the nine different treatments was triplicated.

10

maximise the ability to compare time-integrated shredding activities, and to minimise temporal

changes of macroinvertebrate density dependant competition for space and food among treatments,

the ratio between numbers of remaining leaf packs and macroinvertebrates were maintained at a

constant level by removing equal proportions of leaves and macroinvertebrates during each

sampling. Removed macroinvertebrates were preserved in 96% ethanol. At day 26 the experiment

was terminated and remaining leaves and macroinvertebrates were collected and stored at -18 °C

and in 96% ethanol, respectively.

2.5 Leaf contents of carbon and nitrogen

The ratio between carbon (C) and nitrogen (N) content in leaves was used as a proxy for

microbial biomass, while also reflecting the nutritional quality of leaves for shredding

macroinvertebrates (Gessner and Chauvet, 1994; Gessner et al., 2007). Total carbon and nitrogen

abundance were determined according to the method described in Scrimgeour and Robinson (2003).

The dried leaf packs were ground individually and 5.6 ± 0.4 mg of the ground leaf material from

each pack were weighed into 6 x 4 mm tin cups for analysis by continuous flow Dumas combustion

using a Roboprep CN sample converter (Europa Scientific, Crewe, UK) in line with a Tracermass

spectrometer (Europa Scientific, Crewe, UK). Samples were quantified relative to a leucine/citric

acid mixture of known C and N content.

2.6 Data analysis

For each aquarium the remaining dry mass of leaves was fit to the exponential decay model

mt = m0 * e-kt (1)

where mt is the leaf dry mass remaining at time t, m0 is the initial dry mass, and k (day-1) is the rate

of leaf decomposition. The k was applied as measure for leaf decomposition for microbial

decomposition and macroinvertebrate shredding activity (Abelho, 2001), hence, one k value was

calculated for each aquarium (Appendix A).

11

Two-way analysis of variance was used to analyse for potential differences in rates of leaf

decomposition (k), macroinvertebrate mortality and average C:N ratios among pesticide treatments

(n = 3, P < 0.05). The Shapiro-Wilk test was used to determine homogeneity of variance, and

Bonferoni adjusted Fisher LSD was used as post hoc test for pairwise differences (P < 0.05).

Moreover, we used one-way analysis of variance to analyse for potential differences in rates of leaf

decomposition (k) in the presence or absence of macroinvertebrate shredders among three of the

pesticide treatments. We tested for differences in macroinvertebrate shredding activity among

treatments with 0, 0.1 and 1 μg L-1 alpha-cypermethrin within each of the propiconazole treatments

(0, 50 and 500 μg L-1), and we tested for differences in microbial decomposition rates among

propiconazole treatments (0, 50 and 500 μg L-1) within each of the alpha-cypermethrin treatments

(0, 0.1 and 1 μg L-1) . We used the Bonferoni adjusted Fisher LSD as post hoc test, and for clarity

we decided a priori only to compare the C:N ratio for each treatment to the C:N ratio for the control

group. Homogeneity of variance was confirmed for all data. All ANOVAs and post hoc tests were

conducted using SAS Enterprise Guide 4.2.

Using linear regressions we tested strength of the correlation between rates of microbial litter

decomposition and the C:N ratio of leaves. Furthermore, we tested if the nutritional quality of

leaves (C:N ratio) was a significant predictor for macroinvertebrate shredding activity. We used the

C:N ratios measured on experimental days 3, 6, 11, 18, 26 and the time-weighed average (n = 27) as

input parameters. All regression analyses were performed in SigmaPlot 11.0.

Possible interactions between the two pesticides on microbial decomposition and shredder

activity was tested using the reference model; Independent Action (IA) (Bliss, 1939). Independent

action is originally based on probabilities and states that the probability of being un-affected by a

combination of chemicals is equal to the product of probabilities of not being affected by each of

the chemicals separately. Expressed mathematically, this is:

12

∏=

==n

innBAmix RRRRR

1

........** (2)

Where Rmix is the probability of not being affected by the mixture and RA, RB and Rn denote the

probability of not being affected by chemical A or B up to n chemicals. Even though IA is based on

the assumption of binominal endpoints and probabilities, it has been shown to be a strong descriptor

of joint effects on gradual data, and often being quite similar to predictions using Concentration

Addition (CA) as the reference model (Cedergreen et al., 2008). In this study, average microbial

and shredder leaf decomposition rates of all mixture treatments calculated as a proportion of the

untreated controls were used to calculate the IA predictions.

3. Results

3.1 Microbial leaf decomposition

We found that the rates of microbial leaf decomposition (kmicrobial) were significantly different

among treatments (P = 0.003) with the effect of propiconazole being significant (P < 0.001) (Two-

way ANOVA) (raw data shown in Appendix B). We found no significant effects of alpha-

cypermethrin, and there were no significant interactions on kmicrobial (P = 0.057 and P = 0.159,

respectively). The pure 50 μg L-1 and 500 μg L-1 propiconazole treatments significantly reduced

kmicrobial when compared to controls (P = 0.016 and P = 0.002, respectively). The kmicrobial was not

significantly different between the 50 μg L-1 and 500 μg L-1 propiconazole treatments (P = 0.093)

(Fig. 1). Moreover, the additional presence of alpha-cypermethrin reduced the effect of

propiconazole on kmicrobial, but there was still a tendency of a decreased kmicrobial, but the decrease

was less and not significant (Fig. 1B and C). Comparing the observed decomposition rates of the

mixed treatments with the IA predictions showed that the IA predictions were lower than the

observed decomposition rates and not included in the 95% confidence intervals in three of the four

13

mixtures. This strongly indicates significant antagony between the two chemicals on microbial

decomposition of leaves.

Figure 1. Rates of leaf decomposition facilitated by microbial organisms (kmicrobial) subsequent to a 3 hour pulse exposure to 0, 50 or 500 μg L-1 propiconazole. Each of the propiconazole treatments were combined with the simultaneous exposure to 0 (A), 0.1 (B) or 1 (C) μg L-1 alpha-cypermethrin. Significant effects of the treatment were found for the pure propiconazole treatments (ANOVA P < 0.05, n = 3). Asterisks indicate the level of significance (* indicates P < 0.05 and ** indicates P < 0.005). IA predictions of the mixture treatments based on the single chemical responses are given as open symbols. Error bars indicate 95% confidence intervals.

14

3.2 Macroinvertebrate shredding activity

We found no significant effect of pesticide treatments on macroinvertebrate mortality (P >

0.05) (Appendix C). The macroinvertebrate shredding activity (kshredders) was significantly different

among treatments (P = 0.047), and there was a significant effect of alpha-cypermethrin (P = 0.003),

but not of propiconazole (P = 0.581). We found no interactions between treatments (P = 0.619)

(Two-way ANOVA) (raw data shown in Appendix B). The pure alpha-cypermethrin treatments (0.1

μg L-1 and 1 μg L-1) significantly reduced the shredding activity compared to the control (P = 0.036

and P = 0.003, respectively) (Fig. 2A), but there was no significant difference in shredding activity

between the two alpha-cypermethrin treatments (P = 0.430). Additionally, the alpha-cypermethrin

treatments significantly reduced the macroinvertebrates shredding activity in mixtures with 50 μg L-

1 propiconazole but not in mixtures with 500 μg L-1 propiconazole (P = 0.018 and P = 0.246,

respectively) (Fig. 2B and C). As shredder rates actually increased in response to propiconazole

(probably resulting from the effect of propiconazole on the nutritional quality of leaves as discussed

below), this caused effect estimates to be > 1. As IA theory is based on binary endpoints, effects

above 100% are not possible. From a biological point of view, it does, however, make sense to

consider one chemical as having stimulating effects and another having inhibitory effects, with the

combined effects being a product of the two (discussed in Ohlsson et al., 2010). Using the true

effects of propiconazole on shredder activity (values > 1) the mixed treatments with the IA

predictions showed that the predictions were within the 95% confidence intervals of the observed

shredding rates for both combinations of leaves treated with 0.1 μg L-1 alpha-cypermethrin.

However, for the combination treatments with 1 μg L-1 alpha-cypermethrin and 50 or 500 μg L-1

propiconazole the observed shredding rates were significantly lower than those predicted by IA

(Fig. 2B and C). If the propiconazole treatment increased the shredding rates, the alpha-

cypermethrin treatment decreased shredding rates proportionally or more than expected by IA.

15

Assuming that the propiconazole treatment did not affect the shredding rates, IA predictions were

within the 95% confidence limits for all mixture treatments (Fig. 2B and C).

Figure 2. Rates of leaf decomposition facilitated by macroinvertebrate shredders (kshredder). Leaves were exposed to 0, 0.1 or 1 μg L-1 alphacypermethrin for 3 hours, and after a rinsing procedure they were introduced to the macroinvertebrate shredder assemblage. Each of the alpha-cypermethrin treatments were combined with the simultaneous exposure to 0 (A), 50 (B) or 500 (C) μg L-1 propiconazole. Significant effects of the treatment are indicated with asterisks (* indicates P < 0.05 and ** indicates P < 0.005). IA predictions of the mixture treatments based on the single chemical responses accepting propiconazole induced shredder stimulation are given as open symbols, while predictions not accepting stimulation are given in grey symbols. Error bars indicate 95% confidence intervals.

16

3.3 C:N ratio

We found no significant differences among the average C:N ratios for the treatments (P >

0.05) (Appendix D). Moreover, we found no significant effect of pesticides on the C:N ratios for

experimental days 3, 6, 11, 18 or 26 (P > 0.05, data not shown). The kmicrobial significantly decreased

with increasing average C:N ratio for leaves (R2 = 0.47, P < 0.001, n = 27) (Fig. 3). The slope of the

regression line was significantly different from zero (P < 0.05). In addition, the kshredder significantly

increased with increasing C:N ratio for leaves (R2 = 0.4, P < 0.001, n = 27) (Fig. 3), and the slope of

the regression line was significantly different from zero (P < 0.05). Using the C:N ratio for

experimental days 3, 6, 11, 18 or 26 as alternative to the average C:N ratio for the entire study

period did not improve the level of significance for any of the correlations.

Figure 3. The rate of leaf decomposition (k) facilitated by microbial organisms (A) and an assemblage of macroinvertebrate shredders (B) as a function of the average C:N ratio of leaves that were previously exposed to one of eight different treatments with propiconazole, alpha-cypermethrin or the combination of the two (and control). Dashed lines represent 95% confidence intervals.

17

4. Discussion

4.1 Microbial leaf decomposition

As hypothesised, a short pulse of the triazole fungicide propiconazole significantly reduced

the rate of microbial leaf decomposition over a 26 days period, and this effect tended to increase

with increasing exposure concentration. The results probably reflect a direct toxic effect of

propiconazole on the fungal microbial community caused by the inhibition of C14α-demethylase

which is involved in the biosynthesis of ergosterol, an essential membrane sterol for fungi (Copping

and Hewitt, 1998). Propiconazole is likely to adsorb to the leaf surfaces due its lipophilic properties

(log Kow = 3.7). Since the leaf surface is colonised with microbial organisms such as freshwater

fungi (Bärlocher and Kendrick, 1975a, b) the probability of propiconazole reaching the non-target

freshwater fungi is expected to be relatively high, enforcing the toxic potential of propiconazole.

We did not find a significant effect of pesticide treatments on the C:N ratio for the leaves due

to large variations within treatments, but the significant decrease of microbial decomposition rates

with increasing average C:N ratio of the leaves confirms that the changes in C:N ratio is linked to

the degree of microbial colonisation of the leaves. This indicates that the rate of microbial activity

was a more sensitive endpoint than the microbial biomass estimate measured as C:N ratio due to

lower data variability. However, this rank ordering of endpoint sensitivity receives varying support

in the literature (Duarte et al., 2008a; Fernandes et al., 2009; Gessner and Chauvet, 2002; Schäfer et

al., 2011a). The C:N measurements reflect the sum of living and dead biomass, whereas microbial

activity represents living cells only, and a slow decomposition of dead microbial organisms may be

the reason for the microbial activity being the more sensitive endpoint for fungicidal effects of

propiconazole in our study.

The presence of alpha-cypermethrin seemed to reduce the toxic effect of propiconazole on

rates of microbial decomposition, and when compared to IA mixture toxicity predictions, alpha-

18

cypermethrin showed antagonistic behaviour in the combination treatments. We have found no

obvious explanation for this observation but can only speculate that the presence of alpha-

cypermethrin may alter the sorption characteristics for propiconazole, though this seems unlikely

considering the low concentrations of alpha-cypermethrin compared to propiconazole and the

lipophilic binding mechanisms of both compounds. However, the attempted measurements of

pesticide contents of the leaf material were not successful; hence we cannot confirm nor reject this

hypothesis.

4.2 Macroinvertebrate shredding activity

The macroinvertebrate shredding activity was reduced by 50% compared with the control

group when leaves were previously exposed to the pyrethroid insecticide alpha-cypermethrin (0.1 or

1 μg L-1). Up to 98% of molecules of this substance are (due to its high Kow of 6.9) proposed to

form irreversible sorption complexes with organic material within two hours after the introduction

to freshwater (Lauridsen et al., 2006). Our results therefore most likely reflect that the

macroinvertebrate shredders to some extent ceased to consume the contaminated leaves due to a

direct but sublethal toxic effect of the contaminated leaves or a repelling effect. Similar effects were

observed in a study of Lauridsen et al. (2006) where shredding rates of the caddis fly Sericostoma

personatum were reduced when leaves were previously exposed to 0.3 μg L-1 of the pyrethroid

lambda-cyhalothrin for 24 h. In our study, the presence of two different species of

macroinvertebrate shredders could possibly introduce some functional redundancy to the system,

but reduced shredding activity was observed down to 0.1 μg L-1 which is clearly below the lowest

concentration of observed effects in Lauridsen et al. (2006). The increased sensitivity to alpha-

cypermethrin for one or both macroinvertebrate species therefore could be a potential side effect of

several organisms potentially competing for food and space. Interspecies competition has been

19

shown to significantly increase the effects of pesticide exposure in other studies (Beketov and Liess,

2007).

Propiconazole had no significant adverse effect on the macroinvertebrate shredding activity.

On the contrary, shredder activity seemed to increase with increasing propiconazole exposure. Since

the nutritional quality of microbial cells is four to ten times higher than the nutritional quality of

unconditioned leaves, a reduction of the microbial biomass, resulting from the exposure of

propiconazole, is likely to reduce the nutritional quality of leaves (i.e. increased C:N ratio)

(Bärlocher and Kendrick, 1975a, b). The reduced nutritional quality of propiconazole treated leaves

could have caused the macroinvertebrate assemblage to compensate for a potential nutritional

deficiency by increasing the consumption of leaf litter in order to meet energy requirements for

basic physiological processes (Simpson and Simpson, 1990). Hence, even if propiconazole did not

have a direct toxic effect on the shredders, it could affect the feeding behaviour and nutritional

status of the shredders indirectly through its effect on the microbial colonisation of the leaves and

thereby food quality.

The results of the mixture treatment additionally need to be interpreted in that context. If we

accept that feeding activity increase in the presence of propiconazole, the high alpha-cypermethrin

concentration decrease shredder activity proportionately more in the two propiconazole treatments

compared to the decrease in the pure alpha-cypermethrin treatment. This could be interpreted as

synergy, resulting either from initial increased feeding rates resulting in higher exposure to alpha-

cypermethrin, or from propiconazole having a direct impact on the metabolisation of alpha-

cypermethrin. The latter is believed to be the cause of synergy observed between freely diluted

pyrethroid insecticides on Daphnia magna and other zooplankton in laboratory and microcosm

studies (Bjergager et al., 2011; Nørgaard and Cedergreen, 2010). However, the variance in the test

system is high, and the number of replicates low. And if not accepting an increased shredding

20

activity in the propiconazole treatments, but simply assuming them to be as the pure control, there is

no difference between observed and predicted shredding activity in the mixture treatments. The

interpretation of potential synergistic interactions between the pesticides when adsorbed to organic

matter and when using a non-lethal endpoint such as shredding activity consequently needs more in-

depth studies where exposure can be quantified over time together with the proposed physiological

effects of both propiconazole and alpha-cypermethrin.

4.3 Microbial biomass and nutritional quality

We found no significant effect of pesticide treatment on the C:N ratio of the leaves, but the

average C:N ratios explained 47% of the variation in rates of microbial decomposition, where

increasing C:N ratios were characterised by decreasing rates of microbial decomposition. This

clearly indicates that there was a general effect of the different pesticide treatments, and that the

C:N ratio did reflect microbial abundance and activity. In addition, the correlation shows that there

was no significant functional redundancy in these microorganism communities. Increasing

abundance of more tolerant species, however, has been observed in other studies as a result of

competitive release which is facilitated by decreasing abundance of more sensitive species as a

result of toxic stress (Duarte et al., 2008b; Gessner and Chauvet, 2002). Consistent with our results

Bundschuh et al. (2011) observed a significant reduction of the microbial biomass on leaves after 12

days exposure to 500 μg L-1 of the triazole fungicide tebuconazole. These results confirm that

triazole fungicides are toxic for aquatic microorganisms, and the results from our study extend these

findings into the range of environmentally realistic exposure durations.

As already mentioned, the rate of macroinvertebrate shredding activity significantly increased

with decreasing nutritional quality of leaves (higher C:N ratio) with low nutritional value of leaves

primarily being associated with the fungicide treatments, while the increased shredder activity was

believed to be a compensatory behaviour with the aim of optimising energy input (Simpson and

21

Simpson, 1990). These interrelated responses have important implications for ecosystem

functioning, because reduced nutritional quality of food may have cascading effects reducing the

amount of energy available for growth and reproduction in higher trophic levels. The effect of

decreased food quality and hereby energy resources may be further enhanced under natural

conditions, as the shredding macroinvertebrates in streams that are impacted by agricultural

pesticides may have increased energy demands in order to cope with toxic stress. Our results are

supported by Bundschuh et al. (2011) who showed that triazole fungicides can reduce the overall

microbial biomass, and Zubrod et al. (2011) showed that reduced nutritional value of leaves

resulting from exposure to an azole fungicide increased the food intake by macroinvertebrate

shredders but reduced their physiological fitness. Applying endpoints that are based on scope for

growth assessments are important and highly ecologically relevant supplements to studies on the

activity of secondary producers, as it provides useful information on the potential changes in

interspecies competition and selective forces that arise in consequence of pesticide stress

(Bundschuh et al., 2011; Zubrod et al., 2011; Zubrod et al., 2010).

Conclusions

In this study, we showed that a 3h pulse exposure of environmentally realistic exposure

concentrations and durations of propiconazole significantly reduced average rates of microbial leaf

decomposition for the subsequent 26 days. Moreover, this reduction in leaf decomposition rates was

associated with a reduction in nutrient concentrations of the leaves reflecting reduced microbial

biomass. Exposing leaves to a 3h pulse of alpha-cypermethrin significantly reduced shredding

activity of an assemblage of macroinvertebrate shredders reflecting either direct toxic effects or a

repellent effect. Alpha-cypermethrin was found to act as antagonist on propiconazole in terms of

microbial activity whereas we did not find significant interactions between propiconazole and

22

alpha-cypermethrin on shredding activity or mortality. Lastly, we found that macroinvertebrate

shredding activity significantly increased with decreasing nutrient concentrations of leaves

independent of the pesticide treatments. This effect we believe is a compensatory behaviour to

maintain basic physiological processes. The results from this study have important implications for

understanding the complexity of ecotoxicological effects of pesticides in the field where some toxic

compounds may act on the habitat or food choice of an organism and other compounds may act on

the organism itself, which eventually may increase the total ecotoxicological effect on the

ecosystem structure and function.

Acknowledgements

This study was financed by the Danish Research Council (grant no. 2104-07-0035) and is part

of the RISKPOINT project. We wish to thank Johnny Nielsen, Uffe Mensberg and Marlene Venø

Skjærbæk for laboratory assistance and Ulrik Nørum and Klaus Brodersen for useful feedback on

set-up development and statistical analyses.

References

Abelho, M., 2001. From litterfall to breakdown in streams: A review. The scientific world 1, 656-680.

Arsuffi, T.L., Suberkropp, K., 1989. Selective feeding by shredders on leaf-colonizing stream fungi - Comparison of macroinvertebrate taxa. Oecologia 79, 30-37.

Beketov, M.A., Liess, M., 2007. Predation risk perception and food scarcity induce alterations of life-cycle traits of the mosquito Culex pipens. Ecol. Entomol. 32, 405-410.

Belden, J.B., Gilliom, R.J., Martin, J.D., Ludy, M.J., 2007. Relative toxicity and occurrence patterns of pesticide mixtures in streams draining agricultural watersheds dominated by corn and soybean production. Integrated Environ. Assess. Manag. 3, 90-100.

Bjergager, M.B.A., Hanson, M.L., Lissemore, L., Henriquez, N., Solomon, K.R., Cedergreen, N., 2011. Synergy in microcosms with environmentally realistic concentrations of prochloraz and esfenvalerate. Aquat. Toxicol. 101, 412-422.

Bliss, C.I., 1939. The toxicity of poisons applied jointly. Annals of Applied Biology 26, 585-615. Bundschuh, M., Zubrod, J.P., Kosol, S., Maltby, L., Stang, C., Duester, L., Schulz, R., 2011. Fungal

composition on leaves explains pollutant-mediated indirect effects on amphipod feeding. Aquat. Toxicol. 104, 32-37.

23

Bärlocher, F., Kendrick, B., 1975a. Assimilation efficiency of Gammarus pseudolimnaeus (Amphipoda) feeding on fungal mycelium or autumn shed leaves. Oikos 26, 55-59.

Bärlocher, F., Kendrick, B., 1975b. Leaf-conditioning by microorganisms. Oecologia 20, 359-362. Bøgestrand, J., 2007. Stream conditions, 2006. NOVANA. Scientific report from NERI No. 643 (In

Danish). National Environmental Research Institute, Aarhus University, Aarhus, Denmark, p. 70pp.

Cedergreen, N., Christensen, A.M., Kamper, A., Kudsk, P., Mathiassen, S.K., Streibig, J.C., Sorensen, H., 2008. A review of independent action compared to concentration addition as reference models for mixtures of compounds with different molecular target sites. Environ. Toxicol. Chem. 27, 1621-1632.

Cedergreen, N., Kamper, A., Streibig, J.C., 2006. Is prochloraz a potent synergist across aquatic species? A study on bacteria, daphnia, algae and higher plants. Aquat. Toxicol. 78, 243-252.

Copping, L.G., Hewitt, H.G., 1998. Chemistry and mode of action of crop protection agents. The Royal Society of Chemistry, Cambridge.

Deneer, J.W., 2000. Toxicity of mixtures of pesticides in aquatic systems. Pest Manag. Sci. 56, 516-520.

Dijksterhuis, J., van Doorn, T., Samson, R., Postma, J., 2011. Effects of seven fungicides on non-target aquatic fungi. Water Air Soil Poll. 222, 421-425.

Ding, Y.P., Harwood, A.D., Foslund, H.M., Lydy, M.J., 2010. Distribution and toxicity of sediment-associated pesticides in urban and agricultural waterways from Illinois, USA. Environ. Toxicol. Chem. 29, 149-157.

Duarte, S., Pascoal, C., Alves, A., Correia, A., Cassio, F., 2008a. Copper and zinc mixtures induce shifts in microbial communities and reduce leaf litter decomposition in streams. Freshwat. Biol. 53, 91-101.

Duarte, S., Pascoal, C., Cassio, F., 2008b. High Diversity of Fungi may Mitigate the Impact of Pollution on Plant Litter Decomposition in Streams. Microb. Ecol. 56, 688-695.

Elsaesser, D., Schulz, R., 2008. Mitigation of fungicide pollution in vegetated agricultural surface waters: GIS modelling and monitoring in the field., SETAC Europe 18th annual meeting, Warsaw, pp. 406-407.

Ensminger, M., Bregin, R., Spurlock, F., Goh, K.S., 2011. Pesticide concentrations in water and sediment and associated invertebrate toxicity in Del Puerto and Orestimba Creeks, California, 2007-2008. J. Environ. Monit. Assess. 175, 573-587.

Fernandes, I., Duarte, S., Cassio, F., Pascoal, C., 2009. Mixtures of zinc and phosphate affect leaf litter decomposition by aquatic fungi in streams. Sci. Total. Environ. 407, 4283-4288.

Gessner, M.O., Chauvet, E., 1994. Importance of stream microfungi in controlling breakdown rates of leaf-litter. Ecology 75, 1807-1817.

Gessner, M.O., Chauvet, E., 2002. A case for using litter breakdown to assess functional stream integrity. Ecol. Appl. 12, 498-510.

Gessner, M.O., Gulis, V., Kuehn, K.A., Chauvet, E., Suberkropp, K., 2007. Fungal decomposers of plant litter in aquatic ecosystems, in: Kubicek, C.P.a.D., I.S. (Ed.), Environmental and microbial relationships, 2nd edition. Springer Verlag, Berlin.

Graca, M.A.S., 2001. The role of invertebrates on leaf litter decomposition in streams - A review. Int. Rev. Hydrobiol. 86, 383-393.

Gulis, V., Ferreira, V., Graca, M.A.S., 2006. Stimulation of leaf litter decomposition and associated fungi and invertebrates by moderate eutrophication: implications for stream assessment. Freshwat. Biol. 51, 1655-1669.

Iversen, T.M., Jessen, J., 1977. Life-cycle, drift and production of Gammarus pulex L. (Amphipoda) in a Danish spring. Freshwat. Biol. 7, 287-296.

24

Lauridsen, R.B., Kronvang, B., Friberg, N., 2006. Occurrence of sediment-bound pyrethroids in danish streams and their impact on ecosystem function. Wat. Air Soil Poll. 6, 423-432.

Liess, M., Schulz, R., 1996. Chronic effects of short-term contamination with the pyrethroid insecticide fenvalerate on the caddisfly Limnephilus lunatus. Hydrobiologia 324, 99-106.

Liess, M., Schulz, R., Liess, M.H.D., Rother, B., Kreuzig, R., 1999. Determination of insecticide contamination in agricultural headwater streams. Water Res. 33, 239-247.

Martin, J.D., Crawford, C.G., Larson, S.J., 2003. Pesticides in streams. National Water Assessment Program (NAWQA), USA.

Millenium Ecosystem Assessment, 2005. Millenium Ecosystem Assessment. Ecosystems and human well-being: synthesis. Island Press, Washington, DC.

Nørgaard, K.B., Cedergreen, N., 2010. Pesticide cocktails can interact synergistically on aquatic crustaceans. Environ. Sci. Poll. Res. 17, 957-967.

Nørum, U., Friberg, N., Jensen, M., Pedersen, J., Bjerregaard, P., 2010. Behavioural changes in three species of freshwater macroinvertebrates exposed to the pyrethroid lambda-cyhalothrin: Laboratory and stream microcosm studies. Aquat. Toxicol. 98, 328-335.

Ohlsson, A., Cedergreen, N., Oskarsson, A., Ulleras, E., 2010. Mixture effects of imidazole fungicides on cortisol and aldosterone secretion in human adrenonocortical H295R cells. Toxicology 275, 21-28.

Piscart, C., Genoel, R., Doledec, S., Chauvet, E., Marmonier, P., 2009. Effects of intense agricultural practices on heterotrophic processes in streams. Environ. Poll. 157, 1011-1018.

Rasmussen, J.J., Baattrup-Pedersen, A., Wiberg-Larsen, P., McKnight, U.S., Kronvang, B., 2011. Buffer strip width and agricultural pesticide contamination in Danish lowland streams: Implications for stream and riparian management Ecol. Eng. 37, 1990-1997.

Rasmussen, J.J., Friberg, N., Larsen, S.E., 2008. Impact of lambda-cyhalothrin on a macroinvertebrate assemblage in outdoor experimental channels: Implications for ecosystem functioning. Aquat. Toxicol. 90, 228-234.

Rasmussen, J.J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Monberg, R.J., Kronvang, B., 2012. Impacts of pesticides and natural stressors on leaf litter decomposition in agricultural streams. Sci. Tot. Environ. 416, 148-155.

Schulz, R., Liess, M., 2001a. Acute and chronic effects of particle-associated fenvalerate on stream macroinvertebrates: A runoff simulation study using outdoor microcosms. Arch. Environ. Contam. Toxicol. 40, 481-488.

Schulz, R., Liess, M., 2001b. Toxicity of aqueous-phase and suspended particle-associated fenvalerate: Chronic effects after pulse-dosed exposure of Limnephilus lunatus (Trichoptera). Environ. Toxicol.Chem. 20, 185-190.

Schäfer, R.B., Bundschuh, M., Rouch, D.A., E., S., Von der Ohe, P.C., Pettigrove, V., Schulz, R., Nugegoda, D., Kefford, B., 2011a. Effects of pesticide toxicity, salinity and other environmental variables on selected ecosystem functions in streams and the relevance for ecosystem services. Sci. Tot. Environ. 415, 69-78.

Schäfer, R.B., Pettigrove, V., Rose, G., Allinson, G., Wightwick, A., von der Ohe, P.C., Shimeta, J., Kuhne, R., Kefford, B., 2011b. Effects of Pesticides Monitored with Three Sampling Methods in 24 Sites on Macroinvertebrates and Microorganisms. Environ. Sci. Tech. 45, 1665-1672.

Scrimgeour, C.M., Robinson, D., 2003. Stable isotope analysis and applications, in: Smith, K.A., Cresser, M.S. (Eds.), Soil and environmental analysis: Modern Instrumental Techniques, 3rd ed. Marcel Dekker Inc., New York, NY, USA, pp. 381-431.

Simpson, S.J., Simpson, C.L., 1990. The mechanisms of nutritional compensation by phytophagous insects, in: Bernays, E.A. (Ed.), Insect-plant interactions. CRC Press, Boca Raton, Florida, USA, pp. 111-160.

25

Walker, C.H., 2009. Factors determining the toxicity of organic pollutants to animals and plants, Organic Pollutants: An ecotoxicological Perspective, 2nd ed. CRC Press, London, pp. 17-66.

Wauchope, R.D., 1978. Pesticide content of surface-water draining from agricultural fields - review. J. Environ. Qual. 7, 459-472.

Zubrod, J.P., Bundschuh, M., Feckler, A., Englert, D., Schulz, R., 2011. Ecotoxicological impact of the fungicide tebuconazole on an aquatic decomposer-detritivore system. Environ. Toxicol. Chem. 30, 2718-2724.

Zubrod, J.P., Bundschuh, M., Schulz, R., 2010. Effects of subchronic fungicide exposure on the energy processing of Gammarus fossarum (Crustacea; Amphipoda). Ecotoxicol. Environ. Saf. 73, 1674-1680.

Paper VI

18. årgang nr. 4, december 2011 • 143

Ny viden om effekter af pesticider i vandløb

Jes Jessen Rasmussen, PeteR WibeRg-LaRsen, annette baattRuP-PedeRsen, Rikke JuuL monbeRg, uRsuLa soLaRd

mcknight & bRian kRonvang

Medierne afspejler en stadig offentlig interesse for betydningen af

pesticider i vores miljø. Og pesticider forbindes automatisk med

dansk landbrug. Men er det retfærdigt? Og hvor stor betydning har

pesticiderne i forhold til andre faktorer, som påvirker miljøet? Ny

dansk og udenlandsk forskning kaster lys over effekterne af pesticider

i vore vandløb, og mulighederne for at afbøde negative effekter.

Den stigende globalisering øger konkurrencen på det internationale fødevaremarked, og dermed øges også kravene til en konkurrencedygtig og effektiv landbrugsdrift. Effektiv landbrugsdrift er for længst indført i Danmark, men udvikles til stadighed for at opretholde en høj produktion. Med de igangværende klimaforandringer får danske landmænd fordelen af en længere dyrkningssæson, samt potentielt muligheden for at dyrke andre afgrøder (primært kernemajs). Ændrede klimatiske forhold kan dog medføre en stigning i ’angreb’ fra skadedyr, svampe og ukrudt. Og man kan derfor forvente et øget behov for anvendelse af pesticider og deraf følgende forøget risiko for belastning af vandmiljøet. Et aktuelt dansk projekt (PRECIOUS) under Miljøstyrelsens Pesticidforskningsprogram søger netop at belyse dette ud fra bestemte klimascenarier og modeller. Uanset udfaldet af dette projekt er det rimeligt at forvente, at pesticidbelastningen af vores vandmiljø, herunder vandløbene, i hvert fald ikke bliver mindre end nu. Der er derfor god grund til at søge at gøre status over den nuværende viden om utilsigtede skadevirkninger af pesticider i vandløbsmiljøet.

Det er allerede dokumenteret, at mange af de pesticider, som anvendes i den daglige

landbrugsdrift, kan findes i vandløbene, hvor de potentielt kan udøve skadevirkninger på organismerne i disse. Særligt mange insektmidler, og i et vist omfang også svampemidler, kan have betydelig negativ virkning på mange af de smådyr, der indtager en central rolle i vandløbets økologiske system som blandt andet nedbrydere af organisk stof og føde for fisk. Selv koncentrationer under 0,1 µg/L af de langt mest anvendte insektmidler i dansk landbrug (de såkaldte pyrethroider) kan medføre øget dødelighed og ikke mindst få visse arter til at lade sig føre med strøm

men bort fra en påvirket strækning /1/. Der er tale om koncentrationer, som er mindre end den generelle grænseværdi for pesticider i dansk grundvand, som må anvendes som drikkevand. Der er således god grund til at interessere sig for de utilsigtede effekter af pesticider i danske vandløb.

Gennem det seneste årti har en del af midlerne fra Miljøstyrelsens Pesticidforskningsprogram været tildelt forskningsprojekter, der skulle belyse forskellige aspekter af de potentielle økologiske effekter i vandløb og – til dels søer. Og for tiden er der nye interes

Figur 1. Den mikrobielle omsætningshastighed af bøgeblade (kmicrobial) som funktion af giftigheden af de fundne pesticider (log mTUmicrobial) i 14 fynske vandløb. I hvert vandløb blev prøver indsamlet fra to strækninger, der var forskellige i fysisk kompleksitet og kvalitet (lav og høj).

144 • Vand & Jord

sante projekter undervejs. I denne artikel vil vi imidlertid primært rette fokus mod undersøgelser af effekten af pesticider på vandløbsorganismer, som er udført i andet regi, nemlig som en del af projektet RISKPOINT (finansieret via det Strategiske Forskningsråd), se www.riskpoint.dk. Vi vil desuden pege på, hvor vi mener, at det vil være fornuftigt rent forskningsmæssigt at satse i fremtiden for at beskytte vandløbene mod utilsigtet pesticidbelastning.

Pesticidforekomster og giftighedDer er efterhånden foretaget en del målinger af pesticidindholdet i vandløbsvand, og det er ret veldokumenteret, at de højeste koncentrationer findes i små vandløb under og lige efter større regnbyger, hvor pesticider bliver transporteret fra marker til vandløb primært via dræn og i sjældnere via tilfælde overfladisk afstrømning /2/. Langt de fleste målinger har imidlertid omfattet herbicider (ukrudtsmidler), mens målinger af fungicider (svampemidler) og insekticider (insektmidler) er yderst sparsomme.

Helt ny forskning bekræfter, at der jævnligt opstår situationer med forhøjede pesticidkoncentrationer, og at disse forventes at medføre effekter på vandløbets organismer /3/. Derudover gør nye og forbedrede metoder det muligt at måle de pesticider, der pga. deres fysiske og kemiske egenskaber sjældent (og kun kortvarigt) findes i vandfasen, men derimod bindes til organisk stof og partikler, som aflejres i og på vandløbets bund. Der er især tale om de allerede omtalte insektmidler, pyrethroiderne, som er de mest giftige pesticider for dyrene i vandløb overhovedet, og der er for nyligt gjort fund af relativt høje koncentrationer af 4 forskellige potente insektmidler (heriblandt netop to pyrethroider) i sedimentet i et sjællandsk vandløb /4/. Selvom sedimentbundne pesticider er mindre giftige, end pesticider på opløst form, så er der påvist negative effekter under naturlige forhold. Der er imidlertid stort behov for at forstå tilførsel og virkning af de partikelbundne stoffer, dels fordi vandløbsorganismernes optagsmekanismer af stofferne formodentlig er anderledes, end når de er opløst, og dels fordi virkningen er længere, når stofferne bliver en del af dyrenes fysiske miljø (fx et finkornet sediment) og indgår i deres føde. Disse problemstillinger er meget ufuldstændigt belyst i dansk såvel som international litteratur.

Men hvor giftige er ukrudts og svampemidlerne til sammenligning? Undersøgelser under Miljøstyrelsens Pesticidforskningsprogram tyder på, at ukrudtsmidlerne – i de koncentrationer de forekommer i – næppe har be

tydende effekter på alger, planter og smådyr i vandløbene. Nogle svampemidler er relativt giftige over for smådyr, men vi ved stort set intet om i hvilke koncentrationer, de optræder i vandløbene, og dermed hvor giftige de reelt er.

Biologiske effekter af pesticiderNår et bestemt pesticid skal godkendes, bliver dets giftighed fastlagt ud fra en række standardtests på bestemte standardorganismer. Disse tests udføres over relativt lange tidsrum (48 eller 96 timer) og stoffernes akut dødbringende effekt bestemmes. Forsøgene er derfor ikke miljørealistiske i forhold til de pulse, som stofferne reelt optræder i ude i et vandløb. Desuden optræder stofferne formodentlig sjældent i dødelige koncentrationer derude. Det er derfor mere meningsfyldt at fokusere på andre forhold i fx smådyrenes liv, som kan være nok så betydende som simpel dødelighed. Det kan fx være bevægelsesadfærd, fødeoptagelse, forvandling fra larve til voksen og formering. Påvirkning af disse forhold kan forskyde konkurrenceforholdene imellem arterne og over længere tid medføre ændret artssammensætning og størrelsesfordeling samt påvirke omsætningen af organisk stof i vandløbet.

Effekter af denne art er imidlertid svære at registrere i felten, og generelt er studier med så komplekse formål en mangelvare. Enkelte projekter under Miljøstyrelsens Bekæmpelsesmiddelforskning har dog dokumenteret ef

fekter på rovdyrbyttedyrsinteraktioner samt omsætningen af blade og algebiofilm /5,6/.

Blade som ved løvfald falder ned i vandløb har stor betydning for hele stof og energiomsætningen i disse. Mange smådyr lever af de døde blade og især af de mikroorganismer, som sidder på og nedbryder bladene. Helt ny dansk forskning /7/ har vist, at den mikrobielle nedbrydning af sådanne blade blev klart reduceret i vandløb, der var påvirket af især svampemidler, og at effekterne blev yderligere forstærkede, hvis de fysiske forhold var forarmede, som det er tilfældet i kanaliserede og hårdt vedligeholdte vandløb (Figur 1). Dette giver umiddelbart god mening, fordi den primære gruppe af mikrobielle nedbrydere er mikrosvampe, og fordi disse optager ilt og næringsstoffer fra vandfasen begunstiges de, når iltforholdene er optimale. Optimale iltforhold forekommer i naturligt slyngede vandløb med stedvis hurtig strøm og dermed god tilførsel af ilt fra luften. Den nedsatte mikrobielle omsætningshastighed kunne i laboratoriet /8/ sammenkædes med en ernæringsmæssig forringelse af bladene (lavere indhold af især kvælstof), hvilket kan skyldes en mindre mængde mikrosvampe på bladene (svampene indeholder væsentlig mere kvælstof end bladene selv) (se Figur 2). Smådyrene kompenserede dog ikke – som ellers forventet for denne næringsmæssige forringelse af bladene ved at spise mere af bladene. Trods dette er der grund til at antage, at hvis smådyr indtager føde af ringere værdi, ja

Figur 2. Den mikrobielle omsætningshastighed af bøgeblade (kmicrobial) som funktion af kulstof/kvælstof (C:N) forholdet i bladene. De enkelte datapunkter repræsenterer forskellige pesticideksponeringer forud for selve forsøgsperioden. C:N forholdet afspejler den mikro-bielle biomasse (højt C:N forhold kendetegner lav mikrobiel biomasse). 95 % konfidens-bånd er angivet.


så reduceres også den mængde energi, som de kan råde over til vækst, overlevelse og formering. Der kan derfor teoretisk set være risiko for, at et svampemiddel ligefrem kan fremkalde en kaskadeeffekt igennem et vandløbsøkosystem, hvor nedsat svampevækst medfører nedsat vækst hos smådyrene og efterfølgende nedsat vækst hos de fisk, som lever af smådyrene.

SPEAR – en mulig pesticidindikatorDet kan være urimeligt dyrt og besværligt at foretage omfattende og dybtgående analyser af omsætningsrater af organisk stof, formeringssucces eller ændring af gennemsnitsstørrelsen af organismer som del af et miljøovervågningsprogram. Et muligt alternativ blev lanceret i 2005 af en tysk forskergruppe i form af et nyt pesticidindikatorindeks kaldet SPEAR (SPEcies At Risk). SPEAR indekset (se Boks 1) bygger ligesom vores eget Dansk Vandløbsfauna Indeks (DVFI) på sammensætningen af vandløbenes smådyrssamfund /9/. DVFI er oprindeligt udviklet til at beskrive effekten af forurening med let nedbrydeligt organisk stof (som findes i almindeligt spildevand), men har vist sig også at afspejle andre relevante miljøfaktorer, fx de fysiske forhold. Vi har imidlertid kunnet påvise, at DVFI var dårligt til at afspejle den aktuelle og betydende pesticidbelastning i et sjællandsk vandløb /4/. Til gengæld syntes det tyske SPEAR indeks i samme vandløb at

være en stærkere indikator for pesticideffekter på smådyrssamfundene. Selvom SPEAR indekset i vores øjne lige nu formodentlig er det stærkeste bud på et indikatorsystem til en vurdering af pesticidbelastede vandløb, fandt vi dog også en række mangler, som skal belyses yderligere, førend indekset meningsfyldt kan bruges i overvågningen af

vandløbskvalitet. Således scorer SPEAR indekset væsentligt lavere på vandløbsstrækninger, der er stærkt fysisk forarmede (med ensartede, mudrede/sandede bundforhold), end på tilgrænsende uregulerede og varierede strækninger (Figur 3) /10/. Forklaringen kan naturligvis være helt simpel: nemlig tab af den fysiske variation, som er en betingelse for tilstedeværelsen af et artsrigt samfund af smådyr. En anden mulighed er, at smådyrene i et fysisk forarmet miljø er mere følsomme overfor en pesticidbelastning, fordi de i forvejen er under stress i et fysisk ugunstigt miljø. Ydermere kan de komplekse strømforhold, som er karakteristisk for uregulerede vandløb, skabe ”refugier” for smådyrene (fx partier af tætte puder af vandplanter, eller andre levesteder, hvor strømmen kun i mindre grad fører pesticider hen), når og hvis en puls af pesticid rammer vandløbet. Sådanne refugier vil kun i mindre grad forekomme i regulerede vandløb.

Kan pesticidbelastningen reduceres?Det er efterhånden dokumenteret, at pesticider forekommer i danske vandløb i koncentrationer, der kan have utilsigtede negative økologiske effekter. Og det er muligt at påvise disse effekter i danske vandløb ved at videreudvikle og anvende en allerede eksisterende pesticidindikator. Et tilbageværende men yderst relevant spørgsmål er imidlertid, om det så også er muligt at sikre en effektiv forebyggelse af pesticiders skader i vore vandløb.

Boks 1. SPEAR-indeksetMange forskellige menneskeskabte miljøfaktorer indvirker ofte samti-dig på de biologiske forhold i et vandløb. Hvis man derfor vil anvende fx artssammensætningen af smådyr i et vandløb til at beskrive effekten af disse miljøfaktorer, er det ønskeligt og afgørende at kunne doku-mentere betydningen af hver faktor for sig. Det kræver, at der fokuse-res på netop de arter, som påvirkes af en given faktor. SPEAR indek-set er et godt eksempel. Det ’rangordner’ hver enkelt art i en faunaprø-ve som enten følsom eller ufølsom overfor pesticidbelastning ud fra dels viden om dyrenes fysiologiske tolerance, dels deres evne til at genindvandre efter en given pesticidpåvirkning (typisk puls med forhø-jet koncentration). Genindvandringen afhænger af, hvor hurtigt dyrene kan formere sig, hvor meget afkom de får, og deres evne til at flytte sig rent fysisk. Dyr der formerer sig hurtig, lægger mange æg, og er gode til at indvandre over store afstande er mindre følsomme end arter, som formerer sig langsomt, lægger få æg og kun flytter sig ganske lidt fra deres oprindelige levested. SPEAR indekset udtrykker hvor stor en del af smådyrsarterne, som, på grundlag af rangordningen, er følsomme overfor en given pesticidbelastning. SPEAR indekset har vist sig at korrelere stærkt med den samlede målte giftighed af pesticider i vand-løb. Den samlede giftighed beregnes ved brug af såkaldte ’toxic units’ (TU) ud fra målte koncentrationer af de enkelte pesticider. Toxic units er baseret på effekten af stofferne over for ’standardorganismen’ Daphnia magna i standardiserede laboratorium tests.

Figur 3. SPEAR indeks scorer (% SPEAR abundance) som funktion af fundne pesticiders giftighed for smådyr (log mTUD.magna). Datapunkterne repræsenterer prøver fra 14 fynske vandløb, hvor der i hvert vandløb blev indsamlet prøver fra to strækninger, der var fysisk forskellige (høj og lav habitat kompleksitet).

146 • Vand & Jord

I en række fynske vandløb med ensartede geomorfologiske forhold fandt vi en meget tæt kobling mellem bredden af den dyrkningsfri zone langs vandløbene og koncentrationen og især giftigheden af de pesticider, der blev fundet i de respektive vandløb (se figur 4 og /3/). Det er veldokumenteret, at der opstår effekter på smådyrssamfundene omkring en bestemt tærskelværdi (3) for den samlede giftighed af pesticider (udtrykt ved såkaldte Toxic Units (TU), se også boks 1). Ved at benytte en veletableret matematisk sammenhæng mellem TU og bredden af de dyrkningsfri zoner, samt sammenhængen mellem SPEAR indekset og TU, beregnede vi, at en minimumsbredde på 67 meter er nødvendig for at mindske pesticidpåvirkningen så meget, at det vil være muligt at opnå ’god økologisk kvalitet’ i vandløbet. Det skal dog understreges, at lokale forhold som dræningsintensitet, forekomst af makroporer i jorden, den lokale topografi og de dyrkningsmæssige forhold også er af stor betydning for den samlede mængde pesticid, der vaskes ud fra markerne. Især vil dræn i markerne sende pesticider ud i vandløbene ved store regnskyl uanset bredden af en dyrkningsfri zone langs disse. Det er derfor helt afgørende at videreudvikle modeller for tabet af pesticider fra mark til grundvand og overfladevand, for derved bedre at kunne vurdere effekten af virkemidlet ’dyrkningsfrie bræmmer’, der i vandplanerne primært er tænkt

som en måde at reducere fosfortilførslen til vandløbene på.

Fremtidige forskningsfelterSelvom der således via RISKPOINT projek

tet er høstet megen ny og spændende viden om pesticiders forekomst og virkning i vore vandløb, så er der også basis for at komme nogle spadestik dybere. Vi har i boks 2 opstillet en liste over oplagte forskningsemner, der gerne skulle sikre en mere komplet forståelse af samspillet mellem forskellige typer af menneskeskabte stressfaktorer i vore vandløb, samt en mere komplet forståelse af vandløbsøkosystemernes struktur og dynamik – ikke mindst under påvirkning af pesticider. Det er vores vurdering, at det vil give langt bedre muligheder end nu, for at forebygge og udbedre pesticidernes utilsigtede skadevirkninger.

Referencer/1/ Nørum, U., Friberg, N., Jensen, M.R., Pedersen, J.M. &

Bjerregaard, P. (2010) Behavioural changes in three

species of freshwater macroinvertebrates exposed

Boks 2. Pesticiders forekomst og effekt i vandløb: Vigtige fremtidi-ge forskningsfelterBehovet for ny og forbedret viden om pesticiders betydning i vandløb kan efter vores opfattelse afgrænses til fem forskningsmæssige hoved-områder:• Undersøgelser af tabet fra marker, samt forekomst og skæbne af pe-

sticider, som i særlig grad bindes til partikler og sediment (med sær-ligt fokus på pyrethroider), samt deres økologiske effekter. Der er alle-rede igangsat et indledende studie (PESTFATE) som en del af Miljø-styrelsens Pesticidforskningsprogram.

• Betydningen af dræn og overfladisk afstrømning for transport af pesti-cider (herunder partikelbundet) fra mark til vandløb, herunder betyd-ningen af dyrkningsfrie zoner langs vandløbene.

• Pesticidernes indirekte økologiske effekter, dvs. effekter som kan medføre virkninger igennem flere såkaldte niveauer i vandløbsøkosy-stemet (fx mikrosvampe – smådyr – fisk), og som kun vil være synlige på lang sigt.

• Betydningen af vandløbenes fysiske forhold (substrattyper, vandplan-ter m.m.) for den måde, som fx smådyrene bliver eksponeret for de tilførte pesticider: findes der mindre refugier i vandløbene, hvor små-dyrene kan unddrage sig de værste påvirkninger

• Videreudvikling af screenings værktøjer (biologiske indices – herun-der SPEAR) til identificering af vandløbsstrækninger, hvor der er sær-lig stor belastning med pesticider, og hvor der derfor er behov for en særlig forebyggende indsats. Under Miljøstyrelsens Pesticidforsk-ningsprogram er der desuden iværksat et ’pilotprojekt’ (RAINTOP), hvor en lang række arter af smådyr (med bl.a. forskellig placering og betydning i Dansk Vandløbs Fauna Indeks) screenes for deres føl-somhed over for et modelpyrethroid. Resultater fra dette projekt vil di-rekte kunne indgå ved udviklingen af et dansk pesticidindeks.

Figur 4. Den samlede giftighed for smådyr af fundne pesticider i 14 fynske vandløb som funktion af minimumsbredden af den dyrkningsfri zone i en 1 km lang vandløbskorridor. Da-tapunkter repræsenterer vandprøver, der er indsamlet under 2 episoder med kraftig nedbør. Den sorte linje angiver den egentlige regression, de blå linjer angiver 95 % konfidensbånd, mens de røde linjer angiver 95 % prædiktionsbånd.


PressemeddelelsePlantekongres med fokus på natur, miljø og planlægning

I dagene 10. – 12. januar 2012 er der Plantekongres i Herning. Denne gang med særligt fokus på natur, miljø og planlægning to af dagene.

2011 var første år, vi specifikt henvendte os til folk, der arbejder med miljø, natur og planlægning i det åbne land, og det viste sig at være en stor succes. Mange deltagere har rost kongressen for det faglige indhold, og for at være stedet, hvor debatten stortrives, når både administratorer, forskere, rådgivere og landmænd mødes omkring aktuelle emner. Det oplever man ikke på almindelige faglige kurser, forklarer miljøchef Hans Roust Thysen, Videncentret for Landbrug.

Deltagerne kan vælge mellem 116 sessioner og skræddersy deres helt eget program. Emnerne spænder lige fra serviceeftersyn af § 3områder over kommunikationen mellem jordbrugere og kommuner til debat om det virkelig er kvælstof, der er problemet i vandmiljøet.

Vi har sammensat et stort og fleksibelt program, så alle forventede 2.000 kongresdeltagere kan sammensætte deres eget program efter interesser og behov. Kongressens omfattende program findes på hjemmesiden www.plantekongres.dk<http://www.plantekongres.dk> og indeholder noget for enhver smag, siger Hans Roust Thysen.

Kongressen finder sted i Herning Kongrescenter og organiseres af Videncentret for Landbrug i samarbejde med Aarhus Universitet og Københavns Universitet.

Yderligere oplysningerwww.plantekongres.dk<http://www.plantekongres.dk>Miljøchef Hans Roust Thysen, Videncentret for Landbrug, telefon 30

92 17 01

to the pyrethroid lambdacyhalothrin: Laboratory

and stream microcosm studies. Aquatic Toxicology

98: 328335.

/2/ Styczen, M., WibergLarsen, P. & Aagaard, A. (2003):

Tag pulsen på pesticiderne i vandløb. Vand & Jord

10 (3): 8487.

/3/ Rasmussen, J.J., BaattrupPedersen, A., WibergLarsen,

P., McKnight, U., Kronvang, B. (2011) Buffer strip

width and agricultural pesticide contamination in

Danish lowland streams: Implications for stream and

riparian management. Ecological Engineering 37:

19901997.

/4/ McKnight, U., Rasmussen, J.J., Kronvang, B., Bjerg,

P.L. & Binning, P.J. Integrated assessment of the

impact of chemical stressors on surface water ecosys

tems. Environment International (indsendt).

/5/ Nørum, U., Bjerregaard, P., Friberg, N. & Larsen, S.E.

(2006) Effekter af pulseksponering med pyrethroider

på vandløbsmakroinvertebrater. Bekæmpelsesmid

delforskning fra Miljøstyrelsen nr. 102.

/6/ WibergLarsen, P. & Nørum, U. (2009) Effekter af py

rethroidet lambdacyhalothrin på biologisk struktur,

funktion og rekolonisering i vandløb. Bekæmpelse

smiddelforskning fra Miljøstyrelsen nr. 126.

/7/ Rasmussen, J.J., BaattrupPedersen, A., WibergLarsen,

P., Monberg, R.J. & Kronvang, B. (2011) Impacts of

pesticides and natural stressors on leaf litter decom

position in agricultural streams. Science of the Total

Environment (accepteret til publicering).

/8/ Rasmussen, J.J., Monberg, R.J., BaattrupPedersen, A.,

WibergLarsen, P., Strobel, B. & Kronvang, B. (2011)

Pesticide cocktails and leaf litter decomposition:

Potentially cascading effects. Water, Air & Soil Pollu

tion (indsendt).

/9/ Liess, M. & von der Ohe, P.C. (2005) Analyzing the

effects of pesticides on invertebrate communities in

streams. Environmental Toxicology and Chemistry

24: 954965.

/10/ Rasmussen, J.J., WibergLarsen, P., BaattrupPeder

sen, A., Friberg, N. & Kronvang, B. (2011) Physical

structure of stream microhabitats influences the re

sponse of stream macroinvertebrate communities to

pesticide stress. Environmental Pollution (indsendt).

Jes Jessen Rasmussen er PhD studerende ved Institut for

Bioscience, Aarhus Universitet. Arbejder med effekter af

pesticider og har tidligere arbejdet med klima og stofom

sætning i vandløb.

Peter WibergLarsen er seniorrådgiver ved Institut for

Bioscience, Aarhus Universitet. Arbejder med effekter

af pesticider i vandløb, samt overvågning (NOVANA) af

vandløb.

Annette BaattrupPedersen er seniorforsker ved Institut

for Bioscience, Aarhus Universitet. Arbejder med pestici

ders effekt på plantesamfund i ådale, men primært med

planters økologi i og langs vandløb.

Rikke Juul Monberg er biolog fra Københavns Universitet

med speciale i pesticiders effekt på stofomsætning i

vandløb.

Ursula Solari McKnight er ansat i et PostDoc forløb ved

Danmarks Tekniske Universitet og arbejder med model

lering af transport og effekter af miljøfremmede stoffer.

Brian Kronvang er forskningsprofessor ved Institut for Bi

oscience, Aarhus Universitet. Arbejder bl.a. med transport

af næringsstoffer og pesticider fra oplande til vandløb.

Hans Roust Thysen, Miljøchef.

Paper VII

1

Thresholds for the effects of pesticides on invertebrate communities and leaf

breakdown in stream ecosystems

Ralf B. Schäfer1*, Peter Carsten von der Ohe2, Jes J. Rasmussen3, Ben J. Kefford4, Mikhail A. Beketov5,

Ralf Schulz1 and Matthias Liess5

1 Institute for Environmental Sciences, University Koblenz-Landau, Campus Landau, Fortstrasse 7,

76829 Landau, Germany

2 UFZ, Helmholtz Centre for Environmental Research, Department of Effect-Directed Analysis,

Permoserstrasse 15, 04318 Leipzig, Germany

3 Faculty of Science and Technology, University of Aarhus, Vejlsøvej 25, 8600 Aarhus, Denmark

4 Centre for Environmental Sustainability, School of the Environment, University of Technology

Sydney, Broadway NSW 2007, Australia

5 UFZ, Helmholtz Centre for Environmental Research, Department System-Ecotoxicology,

Permoserstrasse 15, 04318 Leipzig, Germany

* Corresponding author telephone: ++49 (0) 6341 28031536; email: [email protected]

2

Abstract

We compiled data from nine field studies conducted between 1998 and 2010 in Europe, Siberia and

Australia in order to derive thresholds for the effects of pesticides on macroinvertebrate communities

and the ecosystem function leaf breakdown. Dose-response models for the relationship of pesticide

toxicity with the abundance of sensitive macroinvertebrate taxa showed significant differences to

reference sites at levels of 1/1,000 to 1/10,000 of the median acute effect concentration (EC50) for

Daphnia magna, depending on the model specification and whether upstream sections undisturbed by

pesticides were present. Hence, the analysis revealed effects well below the threshold of 1/100 of the

EC50 for D. magna derived from mesocosm studies that corresponds to the threshold incorporated in

the European Union Uniform Principles for pesticide registration. Moreover, the abundances of

sensitive macroinvertebrates in the communities were reduced by 28 to 45% at concentrations related to

1/100 of the EC50 for D. magna. The invertebrate leaf breakdown rate was positively linearly related to

the abundance of pesticide-sensitive macroinvertebrate species in the communities, though only for two

of the three countries examined. We argue that the low effect thresholds observed were not mainly due

to an underestimation of field exposure or confounding factors. From the results gathered we derive that

the current thresholds are not protective for field communities and that risk mitigation measures such as

landscape patches without agricultural disturbances can alleviate effects of pesticides.

Keywords

Risk assessment, pesticides, organic pollutants, toxicity, streams, rivers, field study, ecosystem

function, stream invertebrates

3

Introduction

Freshwater ecosystems are among the most threatened ecosystems in terms of species extinctions and

losses in ecosystem services. One of the major stressors for these ecosystems are pesticides, which are

introduced via point and non-point sources [1]. An efficient protection of freshwater ecosystems

requires the determination of a reliable threshold value for the effects of pesticides. For example, the

Uniform Principles (UP) of the European Union (EU) state that for pesticides “no authorization shall be

granted if the toxicity/exposure ratio for fish and Daphnia is less than 100 for acute exposure [...]” [2].

However, a study conducted in 20 agricultural streams showed a significant change in community

structure already at an acute toxicity/exposure ratio for Daphnia magna in the range of 100 to 1000 [3].

In contrast, a review of mesocosm studies by Wijngaarden et al [4] reported that insecticide

concentrations below the abovementioned toxicity/exposure ratio of 100 that relates to a concentration

of 1/100 of the median effect concentration (EC50) for D. magna are unlikely to cause notable effects.

Hence, these authors suggested that the safety factor of the UP would be protective. However,

mesocosm studies can underestimate effects occurring in the field because (a) they usually exhibit a

lower proportion of species vulnerable to pesticides compared to natural communities [5] and (b) the

joint effects of multiple stressors occurring in the field are rarely considered in mesocosm studies,

though they may influence effect thresholds [6]. The threshold determined from mesocosm studies [4],

i.e. 1/100 of the EC50 for D. magna, has not been compared to a threshold derived from a meta-analysis

based on field studies.

Beside effects on the structure of freshwater communities, pesticides can impede important ecosystem

functions such as leaf breakdown [7] that represents the main energy source in freshwater ecosystems

beside gross primary production [8]. However, to which extent effects on biota propagate to effects on

ecosystem functions has been ranked as one of the most important research questions for the

conservation of biological diversity [9]. Theoretically, the biotic community and ecosystem functions

can be linked in four ways [10]. First, in a near linear way implying effects on biota would lead to a

4

similar decline in ecosystem functions. Second, there may be functional redundancy in the community

and no effects on ecosystem functions would occur up to certain thresholds. Third, the loss of most

species may be compensated whereas the loss of a few so-called keystone species or ecosystem

engineers would result in changes in ecosystem functions. Thus the effects depend on the identity of the

species lost. Fourth, a chemical may alter the functional capacity of species and hence affect ecosystem

functioning without alteration of the community [11]. It is unknown which of these models applies for

the relationship between effects of pesticides on biota and on ecosystem functions and whether this

relationship would be universal.

Species traits have been suggested as a stressor-specific tool in ecological risk assessment [12, 13].

The SPEcies At Risk (SPEAR) indicator for pesticides [3] relies on species traits to calculate the

fraction of pesticide-sensitive species in macroinvertebrate communities. The SPEAR index has been

successfully linked to pesticide toxicity and the leaf breakdown rate in field studies [3, 14]. At the same

time, this indicator showed high discriminatory potential for confounding factors in field studies on

pesticide effects [13] and to be applicable over different biogeographical regions [14-16]. The latter is

especially important because it enables the meta-analysis of studies from different regions.

In this study, we determined thresholds for the effects of pesticides on freshwater ecosystems from a

meta-analysis of field studies. Therefore, we compiled data from various field studies and years in

different regions on the effects of pesticides on freshwater macroinvertebrate communities as detected

using the SPEAR approach as well as on the ecosystem function leaf breakdown. Macroinvertebrates

were selected as structural endpoint since (a) they belong to the most sensitive group of organisms to

pesticides in freshwater communities, (b) trait-based approaches in freshwater ecology are most

advanced for macroinvertebrates and (c) there is a paucity of field studies on the effects of pesticides on

other groups of biota [17]. In addition, we examined how effects on the structure, in terms of the

fraction of pesticide-sensitive species in the communities, are related to effects on the important

ecosystem function of leaf breakdown [17].

5

Experimental section

Selection and description of field studies

The following inclusion criteria were used for the selection of field studies on the effects of pesticides:

(1) at least 5 different streams monitored, (2) selection of pesticides for chemical analysis that are most

likely to represent a risk to macroinvertebrates in the respective region based on recommended pesticide

use information for the respective year (and region) sampled, available toxicity data for D. magna and

results from previous monitoring programs [see 3, 14, 15] and (3) the SPEAR values or leaf breakdown

rates reported. In addition, we included reference sites from studies, where SPEAR values were reported

(Table 1). We focused on studies reporting the trait-based SPEAR indicator, because in contrast to

taxonomical data this indicator has been demonstrated to be applicable over different biogeographical

regions [14, 16]. However, we are not aware of other studies that met the first two criteria and presented

macroinvertebrate community data to assess the effects of pesticides. For example, one study

encompassing sites in 29 different streams [18] was not included since only sediment concentrations for

a limited set of pesticides were reported and the total sediment concentration of the monitored

pesticides was used as a proxy for non-monitored pesticide concentrations. Overall, 9 studies with a

total of 138 sites were included in the present study, of which 7 studies were conducted in different

regions of Europe, and a study in each of Australia and Siberia (Table 1). All studies were conducted

between 1998 and 2010. Except for reference sites and one study in Spain, the pesticide monitoring in

the sites was adjusted to capture episodic runoff events using event-driven water samplers, passive

samplers or suspended particle samplers (Table 1).

6

Table 1. Study regions with number (No.) of sites, biological endpoints reported, number of pesticides measured and Toxic Units (TU) included in this study.

Region No. Sites

Biological endpoints reported

Pesticide monitoring methodsd

No. of pesticides measured

Lowest log TU reported

Highest log TU reported Reference

South Finland 13

SPEARabundance, SPEARPPM abundance and

invertebrate leaf breakdown PS and SPS 10 -5 -4.3 [14]

Brittany, France 16

SPEARabundance, SPEARPPM abundance and

invertebrate leaf breakdown

EWS, PS and SPS 10 -5 -0.4 [14]

Central Germany 20 SPEARabundance EWS 21 -5 -0.7 [3]

Victoria, Australia 24

SPEARpesticides

b and invertebrate leaf

breakdown GWS, PS and

sediment 97 -3.5 -0.2 [7, 15]

Island Funen,

Denmark 14

SPEARpesticides

b and invertebrate leaf

breakdown GWS, EWS and sediment 31 -6.6 -1.7 [19, 20]

Catalonia, Spain 27 SPEAR[%]b GWS 42e -4 -0.8 [21]

Flanders, Belgium 7 SPEAR[%]b - 0f - - [16]

North Germany 11 SPEAR[%]b - 0f - - [16]

Siberia, Russia 6a SPEARorganic

c GWS 0f,g - - [22]

a only reference sites included (sites 3, 5, 6, 7, 9 and 10 of original publication [22]) b reported indicator values were calculated according to Liess & von der Ohe [3]. For this study, the

SPEARpesticides values were calculated from the original data as described below. c for this study the SPEARpesticides indicator was calculated from the original data as described below. d PS = passive sampling, SPS = suspended particle sampling, EWS = event-driven water sampling,

GWS = grab water sampling e in addition, 111 other organic compounds measured f sites were considered as having no pesticide contamination. See references for details. g sum of petrochemicals and surfactants measured

7

Data preparation

Environmental concentrations of pesticides were scaled to acute effects of D. magna using the Toxic

Units (TU) approach of Sprague [23], where the TU for an observed pesticide was calculated as the

compound concentration divided by the respective 48-h median effect concentration (EC50) for D.

magna. Dose-response modelling was used to evaluate the relationship between TU and SPEAR, which

represent pesticide toxicity and community change, respectively. The TUs used here were given as the

maximum TUs of all pesticides across samples per site in the original studies and were reported to have

similar or same explanatory power for biotic endpoints as the sum of TUs of all pesticides across each

or all samples per site [3, 15, 21]. The maximum TU represents the simplest approach because the

estimated pesticide toxicity relies solely on the most toxic pesticide concentration observed per site,

whereas all pesticide concentrations per sample or site contribute to the calculation of the sum of TU.

We used the modified version of the original SPEAR indicator [3] as described in Schäfer et al. [14]

(therein referred to as SPEARPM abundance) in order to compare the indicator values between different

biogeographical regions. For terminological clarity, we refer to this indicator as SPEARpesticides in the

following as suggested by Beketov et al. [24]. For sites for which this version of the SPEAR indicator

was not reported (Table 1), we calculated the indicator according to:

SPEARpesticides

=∑i = 1

n

log (xi+ 1) y

∑i= 1

n

log (xi+ 1)

where n is the number of taxa observed in a sampling site, xi is the abundance of taxon i and y is: 1 if

taxon i is classified as Species At Risk (SPEAR) regarding the traits “physiological sensitivity” and

“dispersal capacity”, otherwise 0. The trait data used were derived from the database associated with the

SPEAR online calculator (http://www.systemecology.eu/SPEAR/index.php).

8

In order to characterize the propagation of effects from pesticide-driven structural changes to

ecosystem functions, the response of the invertebrate-driven leaf breakdown rate (kinvertebrate) [25] to

changes in SPEARpesticides was investigated. The leaf breakdown rate is not stressor-specific and hence

responds to different environmental conditions [26]. In contrast to species traits [27], kinvertebrate varies

over biogeographical regions and would not be expected to be similar across sites without pesticide

contamination due to the influence of other environmental gradients [28]. Indeed, no relevant pesticide

toxicity and only minor variation of SPEARpesticides was detected in the sites from South Finland, whereas

the invertebrate leaf breakdown rate varied strongly between sites in response to temperature [14]. Since

the aim was to examine the link between pesticide-driven community change and invertebrate leaf

breakdown rate, these sites were not considered for further analysis. The invertebrate leaf breakdown

rates from the French, Danish and Australian streams (Table 1) were not comparable as leafs of

different tree species were employed. Therefore, the % change in the leaf breakdown rate kinvertebrate was

calculated for each data set by dividing all values for kinvertebrate by the maximum value obtained from

individually fitted models for kinvertebrate as explained by SPEARpesticides (see below). We used the maximum

value from the models fitted with all data points instead of the maximum value for kinvertebrate from the

respective raw data in order to avoid undue influence of a single data point.

Data analysis

Before analysis the data were divided into sites with and without upstream sections undisturbed from

pesticide contamination, as defined in the original publications, and analysed separately (Table 1). This

was done because the presence of upstream sections undisturbed from pesticide contamination

(hereafter termed undisturbed upstream sections) was demonstrated to alleviate the effects of pesticides

on the macroinvertebrate community as indicated by SPEARpesticides [3, 14, 21]. S-shaped dose-response

curves with TU as concentration and SPEARpesticides as response variable were computed using two-

parameter log-logistic, Weibull I and Weibull II models with the upper limit fixed to the arithmetic

mean of the SPEARpesticides values for reference sites and the lower limit fixed to 0 [29], representing the

9

lowest possible indicator value. In addition, linear, quadratic and cubic regression models were

computed to check for the fit of more parsimonious models. The best-fit model among the s-shaped

dose response models and the polynomial regression models was selected using the Bayesian

Information Criterion (BIC). If a s-shaped dose-response curve represented the best-fit model, we

calculated the effect concentration (EC) for the percentage (p) of reduction in SPEARpesticides for p = 10,

50 and 90%. In addition, the p value was computed for the EC related to the log TU of -2 that is equal

to the safety factor of 100, employed in the UP of the EU. Moreover, we derived the lowest

concentration at which significant differences (significance level = 0.05) to reference sites occur in the = 0.05) to reference sites occur in the

best-fit dose-response model using 95% confidence intervals. Technically, we determined the lowest

concentration for which the 95% confidence interval of the fitted model did not overlap with the 95%

confidence interval for the reference sites. Several of the original studies assigned log TUs of either -5

or -4 to sites where no pesticides were found assuming that this would represent the minimum log TU

for which no pesticide effects would occur (Table 2). Since the minimum TU influences the dose-

response modelling, all analyses were conducted for both minimum reported TUs i.e. we assigned either

a log TU of -5 or -4 to all sites with no pesticide detections. To confirm the results of the dose-response

modelling, analysis of variance (ANOVA) with a priori treatment contrasts was used to identify

significant differences in SPEARpesticides values between reference sites and groups of contaminated sites

in terms of TU. The class boundaries of log TU • -3.5 (reference sites), -3.5 < log TU • -2.5 (lightly

contaminated sites), -2.5 < log TU • -1.5 (moderately contaminated sites) and log TU > -1.5 (highly

contaminated sites) were selected in order to have similar class widths and at least 5 observations in

each class. A similar classification was used in the study of Schäfer et al. [14].

For the examination of the propagation of effects of pesticides on ecosystem functions, the

relationship between %kinvertebrate and SPEARpesticides was modelled using s-shaped dose-response models

and polynomial regression models as described above, except that three-parameter s-shaped dose

response models were fitted since no upper limit could be fixed. First, the modelling was done

10

separately for each country, because in the original studies this relationship had only been examined for

the sites from Brittany, France. Subsequently, the data were modelled jointly. For best-fit linear

regression models, significance of the slope was tested with the t-test and in the case of data from

different countries, analysis of covariance (ANCOVA) was used to detect significant differences

between slopes and intercepts from the countries. ANOVA, ANCOVA and linear regression model

checking (normal distribution of residuals, homoscedasticity and unusual observations) was conducted

according to Fox and Weisberg [30]. All computations and graphics were created with the free and open

source software R (version 2.13.1 for Mac OS X, 10.6.8) [31] including supplemental packages such as

“drc” for dose-response modelling [32].

Table 2. Estimated effect concentrations (EC) in terms of log TU for p = 10, 50 and 90% reduction in SPEARpesticides for the models with and without undisturbed upstream sections (UUS and WUUS, respectively) and with a minimum log TU of -5 (low min.) or -4 (high min.) Estimated EC (in log TU)

p UUS low min. UUS high min. WUUS low min. WUUS high min.

10 -3.3 -2.9 -3.9 -3.2

50 -1.5 -1.3 -1.8 -1.6

90 -0.4 -0.3 -0.4 -0.4

Results

Weibull I models were identified as best-fit dose-response models for TU and SPEARpesticides, except for

sites without undisturbed upstream sections with an assigned minimum log TU of -4. For these data, a

Weibull II model yielded a slightly better fit (Supporting Information Table S1). However, Weibull I

models were fitted for all cases in order to facilitate comparison of the model parameters (Figure 1,

Supporting Information Figure S1 and Table S2). The estimated EC10 for the different models ranged

from a log TU of -2.9 to a log TU of -3.9 depending on a) the availability of undisturbed upstream

sections and b) which minimum TU was assigned (Table 2). The estimated EC50 and EC90 were similar

11

for the two minimum TUs, with a maximum difference in TU of 0.2 and 0.1 log units (Table 2). For an

EC related to a log TU of -2, the fraction of species at risk in the communities in terms of abundance

was reduced by 35% and 45% in sites with undisturbed upstream sections for models with a minimum

log TU of -4 and -5, respectively (Figure 1, Figure S1). In sites without undisturbed upstream sections,

this EC corresponded to 28% and 35% reduction in SPEARpesticides, respectively (Figure 1, Figure S1).

Significant differences (non-overlapping 95% confidence intervals) to reference sites were observed for

log TUs • -2.7 and -3.3 in sites with undisturbed upstream sections and for log TUs • -3.1 and -3.8 in

sites without undisturbed upstream sections when assigning a minimum log TU of -4 and -5,

respectively. Similar results were obtained for the ANOVAs, in which all sites in classes with a log TU

> -3.5 exhibited a significant difference (all p < 0.05) to reference sites (Figure 2).

Figure 1. Dose-response curves with 95% confidence bands (grey) for the relationship between log TU and SPEARpesticides for sites with undisturbed upstream sections (a) and for sites without undisturbed upstream sections (b). Reference sites were assigned a minimum log TU of -4. A small amount of random noise (twice the smallest difference between non-identical adjacent TU values) was added (jittering) to the TU values in order to show all data points in the plot. This affected primarily the sites with minimum TUs.

12

The linear models exhibited the best-fit for the relationships between % kinvertebrate and SPEARpesticides,

except for sites from Brittany, France, for which a Weibull II model yielded a slightly better fit

(Supporting Information Table S1). However, the slope for linear regression models was only

significant for the sites from Victoria, Australia (r2 = 0.46, p < 0.001, n = 23) and Brittany, France (r2 =

0.87, p < 0.001, n = 11), but not for the sites from Denmark (r2 = 0.09, p = 0.31, n = 13), for which no

plausible relationship between SPEARpesticides and %kinvertebrate could be established (data not shown). The

sites from Denmark were not included in the joint dose-response modelling, because the aim was to

derive a joint relationship between SPEARpesticides and %kinvertebrate. The best-fit model for the joint data

from France and Australia was linear (Supporting Information Table S1) and exhibited a good fit

between %kinvertebrate and SPEARpesticides (r2 = 0.51, p < 0.001, n = 34) (Figure 3). In ANCOVA, the

intercepts for data from Brittany and Victoria were significantly different (p = 0.02), whereas the slopes

Figure 2. Arithmetic mean of SPEARpesticides values with standard errors for different classes of toxic units (TU). The sampling sites were divided into sites with undisturbed upstream sections (filled points) and sites without undisturbed upstream sections (open points). Sample sizes of the different classes were 46 and 9 sites with a log TU • -3.5, 20 and 13 sites with -3.5 < log TU • -2.5, 17 and 16 sites with -2.5 < log TU • -1.5 and 6 and 11 sites with log TU > -1.5 for sites with undisturbed upstream sections and sites without undisturbed upstream sections, respectively. All classes with TUs > -3.5 were significantly different (all p < 0.001) to reference sites in ANOVA with treatment contrasts.

13

exhibited no significant differences (p = 0.25). However, if the data of each region were autoscaled

before ANCOVA, neither the intercepts (p = 0.89) nor slopes (p = 0.15) were significantly different.

Discussion

Effect thresholds for macroinvertebrate communities

In our analysis, pesticide effects on the abundance of sensitive invertebrates were found at TUs below

0.01. Concentrations related to the safety factor incorporated in the UP resulted in a 28% to 45%

decline in the abundance of sensitive taxa, depending on whether or not undisturbed upstream sections

were available and which minimum TU was selected in modelling (Figure 1, Figure S1). Similarly, both

of the latter factors (presence of undisturbed upstream sections and minimum TU) influenced the

estimated effect concentrations (Table 2) and the concentration at which significant differences to

reference sites occurred. The models with a higher minimum TU exhibited a better fit in terms of the

BIC for sites with undisturbed upstream sections and a relatively similar fit for sites without

undisturbed upstream sections compared to models with a lower minimum TU (Table S1).

Figure 3. Linear regression model for the relationship between SPEARpesticides and %kinvertebrate and for sampling sites from Brittany, France and Victoria, Australia. The linear model explained 51% of the variation (p < 0.001, n = 34). Note that the intercepts for data from Brittany and Victoria were significantly different (p = 0.02) in ANCOVA (see Results for details).

14

Nevertheless, more field data in the log TU range of -3 to -5 would be needed to substantiate a selection

between both minimum TUs. Irrespective of which minimum TU was employed in modelling, the

effect threshold was determined to be approximately one order of magnitude lower (log TU of -3) than

the safety factor of the UP for sites with undisturbed upstream sections and even lower (log TU of -3.1

to -3.8) for sites without undisturbed upstream sections. However, field studies and results from the

joint analysis of studies with differing methodologies are subject to random and systematic uncertainties

that can lead to wider confidence bands or bias in the dose-response models and would consequently

affect the derived effect threshold. Several uncertainties were identified as random (Supporting

Information Table S3) and presumably resulted in wider confidence bands of the fitted dose-response

curves (Figure 1, Figure S1) as for example, the joint analysis of data obtained from different countries,

years and pesticide sampling methods (Table1). Nevertheless, the variation of observations between the

studies was not much higher than the variation of observations within the individual studies (Figure 1,

Figure S1). Moreover, some of the variation in the relationship between pesticide toxicity and

SPEARpesticides may result from differences in the dose-response relationship of individual compounds i.e.

concentrations relating to for example 1/100 of their EC50 for D. magna may exert different effects on

the abundance of sensitive taxa. Two studies found that the toxicity of a range of different organic

toxicants also explained between 68% and 87% of the variance in terms of r2 in SPEAR [21, 22, 33],

suggesting that the use of toxic units for D. magna as benchmark for the toxicity of different organic

toxicants is adequate and that the associated uncertainty is of minor importance.

Two sources of uncertainty could result in a systematic bias in the derived effect threshold. First, the

underestimation of pesticide toxicity due to underestimated pesticide concentrations or the non-

measurement of ecotoxicologically relevant compounds would lead to a left shift of the dose-response

curve and consequently a decrease in the effect threshold. If pesticide toxicity was underestimated in the

field studies, there should have been notable differences between the most comprehensive (in terms of

employed sampling methods (Table 1) and measured compounds (97)) study on pesticide effects in

15

Victoria, Australia [15] and the other studies included in our analysis, which were not observed (Figure

1, Figure S1). In addition, underestimation of the real concentrations by a factor of 10 to 100 would be

required in order to yield similar effect thresholds than incorporated in the UP. We consider an

underestimation of this order of magnitude as highly unlikely given that most of the field studies

employed the best available sampling techniques, especially targeted at capturing episodic pesticide

input (Table 1). Finally, it would not be expected that all compounds are underestimated in equal

measure and underestimation should therefore increase the variability in pesticide toxicity that relied on

different compounds in all included studies. In fact, all studies exhibited a very good fit (all r2 from

linear models between 0.62 and 0.87) of pesticide toxicity with the respective biotic endpoint.

Second, the response of the SPEAR index to a confounding factor that is highly correlated with

pesticide toxicity would lead to a decrease in effect thresholds (Supporting Information Table S3). The

original studies included in our analysis (Table 1) and two recent studies [34, 35] identified the toxicity

of a stressor as the most important explanatory variable for the respective version of the SPEAR index,

whereas 8 and 9 respectively measured confounding factors exhibited no explanatory power for

SPEAR. Hence, there is only low uncertainty that the indicated effects were not due to pesticide

toxicity. Higher uncertainty remains regarding the mechanism causing the observed effects of

pesticides. Although previous studies (1) found a slightly better relationship for the maximum TU and

biotic endpoints than for the sum of TUs [3, 15, 21] and (2) attributed toxicity of pesticide mixtures to

only a few pesticides [36, 37], other factors may contribute to the observed effects as well. For example,

co-occurring stressors could increase the vulnerability of field communities to pesticides or interact with

pesticide stress [35, 38]. Moreover, sites with higher maximum TUs could more frequently receive

pesticide pulses and this repeated exposure could contribute to the observed effects. Studies with a high

temporal and spatial resolution would be needed to clarify the mechanisms [cf. 39]. Overall, we suggest

that there is low uncertainty that our derived effect threshold for effects of pesticides on

16

macroinvertebrate communities is too low and we therefore conclude that the safety factor incorporated

in the EU Uniform Principles is not protective for freshwater ecosystems.

Causes for differences in effect thresholds derived from field and mesocosm studies

While in laboratory experiments adverse effects on populations have been observed at concentrations

related to a log TU of -4 [40, 41], mesocosm studies have rarely reported effects on macroinvertebrate

communities below a log TU of -2 and a review of mesocosm studies suggested that this threshold

would be protective in the field [4]. Several reasons may explain why effects in the field can be

observed at lower levels [cf. 42]. First, in our study we re-analysed pooled data from different field

studies, which increases the statistical power to detect effects, whereas the review of mesocosm studies

did not re-analyse pooled mesocosm data [4]. Furthermore, our analysis focused on the change of the

abundance of sensitive species in the communities using the trait-based SPEAR approach, whereas

previous mesocosm studies were primarily evaluated using taxonomic changes in the whole community

[35]. In fact, a recent study showed that the SPEAR approach detected effects in mesocosms up to a

factor of 1000 lower than the commonly employed data analysis method of principal response curves

for taxonomic data [35]. Beside these methodical reasons, macroinvertebrate communities in

mesocosms usually contain a lower proportion of vulnerable (e.g. high physiological sensitivity, long

generation time) species than field communities [5] and mesocosm studies rarely consider repeated

exposures and the joint effects of different stressors both of which can increase the effects of pesticides

in the field [35, 36, 38, 43]. Moreover, sublethal long-term effects on merolimnic insects or slow-

reproducing taxa may not be detected in mesocosm studies, which rarely exceed a study period of

several months. We conclude that the difference in effect thresholds derived from field and mesocosm

studies points to an underestimation of effect thresholds in most previous mesocosm studies.

17

Effects thresholds for the ecosystem function of leaf breakdown

The relationship between the community structure in terms of SPEARpesticides and the percentage of

invertebrate leaf breakdown was linear for the streams from Brittany, France and Victoria, Australia

(Figure 3). This means that of the several suggested links between the community structure and

ecosystem functions (see Introduction), pesticide effects on the abundance of sensitive

macroinvertebrates seem to translate to a similar effect on the breakdown rate of leafs by invertebrates.

Hence, in these regions is no greater tolerance of this important ecosystem function to pesticide

contamination and the effect thresholds for the abundance of SPEAR taxa may also apply. Given that

regional case studies on the relationship between pesticides and ecosystem functions are scarce, this

result may not hold for different ecosystem functions [7], and other regions. In fact, no plausible

relationship between SPEARpesticides and the invertebrate leaf breakdown rate was found for the Danish

sites. Similarly, the original study on the Danish streams only reported a statistically significant

relationship between pesticide toxicity and microbial leaf breakdown but not with invertebrate leaf

breakdown [20]. Hence, although pesticide toxicity lead to community change in terms of SPEARpesticides

in the Danish sites (Figure 1), this did not translate to effects on the ecosystem function of leaf

breakdown. This can be explained by the domination of the shredder community by Gammarus pulex,

which is a rather tolerant species due to its ecological traits and is consequently not classified as SPEAR

[20]. In fact, the density of Gammarus pulex was significantly correlated to the leaf breakdown rate (p =

0.02, t-test for Pearson correlation, n = 13), but did not respond to pesticide toxicity (p = 0.29, t-test for

Pearson correlation, n = 13). Thus, the effect threshold for the invertebrate leaf breakdown presumably

depends on the composition of the shredder community, and if non-SPEAR taxa such as Gammarus

pulex dominate a shredder community, there may be functional redundancy up to a certain threshold

before pesticides affect invertebrate leaf breakdown. Finally, the question is to which extent a temporal

difference between pesticide application and leaf input from deciduous trees affects the relationship

between pesticide-driven structural changes and invertebrate leaf breakdown [cf. 44]. The studies in

18

France and Denmark were conducted in the period of peak insecticide application in these regions,

which precedes the period of leaf abscission and consequently the main input of leafs (late autumn) by

several months. Although it is known that community alterations can persist over months [3, 17], it

remains to be shown that the invertebrate leaf breakdown is affected outside of the main season of

pesticide application. However, for streams receiving a relatively constant leaf input from evergreen

forests as for example the streams in the Australian study [45], the influence of seasonality on the

effects of pesticides on invertebrate leaf breakdown should be of minor importance.

Relevance for ecological risk assessment of aquatic ecosystems

The thresholds obtained in our study may be relevant for pesticides and other organic compounds where

macroinvertebrates represent the most sensitive group of taxa. In a study on the concentrations of 331

organic toxicants in large rivers of North Germany, invertebrates were considered as most sensitive for

110 compounds, among them many insecticides and fungicides, whereas algae and fish represented the

most sensitive group for 142 and 79 compounds, respectively [46]. Another study reported that

invertebrates were most sensitive for 225 organic toxicants, whereas algae and fish exhibited highest

sensitivity for 158 and 104 organic toxicants, respectively [33]. In this study, an effect threshold of

1/1000 of the acute EC50 for D. magna was suggested for the derivation of environmental quality

standards (EQS) for river basin specific pollutants, based on an analysis of macroinvertebrate

biomonitoring and chemical monitoring data. Hence, effect thresholds for macroinvertebrates would

also be protective of other aquatic organisms for a wide range of compounds. The relatively good

relationship between pesticide toxicity in terms of TU and SPEAR in our study is remarkable,

considering that the macroinvertebrate data originated from different regions in Europe, Siberia and

Australia. Our study therefore supports the use of trait-based approaches in risk assessment to identify

the impact of anthropogenic stressors on a continental or even global scale [10, 47].

19

Pesticides reduced the abundance of sensitive macroinvertebrate taxa at concentrations below the

safety factor of the UP. Without knowing the temporal and spatial dimension of the reduction in the

abundance of sensitive macroinvertebrate populations in this study, it is not possible to decide whether

the observed effects on the communities were transient or long-term, defined as no recovery until the

spraying period in the consecutive year. In the latter case, the effects would be unacceptable for the

requirements of the EU directive for the placement of plant protection products on the market [2, 48].

However, we suggest that current exposure of freshwater ecosystems to pesticides may be unacceptable

for the requirements of the EU directive. First, one field study showed long-term effects i.e. that no

recovery of the communities occurred until the pre-spraying period of the following year [3]. Second,

given that pesticides are widely applied in agriculture, which represents the dominant land use in the

EU and elsewhere, and that pesticides frequently occur in streams and rivers in concentrations above

effect thresholds [17], the associated reduction in the abundance of sensitive taxa may lead to losses in

biodiversity on a regional scale ( -diversity) as also indicated by other studies [49, 50]. Since a recent

EU Directive required that the risks for biodiversity from pesticide use should be minimised [51], more

pesticide mitigation measures may be needed to comply with this Directive. Our study highlighted on

the basis of a comprehensive data set that undisturbed upstream sections can buffer against adverse

effects of pesticides on the macroinvertebrate community, especially under low pesticide contamination

as indicated by higher effect thresholds (Table 2). Hence, together with other risk mitigation measures

such as pesticide use reduction, buffer strips and vegetated treatment systems [19, 52, 53], the

conservation and increase of landscape patches without agricultural disturbance may alleviate the

effects of pesticides in aquatic ecosystems.

Acknowledgements

The authors thank all persons involved in the field studies analysed in this paper. Mirco Bundschuh and

three anonymous reviewers gave valuable comments that improved the quality of the manuscript. We are

20

grateful to Sabine Duquesne for clarifying the legal situation for pesticide authorisation. PCO was

financially supported through a Deutsche Forschungsgemeinschaft (DFG) postdoctoral fellowship (PAK

406/1). JR received funding from the Danish Research Council (RISKPOINT, Grant No 2104-07-0035).

A visit by BJK to Germany was funded by the University of Technology Sydney.

Author contributions

Study design: RBS, PCO, BJK, MB, ML; Provision of data: RBS, PCO, JR, MB, ML; Data analysis:

RBS; Discussion and interpretation of results: all; Drafting of manuscript: RBS; Revising manuscript:

MB, BJK, PCO, ML.

Supporting Information

A figure with the relationship between pesticide toxicity and SPEARpesticides for a minimum log TU of -5,

a table with the goodness of fit results for the different models, a table with the parameters of the fitted

dose-response models for pesticide toxicity and SPEARpesticides and a table regarding the sources and

potential impact of uncertainties.

References

1. Millenium Ecosystem Assessment Ecosystems and Human Well-being: Synthesis; Island Press: Washington, DC, 2005.

2. European Council Directive 91/414/EEC of 15 July 1991 concerning the placing of plant protection products on the market.

3. Liess, M.; von der Ohe, P. C., Analyzing effects of pesticides on invertebrate communities in streams. Environ. Toxicol. Chem. 2005, 24, (4), 954-965.

4. Van Wijngaarden, R. P. A.; Brock, T. C. M.; Van Den Brink, P. J., Threshold levels for effects of insecticides in freshwater ecosystems: A review. Ecotoxicology 2005, 14, (3), 355-380.

21

5. Beketov, M.; Schäfer, R. B.; Marwitz, A.; Paschke, A.; Liess, M., Long-term stream invertebrate community alterations induced by the insecticide thiacloprid: Effect concentrations and recovery dynamics. Sci. Total Environ. 2008, 405, 96-108.

6. Townsend, C. R.; Uhlmann, S. S.; Matthaei, C. D., Individual and combined responses of stream ecosystems to multiple stressors. J. Appl. Ecol. 2008, 45, (6), 1810-1819.

7. Schäfer, R. B.; Bundschuh, M.; Rouch, D. A.; Szöcs, E.; von der Ohe, P. C.; Pettigrove, V.; Schulz, R.; Nugegoda, D.; Kefford, B. J., Effects of pesticide toxicity, salinity and other environmental variables on selected ecosystem functions in streams and the relevance for ecosystem services. Sci. Total Environ. 2011, doi:10.1016/j.scitotenv.2011.05.063.

8. Webster, J. R., Spiraling down the river continuum: stream ecology and the U-shaped curve. J. N. Am. Benthol. Soc. 2007, 26, (3), 375-389.

9. Sutherland, W. J.; Adams, W. M.; Aronson, R. B.; Aveling, R.; Blackburn, T. M.; Broad, S.; Ceballos, G.; Cote, I. M.; Cowling, R. M.; GA, D. A. F.; Dinerstein, E.; Ferraro, P. J.; Fleishman, E.; Gascon, C.; Hunter Jr, M.; Hutton, J.; Kareiva, P.; Kuria, A.; Macdonald, D. W.; Mackinnon, K.; Madgwick, F. J.; Mascia, M. B.; McNeely, J.; Milner-Gulland, E. J.; Moon, S.; Morley, C. G.; Nelson, S.; Osborn, D.; Pai, M.; Parsons, E. C.; Peck, L. S.; Possingham, H.; Prior, S. V.; Pullin, A. S.; Rands, M. R.; Ranganathan, J.; Redford, K. H.; Rodriguez, J. P.; Seymour, F.; Sobel, J.; Sodhi, N. S.; Stott, A.; Vance-Borland, K.; Watkinson, A. R., One Hundred Questions of Importance to the Conservation of Global Biological Diversity. Conservation biology : the journal of the Society for Conservation Biology 2009.

10. Cadotte, M. W.; Carscadden, K.; Mirotchnick, N., Beyond species: functional diversity and the maintenance of ecological processes and services. J. Appl. Ecol. 2011, 48, (5), 1079-1087.

11. Kefford, B. J.; Schäfer, R. B.; Metzeling, L., Risk assessment of salinity and turbidity in Victoria (Australia) to stream insects' community structure does not always protect functional traits. Sci. Total Environ. 2011, doi:10.1016/j.scitotenv.2011.05.056.

12. Van den Brink, P.; Alexander, A. C.; Desrosiers, M.; Goedkoop, W.; Goethals, P. L. M.; Liess, M.; Dyer, S. D., Traits-Based Approaches in Bioassessment and Ecological Risk Assessment: Strengths, Weaknesses, Opportunities and Threats. Integr. Environ. Assess. Manage. 2011, 7, (2), 198-208.

13. Liess, M.; Schäfer, R. B.; Schriever, C. A., The footprint of pesticide stress in communities - species traits reveal community effects of toxicants. Sci. Total Environ. 2008, 406, (3), 484-490.

14. Schäfer, R. B.; Caquet, T.; Siimes, K.; Mueller, R.; Lagadic, L.; Liess, M., Effects of pesticides on community structure and ecosystem functions in agricultural streams of three biogeographical regions in Europe. Sci. Total Environ. 2007, 382, (2-3), 272-285.

15. Schäfer, R. B.; Pettigrove, V.; Rose, G.; Allinson, G.; Wightwick, A.; von der Ohe, P. C.; Shimeta, J.; Kühne, R.; Kefford, B. J., Effects of pesticides monitored with three sampling methods in 24 sites on macroinvertebrates and microorganisms. Environ. Sci. Technol. 2011, 45, (4), 1665-1672.

16. von der Ohe, P. C.; Prüß, A.; Schäfer, R. B.; Liess, M.; Deckere, E. d.; Brack, W., Water quality indices across Europe - a comparison of the good ecological status of five river basins. J. Environ. Monit. 2007, 9, 970-978.

22

17. Schäfer, R. B.; van den Brink, P. J.; Liess, M. Impacts of pesticides on freshwater ecosystems. In Ecological impacts of toxic chemicals; Sanchez-Bayo, F.; van den Brink, P.; Mann, R. M., Eds.; Bentham: Bussum, NL, 2011.

18. Friberg, N.; Lindstrom, M.; Kronvang, B.; Larsen, S. E., Macroinvertebrate/sediment relationships along a pesticide gradient in Danish streams. Hydrobiologia 2003, 494, (1-3), 103-110.

19. Rasmussen, J. J.; Baattrup-Pedersen, A.; Wiberg-Larsen, P.; McKnight, U. S.; Kronvang, B., Buffer strip width and agricultural pesticide contamination in Danish lowland streams: Implications for stream and riparian management. Ecol. Eng. 2011, 37, (12), 1990-1997.

20. Rasmussen, J. J.; Wiberg-Larsen, P.; Baattrup-Pedersen, A.; Monberg, R. J.; Kronvang, B., Impact of pesticides and natural stressors on leaf litter decomposition in agricultural streams. Sci Total Environ 2011, doi:10.1016/j.scitotenv.2011.11.057.

21. von der Ohe, P. C.; de Deckere, E.; Prüß, A.; Munoz, I.; Wolfram, G.; Villagrasa, M.; Ginebreda, A.; Hein, M.; Brack, W., Toward an integrated assessment of the ecological and chemical status of European river basins. Integr. Environ. Assess. Manage. 2009, 5, (1), 50-61.

22. Beketov, M. A.; Liess, M., An indicator for effects of organic toxicants on lotic invertebrate communities: Independence of confounding environmental factors over an extensive river continuum. Environ. Pollut. 2008, 156, (3), 980-987.

23. Sprague, J. B., Measurement of pollutant toxicity to fish, II-Utilizing and applying bioassay results. Water Res. 1970, 4, (1), 3-32.

24. Beketov, M. A.; Foit, K.; Schäfer, R. B.; Schriever, C. A.; Sacchi, A.; Capri, E.; Biggs, J.; Wells, C.; Liess, M., SPEAR indicates pesticide effects in streams - Comparative use of species- and family-level biomonitoring data. Environ. Pollut. 2009, 157, 1841-1848.

25. Benfield, E. F. Decomposition of leaf material. In Methods in Stream Ecology; Hauer, F. R.; Lamberti, G. A., Eds.; Academic Press: San Diego, 2007.

26. Gessner, M. O.; Chauvet, E., A case for using litter breakdown to assess functional stream integrity. Ecol. Appl. 2002, 12, (2), 498-510.

27. Statzner, B.; Bis, B.; Dolédec, S.; Usseglio-Polatera, P., Perspectives for biomonitoring at large spatial scales: A unified measure for the functional composition of invertebrate communities in European running waters. Basic Appl. Ecol. 2001, 2, 73-85.

28. Allan, J. D.; Castillo, M. M., Stream Ecology: Structure and Function of Running Waters. Springer: Dordrecht, 2007.

29. Ritz, C., Toward a unified approach to dose-response modeling in ecotoxicology. Environ. Toxicol. Chem. 2010, 29, (1), 220-229.

30. Fox, J.; Weisberg, S., An R companion to applied regression. 2. ed.; Sage: Los Angeles, 2011.

31. R Development Core Team R: A language and environment for statistical computing, reference index version 2.13.1. www.r-project.org (October 5 2011),

23

32. Ritz, C.; Streibig, J. C., Bioassay analysis using R. J Stat Softw 2005, 12, (5), 1-22.

33. von der Ohe, P. C.; Dulio, V.; Slobodnik, J.; De Deckere, E.; Kühne, R.; Ebert, R.-U.; Ginebreda, A.; De Cooman, W.; Schüürmann, G.; Brack, W., A new risk assessment approach for the prioritization of 500 classical and emerging organic microcontaminants as potential river basin specific pollutants under the European Water Framework Directive. Sci. Total Environ. 2011, 409, (11), 2064-2077.

34. Schäfer, R. B.; Kefford, B.; Metzeling, L.; Liess, M.; Burgert, S.; Marchant, R.; Pettigrove, V.; Goonan, P.; Nugegoda, D., A trait database of stream invertebrates for the ecological risk assessment of single and combined effects of salinity and pesticides in South-East Australia. Sci. Total Environ. 2011, 409, (11), 2055-2063.

35. Liess, M.; Beketov, M., Traits and stress: keys to identify community effects of low levels ot toxicants in test systems. Ecotoxicology 2011, 20, (6), 1328-1340.

36. Belden, J. B.; Gilliom, R. J.; Martin, J. D.; Lydy, M. J., Relative toxicity and occurrence patterns of pesticide mixtures in streams draining agricultural Integr. Environ. Assess. Manage. 2007, 3, (1), 90-100.

37. Price, P. S.; Han, X., Maximum cumulative ratio (MCR) as a tool for assessing the value of performing a cumulative risk assessment. Int. J. Env. Res. Public Health 2011, 8, (6), 2212-25.

38. Holmstrup, M.; Bindesbol, A. M.; Oostingh, G. J.; Duschl, A.; Scheil, V.; Kohler, H. R.; Loureiro, S.; Soares, A. M.; Ferreira, A. L.; Kienle, C.; Gerhardt, A.; Laskowski, R.; Kramarz, P. E.; Bayley, M.; Svendsen, C.; Spurgeon, D. J., Interactions between effects of environmental chemicals and natural stressors: a review. Sci Total Environ 2010, 408, (18), 3746-62.

39. Liess, M.; Schulz, R., Linking insecticide contamination and population response in an agricultural stream. Environ. Toxicol. Chem. 1999, 18, 1948-1955.

40. Liess, M., Population response to toxicants is altered by intraspecific interaction. Environ. Toxicol. Chem. 2002, 21, (1), 138-142.

41. Liess, M.; Schulz, R., Chronic effects of short-term contamination with the pyrethroid insecticide fenvalerate on the caddisfly Limnephilus lunatus. Hydrobiologia 1996, 324, (2), 99-106.

42. Carpenter, S. R., Microcosm experiments have limited relevance for community and ecosystem ecology. Ecology 1996, 77, (3), 677-680.

43. Heugens, E.; Hendriks, A.; Dekker., T.; Van Straalen, N. M.; Admiraal, W., A review of the effects of multiple stressors on aquatic organisms and analysis of uncertainty factors for use in risk assessment. Crit. Rev. Toxicol. 2001, 31, (3), 247-284.

44. Bergfur, J., Seasonal variation in leaf-litter breakdown in nine boreal streams: implications for assessing functional integrity. Fundam. Appl. Limnol. 2007, 169, (4), 319-329.

45. Tyler, P.; Williams, B. Lakes and Rivers. In Ecology: An Australian perspective; Attiwill, P.; Wilson, B., Eds.; Oxford University Press: South Melbourne, 2003.

24

46. Schäfer, R. B.; von der Ohe, P.; Kühne, R.; Schüürmann, G.; Liess, M., Occurrence of 331 organic pollutants in four rivers of North Germany between 1994 and 2004 and risk assessment for algae, invertebrates and fish. Environ. Sci. Technol. 2011, 45, (14), 6167-6174.

47. Statzner, B.; Beche, L. A., Can biological invertebrate traits resolve effects of multiple stressors on running water ecosystems? Freshwat. Biol. 2010, 55, 80-119.

48. European Commission Guidance Document Sanco/3268/2001 rev.4 on Aquatic Ecotoxicology in the context of the Directive 91/414/EEC.

49. Gibbs, K. E.; Mackey, R. L.; Currie, D. J., Human land use, agriculture, pesticides and losses of imperiled species. Divers. Distrib. 2009, 15, (2), 242-253.

50. Geiger, F.; Bengtsson, J.; Berendse, F.; Weisser, W. W.; Emmerson, M.; Morales, M. B.; Ceryngier, P.; Liira, J.; Tscharntke, T.; Winqvist, C.; Eggers, S.; Bommarco, R.; Part, T.; Bretagnolle, V.; Plantegenest, M.; Clement, L. W.; Dennis, C.; Palmer, C.; Onate, J. J.; Guerrero, I.; Hawro, V.; Aavik, T.; Thies, C.; Flohre, A.; Hanke, S.; Fischer, C.; Goedhart, P. W.; Inchausti, P., Persistent negative effects of pesticides on biodiversity and biological control potential on European farmland. Basic Appl. Ecol. 2010, 11, (2), 97-105.

51. European Commission Directive 2009/128/EC of the European Parliament and of the Council of 21 October 2009 establishing a framework for Community action to achieve the sustainable use of pesticides.

52. Reichenberger, S.; Bach, M.; Skitschak, A.; Frede, H. G., Mitigation strategies to reduce pesticide inputs into ground- and surface water and their effectiveness; A review. Sci. Total Environ. 2007, 384, (1-3), 1-35.

53. Stehle, S.; Elsaesser, D.; Gregoire, C.; Imfeld, G.; Passeport, E.; Payraudeau, S.; Schäfer, R. B.; Tournebize, J.; Schulz, R., Pesticide Risk Mitigation by Vegetated Treatment Systems: A Meta-analysis. J. Environ. Qual. 2011, 40, (4), 1068-1080.

Paper VIII

1

Integrated assessment of the impact of chemical stressors on surface

water ecosystems

Ursula S. McKnighta*, Jes J. Rasmussenb, Brian Kronvangb, Poul, L. Bjerga, Philip J.

Binninga

a Department of Environmental Engineering, Technical University of Denmark,

Miljoevej Building 113, 2800 Kgs. Lyngby, Denmark

b Aarhus University, Department of Bioscience, Vejlsoevej 25, 8600 Silkeborg, Denmark

*Corresponding author: Tel.: +45 4525 1412; Fax: +45 4593 2850; E-mail:

[email protected]

Other author E-mail addresses: [email protected]; [email protected]; [email protected];

[email protected]

2

Abstract

The release of chemicals such as chlorinated solvents, pesticides and other xenobiotic

organic compounds to streams, either from contaminated sites, accidental or direct

application/release, is a significant threat to water resources. In this paper, different

methods for evaluating the impacts of chemical stressors on stream ecosystems are

evaluated for a stream in Denmark where the effects of major physical habitat

degradation can be disregarded. The methods are: (i) the Danish Stream Fauna Index,

(ii) Toxic Units (TU), (iii) SPEAR indices, (iv) Hazard Quotient (HQ) index and (v)

AQUATOX, an ecological model. The results showed that the hydromorphology,

nutrients, biological oxygen demand and contaminants (pesticides and trichloroethylene

from a contaminated site) originating from groundwater do not affect the good

ecological status in the stream. In contrast, the evaluation by the novel SPEARpesticides

index and TU indicated that the site is far from obtaining good ecological status - a

direct contradiction to the ecological index currently in use in Denmark today - most

likely due to stream sediment-bound pesticides arising from the spring spraying season.

In order to generalise the findings of this case study, the HQ index and AQUATOX

were extended for additional compounds, partly to identify potential compounds of

concern, but also to determine thresholds where ecological impacts could be expected to

occur. The AQUATOX results indicate that tetrachloroethylene, glyphosate, and

potentially naphthalene, could adversely impact ecosystems at typical concentrations

reported in streams. The threshold concentrations produced by AQUATOX were

significantly lower than values obtained using the HQ index. The HQ index identified

naphthalene as having a threshold value closest to concentrations reported in streams.

The results demonstrate that some commonly used methods for the assessment of

3

ecological impact are not sufficient for capturing - and ideally separating - the effects of

all anthropogenic stressors affecting ecosystems.

Key words ecological status; EU Water Framework Directive; benthic macroinvertebrates;

contaminated sites; SPEAR index; AQUATOX

1. Introduction

Due to increasing global exploitation of both stream water and groundwater resources, it

is essential to obtain a better understanding of human impacts on, and the connections

between these two systems and the roles they play in maintaining water quality. Society

is becoming increasingly dependent on groundwater for meeting its industrial,

agricultural and domestic water needs, and anthropogenic impacts due to the release of

xenobiotic organic contaminants and intensive use of agricultural chemicals has led to

the degradation of this resource (Hose, 2005). To address this, the EU Water

Framework Directive (WFD) requires member states to evaluate all types of

contamination sources within a watershed in order to assess their direct impact on water

quality and ecosystem health (Hinsby et al., 2008; Theodoropoulos and Iliopoulou-

Georgudaki, 2010; Von der Ohe et al., 2007; Whiteman et al., 2010).

Fulfilling the requirements of the EU WFD is challenging, as surface water

ecosystems are often impacted by a multitude of co-occurring stressors (Von der Ohe et

al., 2011; Sánchez-Montoya et al., 2010; Thrush et al., 2008), and it is difficult to

evaluate and compare them. Traditional approaches for managing aquatic resources

often fail to account for all the potential effects of anthropogenic disturbances on the

biota. Thus, the applicability of current and novel methods for determining ecological

status must be re-assessed. Here we focus specifically on benthic macroinvertebrates,

4

one of the four EU WFD biological quality elements used to characterise the ecological

quality and chemical toxicity of streams, and five methods are utilised: the (i) Danish

Stream Fauna Index (DSFI), (ii) Toxic Units, (iii) SPEcies At Risk (SPEAR) indices,

(iv) U.S. EPA Hazard Quotient (HQ) index, and (v) an ecological model, AQUATOX,

also developed by the U.S. EPA.

Chlorinated solvents, such as trichloroethylene (TCE), and pesticides are among the

most prevalent and serious contaminants of surface and groundwater resources,

particularly in industrialised countries with intensive agriculture such as Denmark

(Brüsch, 2007; Danish EPA, 2010a; Henriksen et al., 2008; Janniche et al., 2011). To

evaluate the effects of these stressors, field sites are needed where the effects of other

anthropogenic stressors can be disregarded. Rasmussen et al. (2011a) showed that the

effects of diffuse source pesticide contamination on stream macroinvertebrate

communities are clouded by the effects of physical habitat degradation in a set of

Danish streams. These findings emphasise the need for field sites with good physical

conditions that do not confound evaluation of the impact of other anthropogenic

stressors.

A suitable field site, located in an area with protected drinking water interests, was

identified in McKnight et al. (2010), involving a TCE groundwater plume discharging

into a stream. The previous study evaluated ecological modelling methods for assessing

the impact of TCE on surface water ecosystems, based on shifts in ecosystem biomass

patterns. In this study, we endeavoured to evaluate various methods for ecological

quality assessment in order to confirm the previous results regarding TCE, and,

recognizing that this catchment is mostly agriculturally-based, add an assessment of

impacts of other agriculturally related stressors (eutrophication and pesticides).

5

Accordingly, a comprehensive field campaign was carried out over two years (2010 and

2011), including an analysis of water chemistry, xenobiotic organic compounds

including pesticides, physical conditions and benthic macroinvertebrate communities

along a gradient of xenobiotic contamination in the stream. Biological monitoring that is

based on benthic macroinvertebrates integrates environmental conditions over time and

thereby reflects the cumulative impact of multiple stresses on the benthic community.

Thus, the evaluation of the ecological status in freshwater ecosystems using data on

benthic macroinvertebrate populations has become an indispensable tool for the

ecological risk assessment of freshwater pollution (Camargo, 1994).

The purpose of this study was to use observations of contaminants and benthic

macroinvertebrate communities to: (1) assess the applicability of different ecological

evaluation methods for determining the impact of selected pollutants (eutrophicants,

xenobiotic organic compounds and pesticides), (2) determine threshold values for

ecological impact of the detected pollutants by use of AQUATOX and the HQ index,

and (3) consider additional xenobiotic organic compounds from contaminated sites and

pesticides that are frequently found in groundwater to generalise the findings of the field

study.

This paper considers several specific contaminants from within some broader

categories of pollutants: eutrophicants (defined as substances such as nutrients that lead

to rapid growth of microorganisms in surface water and resultant de-oxygenation (see

e.g. Camargo and Alonso, 2006; Friberg et al., 2009); xenobiotic organic compounds

(defined here as organic compounds such as chlorinated solvents and gasoline

compounds originating in groundwater from contaminated sites); and pesticides

(selected herbicides, insecticides and fungicides).

6

2. Materials and Methods

2.1 Site description

The study stream flows past the town of Lille Skensved located on Sjaelland, Denmark,

where the catchment is characterised by a low elevation, clayey/loamy soils, a temperate

climate and an average regional precipitation of 500 mm yr-1. The secondary aquifer at

Lille Skensved is contaminated by TCE originating from an auto lacquer shop, where a

leaking storage tank was found in 1993 resulting in a plume extending up to 1,000 m

downstream (Fig. 1). Although little data exists regarding the source zone, measured

TCE concentrations (in the mgL-1 range) reveal the presence of a separate phase

contamination and indicate that the source will persist for many decades. For more

details on the field site geology, TCE plume and remediation history, see McKnight et

al. (2010).

Figure 1. Propagation of the TCE contaminant plume at the Skensved site, including sampling locations for xenobiotic organic compounds (circles), event-triggered samplers (pluses) and the five kick-samples (numbered).

7

Table 1: Environmental parameters measured in June and August 2010, including descriptive statistics, for the indicated sampling locations along Skensved stream (compare Fig. 1) and NOVANA control sites (C1-C6). Dashes indicate that data was not available.

June 2010 data August 2010 data Parameter Site 1 Site 3 Site 1 Site 2 Site 3 Site 4 Site 5 Mean Median St. Dev. C1 C2 C3 C4 C5 C6

Oxygen [mgL-1] - - 8.62 9.70 9.70 8.90 8.80 9.14 8.90 0.52 9.24 9.17 10.25 8.92 9.63 10.10 BOD [mgL-1] 1.18 1.03 1.59 1.14 2.36 2.34 1.26 1.56 1.26 0.57 1.3 0.95 1.01 0.66 0.50 1.20 Conductivity [µS cm-1] - - 584 580 585 591 610 590 585 11.85 - - - - - - pH - - 7.83 8.03 8.07 7.97 7.99 7.98 7.99 0.09 7.81 8.35 7.96 7.86 7.70 8.15 Ammonium [mgL-1]a 0.06 0.03 0.05 0.03 0.02 0.03 0.05 0.04 0.03 0.01 0.06 0.02 0.09 0.01 0.01 0.01 Nitrate [mgL-1] a 6.08 6.14 4.79 4.66 4.67 4.59 4.44 5.05 4.67 0.73 - - - - - - Phosphate [mgL-1] a 0.07 0.06 0.21 0.21 0.21 0.21 0.19 0.17 0.21 0.07 0.03 0.06 0.03 0.03 0.02 0.06 Total-N [mgL-1] b 5.50 6.64 5.33 4.95 5.33 5.10 5.00 5.41 5.33 0.58 - - 1.15 - - - Total-P [mgL-1] b 0.12 0.12 0.25 0.25 0.29 0.30 0.23 0.22 0.25 0.07 - - 0.05 - - - Chloride [mgL-1] - - 30.04 31.68 30.72 33.22 37.33 32.60 31.68 2.90 - - 57.0 - - - Sulphate [mgL-1] - - 46.89 46.92 46.84 47.37 49.12 47.43 46.92 0.97 - - 21.80 - - - Calcium [mgL-1] 135.65 139.03 93.21 94.78 90.93 90.53 92.55 105.24 93.21 22.00 81.45 Iron [mgL-1] 0.06 0.05 0.85 0.05 0.07 0.10 0.06 0.18 0.06 0.30 0.29 0.58 1.04 0.40 0.79 0.57 Potassium [mgL-1] 5.17 5.63 7.07 7.41 7.30 7.35 7.12 6.72 7.12 0.92 - - - - - - Magnesium [mgL-1] 10.93 11.57 9.87 9.71 9.55 9.61 9.80 10.15 9.80 0.78 - - - - - - Manganese [mgL-1] 0.01 c 0.01 c 0.03 0.01c 0.01 d 0.01 c 0.02 0.02 0.01 0.01 - - - - - - Sodium [mgL-1] 35.73 38.26 23.58 24.54 23.41 26.41 27.49 28.49 26.41 6.04 - - 31.18 - - -

a Sample values taken from filtered samples. b Sample values taken from unfiltered samples. c Calculated concentrations were below detection limit: the value is not reliable. d Calculated concentrations were higher than detection limit but below quantification limit: the value is uncertain.

8

The stream is characterised by coarse substrate, high sinuosity and high turbulence

(Table 1). The majority of the catchment is used for agriculture, and dominant crop

types are wheat, barley and oilseed rape. In total, a 3.3 km stretch of the Skensved

stream was surveyed to determine hydromorphological and physicochemical

parameters, including in-stream vegetation and flow (stream discharge), as well as

characterise benthic macroinvertebrates and xenobiotic organic compounds, as

described in the following sections.

2.2 Hydromorphology

To characterise the hydromorphology of Skensved stream, for each sampling site, five

cross-sectional transects at two meter intervals along the stream were surveyed and

results are given in the Supplementary Material (Appendix A, Table A). At each

transect, wetted width (W), depth (D) and water velocity (at 0.4 x depth) (U) were

measured at four points corresponding to 25, 50, 75 and 100% of the wetted width using

a flow-meter (Höntzsch μP-TAD). The discharge was calculated for each transect (D x

W x U). Four rectangular plots were established between each pair of transects (2 m x

25% of wetted width). In each plot, substratum type and the total macrophyte coverage

were estimated. Submergent and emergent macrophytes were identified to the lowest

possible taxonomical level and proportional coverage was estimated for each taxon.

The physical habitat quality at each sampling site was assessed using the Danish

Habitat Quality Index (DHQI) (Pedersen et al., 2006). The habitat survey was

conducted on a 50 m reach that included the location for kick sampling. The DHQI

assesses the quality of physical habitats evaluating 17 descriptors, and the final score

ranges from -12 to 63. The threshold level for good physical habitat quality is 26

9

(Dunbar et al., 2010). The difference between the average DHQI score for the Skensved

sites and the control sites was tested using a t-test (P<0.05).

To characterise streamflow, we used the 2010 hydrograph for the Lille Vejle stream

(see Appendix A, Figure A) as a surrogate measure for daily discharge in Skensved

stream (since no direct data for Skensved from 2010 exists). Lille Vejle stream is

located just north (ca. 13.5 km) of Skensved stream. For comparative purposes, daily

discharge is normalised according to the size of the catchment area (26 km2 for Lille

Vejle).

2.3 General water chemistry

Concentrations of oxygen and macro- and micro-nutrients in the stream water, as well

as biological oxygen demand (BOD5), conductivity, pH and temperature were measured

at all sampling sites (see Fig. 1 and Table 1). Conductivity and oxygen concentrations

were measured twice using a WTW multi-350i meter; pH was measured with a (YSI-

60) pH-meter. Water samples were collected twice in 2010 (June and August) in a well-

mixed part of the stream and analysed for a series of substances.

The following parameters were analysed according to European standards: BOD5

(DS/EN 1899 1999), ortho-phosphate (DS/EN 1189-1997) and ammonia-N (DS 11732

2005). Nitrate-N was analysed using Lachat-methods (Lachat Instruments, USA,

Quickchem. No. 10-107-06-33-A (Salycate method)). Chloride concentration was

measured using silver nitrate (AgNO3) (Clesceri et al., 1989). Concentrations of total-N

and total-P were measured (unfiltered samples) by the Kjeldahl-N method (Kjeldahl,

1883) and Danish standard (DS-291), respectively.

Water samples for cation analysis were immediately filtered in the field through a

0.45 µm cellulose filter into 50 mL PE containers, preserved by addition of 4 M HNO3

10

(pH<2), and stored at 4°C until analysis. A Varian Vista MPX Axial View Inductively

Coupled Plasma (ICP) OES was used for all measurements, with a Varian SPS3 auto

sampler used for sample introduction. The laboratory control was prepared from 1,000

mgL-1 single element stock solution (Perkin-Elmer). Calibration solutions were prepared

from 100 mgL-1 multi-element standards CCS-4 and CCS-6 (Inorganic Ventures), with

a calibration range of 20 to 25 mgL-1. All solutions, including blanks and samples, were

prepared from Milli-Q water and stabilized with 1% v/v concentrated nitric acid. The

detection limit for calcium, magnesium and sodium was 7.05 µgL-1, 7.16 µgL-1 for iron

and manganese, and 28.09 µgL-1 for potassium.

2.4 Sampling of xenobiotic organic compounds and pesticides

Samples for benzene, toluene, ethylbenzene, m-/p- and o-xylene (BTEX), naphthalene,

and the chlorinated solvents PCE, TCE, trans- and cis-1,2-DCE, 1,1-DCE, 1,1-DCA and

1,1,1-TCA, were collected in 40 mL glass vials, at 23 locations (Fig. 1). Samples were

immediately preserved using 4 M H2SO4 and stored at 4°C. The analytes were separated

and identified by GC/MS using an Agilent 7980 gas chromatograph system equipped

with an Agilent 5975C electron impact (70 eV) triple-axis mass-selective detector.

Detection and quantification limits were determined as described by Winslow et al.

(2006). The detection limit for all BTEX compounds were 0.11 µgL-1, except m,p-

xylene (0.22 µgL-1), and 0.14 µgL-1 for naphthalene. Other detection limits were 0.1

µgL-1 for TCE, 0.05 µgL-1 for cis-DCE, and ranged from 0.01-17.7 µgL-1 for all other

chlorinated compounds.

The event-triggered water sampler was deployed at site 3 (Fig. 1), and water

samples were analysed for a broad selection of the most commonly applied herbicides,

and for a series of banned compounds most commonly found in groundwater (Table 2).

11

During the main agricultural pesticide application season (May and June), event-

controlled runoff sampling systems were used to characterise exposure to diffuse source

pesticide contamination of the stream (Liess and Von der Ohe, 2005) caused by heavy

precipitation (defined as ≥10 mm). Each sampler consisted of a 1-L glass bottle

mounted in the main flow channel of the stream with the tube top position 5 cm above

the water level (Liess and Von der Ohe, 2005). Rising water level triggered sampling,

where the bottles were then filled passively through small (0.5 cm diameter) plastic

tubes emerging from the bottle top.

The bottles were retrieved within 24 hours after each heavy precipitation event and

stored at 4 ○C until analysis. In total, four sampling events triggered the systems in 2010

and 2011. The two events in 2010 occurred on May 15th and May 30th – to – June 1st,

with 17.5 mm and 13 mm precipitation, respectively. Additionally in 2010, grab

samples were collected on August 10th during base-flow conditions in the streams in

order to characterise pesticides mainly originating from base-flow groundwater

discharge. The two events in 2011 occurred on May 22nd and June 8th, on days having

11 mm and 12 mm precipitation, respectively.

Time-integrated sampling of the bed sediment was conducted using a suspended

particle sampler that was deployed in the main flow channel 10 cm above the stream

bed at site 3 during the period from the beginning of May to the end of June 2011 (Fig.

1). The full description and mechanistic details are given in Laubel et al. (2001).

Pesticide analyses for the event-controlled samplers and grab sampling were conducted

by Eurofins Miljoe A/S Laboratories, and the sediment sample was analysed at the

Swedish University of Agriculture (Uppsala, Sweden; Phillips et al., 2000). Analyses of

all the samples were based on solid phase extraction, and the final extract was analysed

12

by GC-MS or LC-MS. The minimum detection limit was 0.01 μgL-1 for all pesticide

compounds in water samples and 1 ng g-1 sediment (dry weight).

2.5 Benthic macroinvertebrate sampling

Benthic macroinvertebrates were sampled before (March) and after the main pesticide

application season (August), using a standardised kick sampling procedure (25 × 25 cm

hand net, 0.5 mm mesh size) (Laubel et al., 2001). Five locations were chosen along

Skensved stream (Fig. 1). At each location, four kick samples were taken across each of

three transects at positions located at distances 10 %, 50 %, 75 % and 100 % from the

stream bank. The 12 sub-samples were pooled into one sample and preserved in 70 %

ethanol. Macroinvertebrates were identified to the species level (when possible;

otherwise genus) with only a few exceptions: Oligochaeta (order), Chironomidae (sub-

family), Ostracoda (order), Heteroptera (family), Simuliidae (family) and Psychodidae

(family).

2.6 Control sites

Control sites with “Least Disturbed Conditions” (Stoddard et al., 2006) in the region of

Skensved stream from the Danish monitoring programme (NOVANA) were identified

and data for them was extracted from the ODA database (https://oda.dk). Selection

criteria were: (1) physically unmodified streams, (2) no contaminated sites or other

(known) discharges impacting the stream, (3) the majority of the catchment, i.e. > 90 %,

should be forest or (wet or dry) meadows and (4) the streams must have at least a good

ecological status according to the DSFI score (see also section 2.7.1).

Six streams fulfilled these criteria, and available data from the NOVANA database

consisted of macroinvertebrate samples, characterisation of substrate, vegetation,

hydrological parameters and water chemistry (see Appendix A, Table A). Data from the

13

most recent available year (2009) was selected and includes information on benthic

fauna from April, substrate and vegetation from June, and hydrological parameters and

water chemistry from 2-12 dates per year.

2.7 Ecological assessment methods

2.7.1 Danish stream fauna index

The Danish Stream Fauna Index (DSFI) is currently the method used by Denmark for

the biological assessment of running waters in compliance with the EU WFD, and

reports the status of oxygen sensitive species in a stream. Since oxygen levels are

affected by a number of contaminants, e.g. increased BOD5 or high nutrient levels, the

index also provides some indication of the chemical status of a stream. The index is

based on the presence/absence of a series of select species that are known to be

intolerant or very tolerant to oxygen depletion (e.g. facilitated by organic pollution)

(Dall and Lindegaard, 1995). The sampling procedure for DSFI is standardised,

endeavouring to sample all microhabitats at a site. A DSFI index-value of 7 represents

high ecological status (unpolluted conditions) under the EU WFD (European

Commission, 2000); other water quality classes include good, with values of 5-6;

moderate, with a value of 4; poor, with a value of 3; and bad, with values of 1-2

(Danish EPA, 2011).

2.7.2 Toxic units

We applied toxic units (TU) as a measure for xenobiotic and pesticide toxicity,

calculating TU for all compounds detected in each sample (Tomlin, 2001). TU are

calculated according to equation 1:

=( . ) log( / 50 )D magna i iTU C LC (1)

14

where TU(D.magna) is the toxic unit for pesticide i, Ci is the measured concentration of

pesticide i and LC50i is the corresponding 48 h LC50 value for D. magna exposed to

pesticide i. Both the maximum TU and summed TU were calculated, the latter

consisting of all the compounds detected in each water sample. For summed TU, the

suggested threshold value for observed acute effects in the field is ≥-3.0 (Liess et al.,

2008), which is based on results from the SPEARpesticides index (Liess and Von der Ohe,

2005). The summation of all TUs is based on the principle of toxic additivity; as the

number of components in a toxic mixture increases, the range of deviation from toxic

additivity has been suggested to decrease (Warne and Hawker, 1995). Differences

between summed concentrations of all compounds and summed TU in base-flow and

storm-flow water samples were tested using t-tests.

In order to compare the potential toxicity of the pesticides that were sorbed to the

sediment with the summed TU for water samples, the summed TU for the sediment

sample was calculated. In the calculations it is necessary to account for the fact that the

sorption of pesticides to organic micro-particles can reduce their acute toxicity to

macroinvertebrates by up to a factor of 400 compared to the toxicity of fully dissolved

pesticides (Hill, 1989). The calculation of TU using equation (1) therefore multiplies the

LC50 values for 48 h exposure of D. magna by a factor of 400 and so decreases the

ecological effect of the measured sediment concentrations.

2.7.3 SPEcies At Risk indices

The SPEcies At Risk indicator for pesticide contamination (SPEARpesticides) was

originally developed to detect the effects of periodic pesticide contamination on stream

macroinvertebrates as a result of normal agricultural practice. Macroinvertebrates are

classified as being at risk or not at risk due to pesticides according to their physiological

15

sensitivity to pesticides, life-cycle characteristics and recovery potential (Liess and Von

der Ohe, 2005). The sampling procedure is the same as used in the DSFI. The biological

traits were compiled using the freely available online SPEAR calculator

(http://www.systemecology.eu/SPEAR/index.php). After the “species at risk” were

defined, the SPEARpesticides index was computed as the relative abundance of the fraction

of sensitive taxa for each site, according to:

1

1

log( 1) **100

log( 1)=

=

+=

+

n

jjpesticides n

jj

x ySPEAR

x (2)

where n is the number of taxa, xj is the abundance of the taxon j, and y is equal to 1 if

taxon j is classified as “at risk”, otherwise 0 (Beketov et al., 2009). The SPEARpesticides

index has been connected to the EU WFD categories for ecological status; the

recommended threshold value characterising good ecological status is 33% SPEAR

(Beketov et al., 2009).

Additionally, a SPEARorganics index was computed as the arithmetic mean of the

species sensitivities for all species in the sample relative to that of D. magna:

.log( 50 / 50 )=j D magna jS LC LC (3)

where Sj is the sensitivity of the taxon j. These values reflect taxon-specific sensitivity

to organic contaminants in general (including both xenobiotics and pesticides), but not

to a particular chemical (Beketov and Liess, 2008). Since only the physiological

sensitivity of a taxon is considered for a given chemical, and not the ecological traits of

the taxon, the SPEARorganics index is designed for detecting chronic exposure. EU WFD

ecological status categories are currently not available for the the SPEARorganics index.

What is evident, however, is that the more negative this index value becomes, the more

serious a pollution event is with respect to the existing benthic macroinvertebrate

16

community structure.

The temporal difference in SPEARorganics and SPEARpesticides between the March and

August samples were evaluated using a t-test (P < 0.05, n = 5). Furthermore, we used

Pearson’s product moment to test the correlation between SPEARorganics and the

summed TU in the August base-flow samples (for pesticides, TCE and DCE), the total

concentration of pesticides, and TCE and DCE in the August samples (P < 0.05).

Conformity of data with a normal distribution and homogeneity of variance were

confirmed prior to performing the statistical tests (P < 0.05).

2.7.4 Hazard quotient index

The HQ index is applied to assess the likelihood of ecological impacts for observed

concentrations in Skensved stream. The HQ method utilises the ratio (or quotient) of an

exposure concentration divided by an effect concentration (equation 4), where an HQ

equal to one represents the threshold for ecological risk (U.S. EPA, 1998), and is

particularly used for chemicals where benchmark toxicity values are widely available:

/ 50=i i iHQ C LC (4)

where HQi is the hazard quotient for compound i, Ci is the concentration measured or

estimated at the point of exposure for compound i, and LC50i is the effect concentration

for compound i, which is a benchmark aqueous-phase toxicity value (e.g. LC50, EC50,

NOAEC), and represents the dose or lethal concentration where 50% of the test

population is killed. Here, the LC50 for test species representing different taxonomic

groups (fish, macroinvertebrates, macrophytes and micro-algae) was chosen, as this is

the benchmark utilised to produce ecotoxicology data for unknown species.

Where possible, acute (48 h) toxicity values were extracted from the ECOTOX

(U.S. EPA, 2011) or PAN databases (Kegley et al., 2008). If data were not available for

17

a particular species, the web-based (and freely available) interspecies correlation

estimation calculator was used (Raimondo et al., 2010), which uses least square

regression to predict acute toxicity (i.e. the LC50 value) to a species, genus or family

from the known toxicity of the chemical to a surrogate species. As with the TU method,

we calculated the HQ for the sediment sample using the acute LC50 values for 48 h

exposure of D. magna, but multiplied by a factor of 400 in order to account for the

reduced toxicity with respect to expected reduced bioavailability of pesticides bound to

micro-particles (see also section 2.7.2).

The HQ index is also used to determine threshold concentrations at which

contaminant impacts may be observed for a selected group of compounds typically

found in groundwater. Furthermore, these values were calculated for a broader range of

reference species. In addition to D. magna which is widely used as the standard

ecotoxicological indicator organism for chemicals in the environment (Baird et al.,

1989; Vandenbrouck et al., 2010), we also included sediment feeding chironomids,

predatory stoneflies (Plecoptera), and the game fish Brown trout (Salmo trutta), the

latter three being found in Skensved stream whereas D. magna are normally associated

with slow moving or standing waters (Allen, 1995).

2.7.5 AQUATOX

AQUATOX is a comprehensive, process-based ecological model for simulation of an

aquatic ecosystem together with the environmental fate and effects of various pollutants,

such as nutrients and organic chemicals (see Park and Clough, 2004; Park et al.,2008;

Sourisseau et al., 2008, and references therein). Here the model is used to determine if

measured contaminant levels result in notable changes from a control (no contaminant)

18

system, thereby evaluating ecological impacts for the 300 m section of the Skensved

stream impacted by the TCE groundwater plume (see also Fig. 1). AQUATOX is also

used to determine threshold concentrations, specifically, concentrations where a

detectable change to the modelled ecosystem occurs.

The model was calibrated to the 10 year average stream discharge data (1995-

2004), 2005 measured TCE concentration (similar to McKnight et al. (2010)), and

observed average BOD5 in 2010 (Sites 2-4; Appendix A, Table A). The BOD5

calibration was done by converting the measured BOD5 value into detritus loadings for

organic matter in the water column, following equation 148b in Park and Clough

(2010), then multiplying by 5 assuming 80 % refractory detritus (BOD5 represents labile

detritus). It should be noted that D. magna was replaced by a compartment containing a

representative species that is expected to be present at Skensved i.e. Gammarus pulex,

and which fills this ecological niche (suspended feeder).

Finally, a pseudo-sensitivity analysis – in the form of a scenario analysis – was

conducted (only for TCE) to assess the dominant controls potentially affecting stream

ecosystems, particularly the hydromorphological parameters The results are available in

the Supplementary Material (Appendix C).

3. Results and Discussion

3.1 Field surveys of linkages between pressures and ecological quality

3.1.1 Hydromorphological conditions

There was no significant difference between the average Danish Habitat Quality Index

score (DHQI) for the Skensved sites compared to control sites (P > 0.05). The site-

specific DHQI score for the Skensved sites ranged from 27 to 44, which are all above

19

the threshold value characterising good physical habitat quality. The influence of local

physical habitat quality on macroinvertebrate communities is well-known (Dunbar et

al., 2010; Friberg et al., 2009), and a DHQI score of 26 or above has been shown to be

indicative of good ecological status (DSFI score ≥ 5) (Wiberg-Larsen et al., 2010).

Consequently, the physical conditions in Skensved stream are not considered to be a

constraining factor for achieving good ecological status.

3.1.2 General water chemistry

In general, concentrations of measured macro- and micro- nutrients along the stream

stretch were comparable to control conditions (Table 1) (Boutrup et al., 2007).

Specifically, concentrations of nitrate-N were highest in early summer probably due

to tile drainage from loamy agricultural fields. Nitrate-N concentrations were lower

during summer and early autumn when tile drain flow ceased and base-flow conditions

were dominating (Fig. 2) (Kronvang and Bruhn, 1996). In contrast to the nitrate

observations, phosphate-P concentrations were higher in the late summer, i.e. during

low or base-flow conditions (Fig. 2), indicating that the main source of phosphate-P was

local point sources (e.g. septic tanks from urban settlements) (Kronvang and Bruhn,

1996).

Biological oxygen demand (BOD5) was highest (maximum 2.4 mgL-1) in August at

sites 3 and 4 (Fig. 1), where the connectivity between groundwater and surface water

was expected to be highest. This most likely reflects the transport of humic substances

from the groundwater deposits to the stream at these sites. In consequence, BOD5

briefly increased when dilution was reduced due to the decreased discharge in August.

Consequently, high BOD5 at sites 3 and 4 was probably only a problem for 4 to 6 weeks

during the summer (see Appendix A, Fig. A), and thus had no major impact on the

20

macroinvertebrate community (Friberg et al., 2010). These results clearly suggest that

concentrations of BOD5 and nutrients were of minor importance and therefore should

not affect the aim of obtaining good ecological status at any of the Skensved stream

sites.

The site-specific DSFI scores confirm this, as all sites were characterised by DSFI

scores of 4 in March and August (Table 2) indicating no site-specific or temporal effects

of BOD5. However, a DSFI score of 4 is indicative of moderate ecological status

probably reflecting effects of other stressors. Since the DSFI is intended for detecting

effects of organic pollution (BOD5), and because it only employs a categorization into

seven groups, the DSFI index is only poorly suited to capture the effects of other

stressors (see e.g. Friberg et al., 2009).

Site SPEARorganics SPEARpesticides SPEARpesticides water quality

class DSFI

#1 – March -0.50 11.05 2 4 #1 – August -0.52 4.32 1 4 #2 – March -0.35 12.64 2 4 #2 – August -0.51 6.85 1 4 #3 – March -0.39 14.67 2 4 #3 – August -0.36 12.56 2 4 #4 – March -0.45 19.95 2 4 #4 – August -0.32 12.84 2 4 #5 – March -0.39 16.59 2 4 #5 – August -0.39 8.23 1 4 C1a (2004) -0.36 34.67 4 6 C2a (2004) -0.14 49.69 5 6 C3 (2005) -0.30 46.48 5 7 C4 (2004) -0.18 43.60 4 6 C5 (2004) -0.46 32.21 3 5 C6 (2004) -0.24 38.35 4 6 a Indicates control sites that were included in the sampling campaign conducted in 2011.

Table 2: SPEAR index results for all sampled sites in spring and summer 2010 (compare Fig. 1), as well as the latest information for the control sites (C1-C6) (Boutrup et al., 2007).

21

3.1.3 Xenobiotic organic compounds and pesticides from groundwater

The results of the 2010 field campaign produced comparable results to the 2005 studies

for TCE concentration in stream water (McKnight et al., 2010) albeit with the maximum

concentration being approximately 20 times lower, peaking at 0.76 ugL-1 (Fig. 2), and

the peak location being shifted ca. 200 m downstream. The fact that these values are

much lower compared to the 2005 campaign is not surprising as the field site has been

under hydraulic containment (pump-and-treat) since 1999.

In total ten different pesticides (herbicides) were detected during August base-flow

conditions reflecting pesticide entry mainly via groundwater inflow (see also Appendix

B, Table B for the complete list of compounds screened, detected, and their

concentrations). Of the sixteen herbicides detected in total (both base-flow and storm-

flow water samples), five had maximum concentrations in August during base-flow,

indicating groundwater inflow as an important source for pesticides in Skensved stream.

Figure 2. TCE (solid line) and cis-1,2,-DCE (dashed line) concentrations measured over distance in the stream water at Skensved in 2010. The locations where kick sampling was conducted are also given (diamonds along x-axis); note that the 0 distance location corresponds to the kick-sampling site 1 (Fig. 1).

22

To evaluate the chemical toxicity, the summed TU for the groundwater-based

pollutants (TCE, cis-1,2-DCE and pesticides) ranged from -4.0 to -3.7, which is an order

of magnitude below the suggested threshold value (-3.0) where acute effects in the field

can be expected (compare August data, Table 3). Due to the continuous input of

groundwater-based pollutants, we additionally used the SPEARorganics index to evaluate

potential ecological effects (Beketov and Liess, 2008). SPEARorganics ranged from -0.52

to -0.32 at the five sampling sites. SPEARorganics showed no significant correlation with

the summed TU for the respective sites, and there was no significant difference in

SPEARorganics between March and August samples (Table 2). Moreover, the observed

SPEARorganics values for the Skensved sites are within the range of SPEARorganics values,

which have been found previously in uncontaminated streams (Beketov and Liess,

2008). Consequently, groundwater-based inflow of TCE, cis-1,2-DCE and pesticides

should not show any detectable effects on the macroinvertebrate communities in

Skensved stream.

These results are also comparable to the results determined for the HQ index (Table

3), utilising the maximum concentration detected for all compounds in stream water

(see Appendix B for pesticides; Fig. 3 for xenobiotics) – for D. magna. Specifically, the

HQ values for all detected concentrations were orders of magnitude below the threshold

value of one, which indicates the potential for ecological risk. The herbicide Dinoseb

had the highest value (HQ = 2.4E-04), and this was still four orders of magnitude below

the threshold. These results showed that observed concentrations at Skensved for

compounds originating in groundwater are far below those required for an ecological

impact according to the HQ index.

23

TU May/June

2010 [mgL-1]

TU August 2010 [mgL-1] TU May/June 2011 [mgL-1]

Sediment concentrations [μg kg-1 DW

sediment] Compounds

LC50 D. magna [mgL-1]

(U.S. EPA, 2011) Site #3 Site

#1 Site #2

Site #3

Site #4

Site #5

Site #1

Site #2

Site #3

Site #4

Site #5

Maximum HQ

D. magna [-] Site #4

2,4-D (H) 100 4.6E-5 - - - - - - - - - - 4.6E-07 - 4-CPP (H) 40 1.6E-5 - - - - - - - - - - 4.0E-07 - BAM (H) 180 2.8E-5 3.0E-5 3.7E-5 3.7E-5 3.4E-5 3.3E-5 2.0E-5 3.0E-5 1.6E-5 2.3E-5 2.6E-5 2.1E-07 - Bentazone (H) 64 9.2E-5 - 1.4E-5 1.3E-5 1.5E-5 1.6E-5 2.0E-5 3.0E-5 1.7E-5 4.3E-5 3.0E-5 1.4E-06 - Desethylatrazine (H) 6.9 - - - - - - - - - - - 1.6E-06 - Dichlorprop (H) 100 - - - - - - 5.1E-5 1.8E-5 - 2.2E-5 - 5.1E-07 - Diflufenican (H) 0.24 - - - - - - - - - - - 4.06E-04 39 Dinoseb (H) 0.2 1.3E-5 2.2E-5 2.8E-5 2.7E-5 3.2E-5 4.8E-5 5.1E-5 1.8E-5 - 2.2E-5 - 2.4E-04 - Diuron (H) 5.7 2.0E-5 - - - - - - - - - - 3.5E-06 - DNOC (H) 5.7 1.7E-4 - 1.0E-5 - 1.2E-5 - 1.8E-5 1.4E-5 4.2E-5 1.9E-5 1.6E-5 3.0E-05 - Hydroxyatrazine (H) 85 - - - 1.1E-5 1.1E-5 1.1E-5 - - - - - 1.3E-07 - Isoproturon (H) 0.6 1.7E-5 - - - - - - - - 5.9E-5 - 9.8E-05 - MCPA (H) 190 9.2E-4 5.1E-5 6.3E-5 6.2E-5 6.2E-5 6.1E-5 4.9E-4 2.2E-4 1.0E-3 2.4E-4 7.8E-4 5.3E-06 - Metamitron (H) 5.7 3.7E-4 - - - - - - - - - - 6.5E-05 - Simazine (H) 1.1 5.3E-5 - - 1.0E-5 1.1E-5 1.0E-5 - - - - 4.8E-05 - Terbutylazine (H) 21.2 2.0E-5 2.1E-5 2.4E-5 2.5E-5 2.6E-5 2.6E-5 2.0E-4 8.1E-5 2.3E-4 1.5E-4 4.5E-5 1.1E-05 - Trichloroacetic acid (H) 3,100 9.9E-5 2.0E-5 1.9E-5 1.9E-5 1.9E-5 2.3E-5 - - - - - 3.2E-08 - Boscalid (F) 5.33 - - - - - - 2.5E-5 2.6E-5 1.1E-4 8.9E-5 1.3E-5 2.1E-05 - Epoxiconazole (F) 8.69 - - - - - - - - 2.9E-5 2.2E-5 - 3.3E-06 - Hexachlorobenzene (F) 0.005 - - - - - - - - - - - 2.50E-03 5 Chlorpyrifos (I) 0.0001 - - - - - - - - - - - 7.50E-02 3 Deltamethrin (I) 0.00056 - - - - - - - - - - - 6.25E-02 14 Lambda-cyhalothrin (I) 0.00036 - - - - - - - - - - - 4.10E-02 5.9 HCH-� (lindane) (I) 1.6 - - - - - - - - - - - 3.13E-06 2 TCE 18,000 - - 1.3E-4 4.3E-4 5.9E-4 4.1E-4 - - - - - 4.2E-08 - cis-1,2-DCE 220,000 - - - - 1.2E-4 7.9E-4 - - - - - 6.4E-10 - Summed TU -3.6 -4.0 -3.9 -3.9 -3.8 -3.7 -3.6 -4.0 -4.3 -3.6 -4.9 -0.74 a Summed conc. [mgL-1] 1.9E-3 1.4E-4 3.2E-4 6.4E-4 9.3E-4 7.3E-4 7.4E-4 4.2E-4 1.4E-3 6.7E-4 9.1E-4 - Number of compounds 13 7 9 10 11 11 7 7 7 9 6 6

a Calculated with a safety margin of 1,000 due to expected lower toxicity for compounds that are adsorbed to sediment.

Table 3. TU and HQ calculation results, including summed concentration and number of compounds, for Skensved sites and sampling dates. Note that the maximum concentration value was chosen for the May/June sampling dates. H stands for herbicide, F for fungicide. Xenobiotic organic compounds were only sampled during August 2010.

24

3.1.4 Pesticides related to spring spraying

In total, 18 of the 35 pesticides screened were detected in the water samples (Appendix

B). Sixteen of the detected compounds were herbicides, the other two were fungicides;

no insecticides were detected in either year. The average summed concentration of

pesticides for all storm-flow samples in May/June 2010 and 2011 (0.85 ± 0.38 μgL-1)

was not significantly different from the average summed concentration for the 2010

August base-flow samples (0.55 ± 0.32 μgL-1) (P = 0.212). Moreover, the average

summed TU for all storm-flow samples (ranging from -4.9 to -3.6) was not significantly

different from the average summed TU for the August base-flow samples (P = 0.714),

and were below the threshold for expected impacts (≥-3.0). These findings were again

comparable to the results determined for the HQ index, now focused on the pesticides

related to spraying in the spring season in stream water, indicating that the observed

concentrations in Skensved stream are orders of magnitude below the HQ threshold

value of one (Table 3).

However, six pesticides were detected in the sediment sampled during May-June in

2011 (Table 3), one herbicide, one fungicide and four insecticides. Both the fungicide

hexachlorobenzene and the insecticide HCH-gamma (lindane) are EU priority

pollutants. The pesticides that were detected with the suspended sediment sampler have

moderate to highly lipophilic physicochemical properties indicating that they were

transported from adjacent fields sorbed to organic particles (Liess et al., 1996),

preferential fracture flow paths or drainage systems during heavy rain falls. This, in

conjunction with the low half-lives associated with these compounds (U.S. EPA, 2011),

supports our conjecture that they have most likely originated from spraying in the spring

season and not from groundwater base-flow.

25

The summed TU for the sediment sample was -0.74, more than two orders of

magnitude above the threshold for expecting ecological effects in the field. In fact, this

is the highest summed TU ever observed for Danish streams (compare with Rasmussen

et al. 2011b, Friberg et al., 2003). Notably, the result of the TU calculation for sediment

contrasted markedly with those determined by the HQ index, where values were

between two and six orders of magnitude below the threshold of one (Table 3). The

reason for this lies in how the threshold values are defined: the HQ index threshold only

relates to (LC50 acute toxicity) values that result from ecotoxicological lab testing,

whereas the TU threshold was defined via the SPEARpesticides index which is based on

fully-integrated population responses in the field and therefore will be more sensitive.

Application of the SPEARpesticides index resulted in values ranging from 11.1 % to

19.9 % SPEAR abundance in March and from 4.3 % to 12.8 % SPEAR abundance in

August. The decrease in average SPEARpesticides from before the spring pesticide

spraying season (March) to after (August) was significant (P = 0.002). Moreover, the

SPEARpesticides in the six control streams (ranging from 32.2 % to 49.6 %, Table 2) was

significantly higher than all SPEARpesticides values in both the March and August

samples from Skensved stream (P < 0.001). Using SPEARpesticides as an ecological

indicator tool, the temporal dynamics in the macroinvertebrate community structure

clearly showed a response to the pesticide contamination that was documented by the

suspended sediment sample. Furthermore, the SPEARpesticides scores correspond to poor

ecological status for all March samples and to poor-to-bad status for the August

samples. Considering that the currently-used ecological indicator (DSFI) showed no

temporal changes, nor a response to pesticide pollution, our results highlight the

importance of using the appropriate ecological indicator tools when conducting

26

investigative surveys of surface water ecosystems. Moreover, our results highlight the

importance of considering the sediment in the evaluation of pesticides in streams.

The low SPEARpesticides values in March (before the main pesticide application

season) could reflect the fact that the macroinvertebrate community structure has

adapted to several decades of agriculture in the catchment, which may have slowly

reduced the abundance of sensitive species. Such long-lasting effects have been

documented in two previous studies (Liess and Von der Ohe, 2005 and Von der Ohe et

al., 2009). However, more field studies with high levels of temporal detail are needed to

further document the long-term effects of the agricultural past.

3.2 Evaluating the ecological impact with AQUATOX

3.2.1 TCE model results

The application of the AQUATOX model in this section is undertaken to improve the

understanding of TCE on the stream ecology. This forms the basis for determination of

the “loading threshold range” (section 3.2.2) and extension of our results to a broader

group of contaminants (section 3.3.3). Figures 3 and 4 present the ecological impact

results for TCE for a variety of parameters. The calculated bioaccumulation factor

(BAF) as a function of time for five species (Chironomid, Caddisfly, Mayfly, Stonefly

and Brown trout) is presented and can be compared with the TCE concentration in

stream water in Fig. 3a, and the TCE half-life in sediment in Fig. 3b, for a modelled

timeframe of three years. Most BAF values stayed constant over the entire 3 year

simulation period with two notable exceptions. The modelled fish species Brown trout

consistently had an elevated value during the summer months. This was only slight for

the adult species (ca. 7 L kg-1, Fig. 3a), but was quite large for the juvenile species

(maximum of ca. 240 L kg-1, only depicted in Fig. 3a), corresponding to the elevated

27

TCE concentration in stream water.

In contrast, the modelled sediment feeder Chironomid had an elevated BAF during

the winter months (of ca. 35 L kg-1, Fig. 3a). The reason for this can be seen in Fig. 3b,

which plots BAF versus TCE half-life in the sediment and clearly shows an elevated

TCE concentration in the streambed sediment during the winter months. It should be

noted that these results (i.e. including BOD5 calibration) are slightly different to those of

McKnight et al. (2010) who presented somewhat elevated BAF values for all modelled

species. However, the overall conclusions are similar to those of the earlier work.

Figure 3. AQUATOX results for bioaccumulation factor and (a) TCE concentration (including juvenile brown trout) and (b) TCE half-life in sediment for the 300 m groundwater-impacted stream stretch, and (c) half-lives for both TCE in water and sediment, as well as time required for 95% TCE loss in both media.

28

3.2.2 Threshold findings for TCE

The predicted base case point source loading into surface water was increased (or

decreased) from 5.5 kg yr-1 – as measured at the site and resulting in maximum

modelled TCE stream water concentrations of 10 ugL-1 – by factors of ten in order to

establish the “loading threshold range” at which toxicant stress could perturb the

modelled AQUATOX ecosystem. Figure 4 presents the stream discharge and predicted

biomass pattern for two species – chironomid (Fig. 4a) and stonefly (Fig. 4b) – for a

modelled timeframe of three years.

The results show that there was little deviation between the control and loading

scenarios for the biomass patterns of both species up to 55 kg yr-1. Thus, the threshold

for impact for TCE lies between 55 and 550 kg yr-1 for both species, where the

predicted biomass decreased by ca. 50 %. Results also show that stream discharge was

the limiting factor most influencing the modelled biomass concentration for all species.

The model thus supports the evaluation in section 3.1 by the other four methods,

indicating that TCE does not affect the attainment of good ecological status in Skensved

stream.

3.2.3 Threshold findings for other compounds

Both the AQUATOX model and the HQ index were used to generalise the findings in

the case study to other compounds of interest, as well as to evaluate chemical impacts

from a species-specific perspective. Specifically, the models were extended to

contaminants that are typically arising from contaminated sites (benzene, PCE, and

naphthalene), or pesticides found in Danish groundwater, ultimately representative of a

range of contaminant concentration scenarios (i.e. indicative of different flow

conditions). The AQUATOX results for all compounds are given in Table 4 for three

29

selected organisms: Chironomid, Stonefly and Brown trout. With respect to TCE, the

“loading threshold” for all organisms ranged from 55 – 550 kg yr-1, which is well above

the actual site-specific loading determined for the site. However, PCE and naphthalene

produced lower “loading threshold ranges” than TCE for at least one modelled

organism. It is interesting to note that, in general, the thresholds determined for

benzene, TCE, naphthalene and PCE corresponded to typical contaminant mass

discharge ranges that could be expected at contaminated sites leaching into groundwater

(ITRC, 2010).

Figure 4: AQUATOX results for predicted biomass pattern for (a) chironomid and (b) stonefly versus water volume for the control (no pollutant) and four TCE loading scenarios.

30

In order to better compare these results with actual concentrations in stream water

and the HQ index, a “biomass perturbation concentration” was also determined (Table

4). This value is the lowest concentration causing the predicted biomass pattern to differ

significantly from the control simulation (i.e. no toxicant present), as shown by the

positioning of the arrow in Fig. 5. These values can be compared with examples of

concentration values currently measured in the field and reported in the literature (Table

4) and range from 0.001 to 0.023 mgL-1 for xenobiotics (see e.g. Conant et al., 2004;

Gomez-Belinchon et al., 1991; McKnight et al., 2010; Yamamoto et al., 1997) and from

0.001 to 0.3 mgL-1 for pesticides (McKnight et al., 2011; Styczen et al., 2003). In

addition to PCE, the compounds naphthalene and glyphosate had perturbation

concentrations close to or below values actually measured in surface water.

a Regression necessary to produce ecotoxicity data (after Raimondo et al., 2010).

Compound

Chironomid loading

threshold range

[kg yr-1]

Stonefly loading

threshold range

[kg yr-1]

Brown trout loading

threshold range

[kg yr-1]

Chironomid biomass

perturbation conc.

[mgL-1]

Stonefly biomass

perturbation conc.

[mgL-1]

Brown trout biomass

perturbation conc.

[mgL-1]

Concentration in surface water

(location) [mgL-1]

TCE 55 - 550 55 - 550 55 - 550 a 0.4 0.55 0.04 a 0.017 (Denmark) Benzene 55 - 550 55 - 550 55 - 550 a 0.35 0.38 0.5 a 0.011 (Japan) PCE 5.5 – 55 a 55 - 550 55 - 550 a 0.007 a 0.03 0.05 a 0.023 (Canada) Naphthalene 55 - 550 0.5 - 5.5 5.5 - 55 a 0.55 0.002 0.02 0.001 (Spain) MCPA >55 000 a 55 - 550 a >55 000 >120.0 a 0.6 a >120.0 0.003 (Denmark) Metamitron 550 - 5 500 a 55 - 550 a 550 - 5 500 a 6.0 a 0.18 a 4.0 a 0.001 (Denmark) Glyphosate 55 - 550 55 - 550 a 0.5 - 5.5 4.4 0.16 a 0.005 0.3 (Denmark)

Table 4. AQUATOX loading threshold ranges and biomass perturbation concentrations for selected xenobiotic organic compounds, pesticides and selected organisms, including Chironomid, Stonefly and Brown trout. TCE, Benzene, PCE and naphthalene are expected to arise from contaminated sites, so these loading threshold ranges can be compared to typical mass discharge values. For the pesticides expected to discharge from groundwater base-flow along the entire stretch, the loading threshold ranges (in italic) are only a means to calculate the biomass perturbation concentration. Also given are examples of surface water concentrations reported in the literature.

31

For comparison purposes, the HQ index results are presented in Table 5 for the

same subset of compounds and species considered above. We first calculated HQ using

a concentration that is at the high end of the range reported in the literature. MCPA and

glyphosate had the highest values (HQ = 0.1), but these were still an order of magnitude

below the recommended threshold value of one. We then calculated the concentration

values needed to reach the threshold and compared them to the values reported in the

literature. Naphthalene had a threshold value of 0.011 mgL-1, and was thereby also the

closest to the actual measured and reported values in the literature, although again, it

differed by an order of magnitude. PCE, MCPA, metamitron and glyphosate also had

fairly low thresholds for at least one species (ranging from 0.02 to 3.0 mgL-1), but these

are still at least one order of magnitude below the threshold (MCPA), and in most cases,

far below actual measured concentrations in surface water. These results suggest that

contaminant concentrations have to be well above the values being reported in the

literature before the HQ index will predict an ecological impact.

It is interesting to note that the HQ index results discussed above are species-

dependent. For example, the lowest threshold concentrations for naphthalene, PCE,

MCPA and metamitron were obtained either for the predatory invertebrate Stonefly or

for the sediment feeder Chironomid, and not for (the suspended feeder) D. magna. In

fact, the pesticide glyphosate was the only compound for which the D. magna HQ

threshold concentration provided the lowest (i.e. most conservative) value. This finding

is worrying, considering that D. magna is often used as a standard ecotoxicological

indicator.

When comparing the HQ index results to AQUATOX, it should be mentioned that,

although the modelling in AQUATOX employs the same LC50 values that are used in

32

the HQ calculations, the model also includes more sensitive values (i.e. EC50 growth

and reproduction) to assess food web interactions. Not surprisingly, the perturbation

concentrations were significantly lower than the HQ index threshold values. In about

half the cases, the AQUATOX concentration thresholds were about 100 times lower

than those obtained using the HQ index (compare Tables 4 and 5).

4. Implications for the EU WFD

The results in this paper clearly demonstrate the need for re-evaluating existing

ecological indices and ensuring that we are using the best available methods to

determine the ecological status of streams. It is essential that the indices used are

capable of capturing all the effects of anthropogenic stressors that could be (or have

been) impacting ecosystems.

The SPEAR indices were capable of distinguishing stressor effects, i.e. for

xenobiotic organic compounds and pesticides. Furthermore, SPEARpesticides was

additionally able to capture seasonal trends for pesticide application. Further work is

still needed in order to connect the SPEARorganics index to the EU WFD ecological

classes. In contrast, the DSFI index could neither distinguish stressor effects, nor

capture seasonal effects, perhaps due to the fact that its’ intended use is for detecting the

effects of organic-caused oxygen depletion. Others have obtained similar results, for

example, the German Saprobic index – which was also constructed to detect the effects

of organic pollution – was less successful in capturing the effects of contaminants than

SPEARpesticides (Schletterer et al., 2010).

It is interesting that the TU results, which are only a measure of chemical toxicity,

were also capable of distinguishing stressor effects, identifying the compounds found in

the sediment sample as being the only significant factor for ecological status. In

33

Table 5: HQ results for a range of xenobiotic organic compounds and pesticides and selected organisms, using actual measured concentrations in surface water reported in the literature (from Table 5). Also given are the HQ threshold results, i.e. concentrations required for a hazard quotient equal to 1.

Compound HQ

D. magna [-]

HQ Chironomid

[-]

HQ Stonefly

[-]

HQ Brown trout

[-]

D. magna threshold [mg L-1]

Chironomid threshold [mg L-1]

Stonefly threshold [mg L-1]

Brown trout threshold [mg L-1]

TCE 9.4E-04 4.1E-04 2.4E-04 0.001 a 18.0 42.0 70.0 16.5 a Benzene 1.9E-04 3.2E-04 8.5E-05 2.5E-04 a 59.6 34.0 130.0 44.8 a PCE 0.003 0.02 a 0.006 0.005 a 9.1 1.3 a 3.6 4.3 a Naphthalene 4.6E-04 3.6E-04 0.09 a 0.001 a 2.2 2.8 0.011 a 1.4 a MCPA 2.7E-04 0.01 0.13 a 1.0E-05 11.0 0.4 0.023 a 300.0 Metamitron 9.8E-06 2.5E-05 a 9.1E-04 a 4.0E-06 a 101.7 40.2 a 1.1 a 247.5 a Glyphosate 0.1 0.01 a 0.05 a 0.04 3.0 55.0 a 6.2 a 7.6 a Regression necessary to produce ecotoxicity data (after Raimondo et al., 2010).

34

contrast, the HQ index and the DSFI index were not capable of isolating this stressor.

More work is needed, however, to determine whether the calculation method applied to

determine the HQ is appropriate for contaminants bound to sediment.

Our study emphasized that contaminated sites may also impact streams; although

no ecological effects were found, the AQUATOX model simulations showed that flow

conditions and other contaminants may be more sensitive to stream ecology.

Accordingly, it may be challenging to find an index that can truly separate and identify

the most important stressors on a stream environment, but given the link between

science and policy – where such indices are used by policy makers in defining water

quality limits – it becomes a crucial issue (Kitsiou and Karydis, 2011).

Finally, the overall results are similar to other studies which demonstrate ecological

degradation due to agro-industrial runoffs and hydromorphological alterations in

streams. However, in contrast to other studies we have shown that for river restoration

to be successful, risk-mitigation procedures are needed not only along the mid- and

lower reaches of rivers, but also on the upper reaches. To date, the upper parts of

catchments have not been considered under the WFD in most EU countries.

5. Conclusions

This study has shown that traditional approaches for determining ecological impact fail

to account for all potential stressors affecting benthic macroinvertebrate populations in

streams. In particular:

• Hydromorphology, macro- and micro-nutrients and BOD5 were comparable to

control conditions, so these pressures are therefore not likely to have obstructed

the obtainment of good ecological status at Skensved stream. This was also

35

shown by the DSFI score obtained for all sampling locations. However, the

DSFI indicated only moderate ecological conditions, probably reflecting the

effects of other stressors.

• All methods applied in this study confirmed that the xenobiotic stressor TCE,

discharging into the stream from a contaminated site, did not impact benthic

macroinvertebrates at measured stream concentration levels.

• Many pesticides were measured under stream base-flow conditions, indicating

that groundwater inflow is an important source of pesticides to the Skensved

stream. However, the methods applied (TU, SPEARorganics, and HQ index) could

detect no significant effects to the macroinvertebrate communities. Similar

results were obtained (TU and HQ index) for pesticides thought to originate

from the spring spraying season, at observed concentrations in the stream water.

• The novel SPEARpesticides index and TU indicated, however, that Skensved

stream was far from obtaining good ecological status due to pesticide

contamination. The response is most likely reflecting the findings of pesticides

bound to sediment, presumably related to the spring spraying season.

• The SPEARpesticides index was also capable of capturing a strong seasonal effect,

which is thought not to occur in un-impacted streams. Specifically, the

SPEARpesticides scores corresponded to poor ecological status before the pesticide

spraying season (March samples), and to poor-to-bad status after (August

samples). However, the generally very low index scores before the spraying

season indicate that the stream ecosystem may reflect the catchments’ long

history of intensive agriculture.

36

• The evaluation by the AQUATOX model revealed that stream discharge was

found to be the factor most limiting the modelled biomass concentration for all

species – pointing to the importance of hydromorphology in the obtainment of

good ecological status.

• Threshold findings were determined for additional xenobiotic organic

compounds arising from contaminated sites and pesticides that are frequently

found in groundwater. AQUATOX results indicated that PCE, glyphosate, and

naphthalene could adversely impact ecosystems at concentrations observed and

reported in streams. Not surprisingly, the threshold concentrations produced by

AQUATOX were significantly lower than values obtained using the HQ index.

The HQ index identified naphthalene as having a threshold value closest to

concentrations reported in streams.

The results presented reflect the importance for identifying and implementing

suitable ecological assessment methods that are capable of capturing (and ideally

separating) the effects of all anthropogenic stressors potentially affecting ecosystems, in

order to assess compliance with the goals of the EU WFD. Results demonstrate that

some commonly used methods for the assessment of ecological impacts are not

sufficient for this purpose. Alternatives must be considered and may lead to a

determination of poorer ecological status in many surface water bodies.

Acknowledgements

The authors gratefully acknowledge the support of the Danish Research Council project

RiskPoint (grant no. 2104-07-0035). We also thank Jonathan Clough, Dr. Richard Park

and Marjorie C. Wellman for their timely advice and support with AQUATOX. We

would also like to acknowledge Uffe Mensberg and Henrik Stenholt (NERI field

37

technicians), Mikael E. Olsson and Morton Andreasen (DTU lab technicians) and Bent

Skov (DTU field technician). In addition, field data was collected and supported by

Nanna I. Thomsen, Maria C. Loinaz and Daniele Promio.

References

Allen JD. Stream Ecology. Structure and function of running waters. Dordrecht, The Netherlands: Kluwer Academic Publishers; 1995.

Baird DJ, Barber I, Bradley M, Calow P, Soares AMVM. The Daphnia bioassay: a critique. Hydrobiologia 1989; 188/189: 403-6.

Beketov MA, Liess M. An indicator for effects of organic toxicants on lotic invertebrate communities: Independence of confounding environmental factors over an extensive river continuum. Environ Pollut 2008; 156: 980-87.

Beketov MA, Foit K, Schäfer RB, Schriever CA, Sacchi A, Capri E, et al. SPEAR indicates pesticide effects in streams - Comparative use of species- and family-level biomonitoring data. Environ Pollut 2009; 157: 1841-48.

Bianchin M, Smith L, Barker JF, Beckie R. Anaerobic degradation of naphthalene in a fluvial aquifer: A radiotracer study. J Contam Hydrol 2006; 84: 178-96.

Boutrup S, van der Bijl L, Jensen PN, Svendsen LM, Grant R, Bøgestrand J, Jørgensen TB, Ellemann T, Ærtebjerg G, Bruus M, Søgaard B, Thorling L, Dahlgren, K. Water quality and Nature: Condition, development and scientific compilation. National Environmental Research Institute, Aarhus University: Faglig report fra DMU Nr 646; 2007 [In Danish].

Brüsch W. Almene vandvaerkers boringskontrol af pesticider og nedbrydningsprodukter. Danish EPA: Technical Report Nr. 26; 2007 [In Danish].

Camargo JA. The importance of biological monitoring for the ecological risk assessment of freshwater pollution: A case study. Environ Int 1994; 20: 229-38.

Camargo JA, Alonso Ãl. Ecological and toxicological effects of inorganic nitrogen pollution in aquatic ecosystems: A global assessment. Environ Int 2006; 32: 831-49.

Clesceri LS, Greenberg AE, Eaton AD. Standard methods for the examination of water and wastewater. 20th edition. Washington, DC, USA: American Public Health Association; 1989.

Conant B, Cherry JA, Gillham RW. A PCE groundwater plume discharging to a river: influence of the streambed and near-river zone on contaminant distributions. J Contam Hydrol 2004; 73: 249-79.

Dall PC, Lindegaard C. An overview of Danish freshwater macroinvertebrates for the evaluation of pollution effects in lakes and streams. Copenhagen University, Denmark: Ferskvandsbiologisk Laboratorium; 1995.

Danish EPA. Biological assessment of watercourse quality. Danish EPA, Ministry of Environment and Energy: Guidelines, Nr. 5; 1998 [In Danish].

Danish EPA. Bekaempelsesmiddelstatistik 2009. Danish EPA: Orientering fra Miljøstyrelsen; 2010a [In Danish].

Danish EPA. Bekendtgørelse om miljøkvalitetskrav for vandområder og krav til udledning af forurenende stoffer til vandløb, søer eller havet. Danish EPA; 2010b [In Danish].

Danish EPA. Bekendtgørelse om fastsættelse af miljømål for vandløb, søer, kystvande, overgangsvande og grundvand. Accessed:

38

(https://www.retsinformation.dk/Forms/R0710.aspx?id=127762#K1); November 2, 2011 [In Danish].

Dunbar MJ, Pedersen ML, Cadman D, Extence C, Waddingham J, Chadd R, Larsen SE. River discharge and local-scale physical habitat influence macroinvertebrate LIFE scores. Freshwater Biol 2010; 55:226-42.

European Commission. Establishing a framework for community action in the field of water policy. European Commission: Directive 2006/06/EC, October; 2000.

Friberg N, Sandin L, Pedersen ML. Assessing the Effects of Hydromorphological Degradation on Macroinvertebrate Indicators in Rivers: Examples, Constraints, and Outlook. Integrated Environmental Assessment and Management 2009; 5: 86-96.

Friberg N, Skriver J, Larsen SE, Pedersen ML, Buffagni A. Stream macroinvertebrate occurrence along gradients in organic pollution and eutrophication. Freshwater Biol 2010; 55: 1405-19.

Gomez-Belinchon JI, Grimalt JO, Albaigès J. Volatile organic compounds in two polluted rivers in Barcelona (Catalonia, Spain). Water Res 1991; 25: 577-89.

Hill, IR. Aquatic organisms and pyrethroids. Pesticide Science 1989; 27: 429-57. Henriksen HJ, Troldborg L, Hoejberg AL, Refsgaard JC. Assessment of exploitable

groundwater resources of Denmark by use of ensemble resource indicators and a numerical groundwater-surface water model. J Hydrol 2008; 348: 224-40.

Hinsby K, Condesso de Melo MT, Dahl M. European case studies supporting the derivation of natural background levels and groundwater threshold values for the protection of dependent ecosystems and human health. Sci Total Environ 2008; 401: 1-20.

Hose GC. Assessing the Need for Groundwater Quality Guidelines for Pesticides Using the Species Sensitivity Distribution Approach. Hum Ecol Risk Assess 2005; 11: 951-66.

ITRC. Use and Measurement of Mass Flux and Mass Discharge. Washington, D.C.: Interstate Technology & Regulatory Council, Integrated DNAPL Site Strategy Team, www.itrcweb.org; 2010.

Janniche GS, Mouvet C, Albrechtsen HJ. Vertical small scale variations of sorption and mineralization of three herbicides in subsurface limestone and sandy aquifer. J Contam Hydrol 2011; 123: 167-77.

Kegley SE, Hill BR, Orme S, Choi AH. PAN Pesticide Database. North America, San Francisco, CA: Pesticide Action Network; 2008.

Kitsiou D, Karydis M. Coastal marine eutrophication assessment: A review on data analysis. Environ Int 2011; 37: 778-801.

Kronvang B, Bruhn AJ. Choice of sampling strategy and estimation method for calculating nitrogen and phosphorus transport in small lowland streams Hydrol Process 1996; 10: 1483-1501.

Kronvang, B.; Iversen, H.L.; Hans, L.; Vejrup, K.; Mogensen, B.B.; Hansen, A.M.; Hansen, L.B. Pesticides in Streams and Subsurface Drainage Water within Two Arable Catchments in Denmark: Pesticide Application, Concentration, Transport and Fate. Danish EPA: Danish Ministry of the Environment; 2003.

Laubel A, Kronvang B, Fjordback C, Larsen S. Time-integrated suspended sediment sampling from a small lowland stream. International Association of Theoretical and Applied Limnology 2001; 28: 1420-24.

Lauridsen, R.B.; Kronvang, B.; Friberg, N. Occurrence of sediment-bound pyrethroids in Danish streams and their impact on ecosystem function. Water, Air and Soil Pollution 2006; 6:423-32.

Liess, M.; Schulz, R.; Neumann, M. A method for monitoring pesticides bound to suspended particles in small streams. Chemosphere 1996; 32:1963-69.

Liess M, Von der Ohe PC. Analyzing Effects of Pesticides on Invertebrate Communities in Streams. Environ Toxicol Chem 2005; 24: 954-65.

39

Liess, M.; Schäfer, R.B.; Schriever, C.A. The footprint of pesticide stress in communities - Species traits reveal community effects of toxicants. Sci Total Environ 2008; 406:484-90.

McKnight US, Funder SG, Rasmussen JJ, Finkel M, Binning PJ, Bjerg PL. An integrated model for assessing the risk of TCE groundwater contamination to human receptors and surface water ecosystems. Ecol Eng 2010; 36: 1126-37.

McKnight US, Rasmussen JJ, Kronvang B, Bjerg PL, Binning PJ. Occurrence of pesticides in surface and groundwater in two catchments on Sjaelland, Denmark. Vandkvalitet I grundvand/overfladevane – hvordan griber vi det an? ATV Jord og Grundvand, Mødenr. 76; 2011: 25-32. ISBN 978-87-913-13585 [In English].

Park RA, Clough JS. AQUATOX (Release 2): Modeling Environmental Fate and Ecological Effects in Aquatic Ecosystems. Technical Documentation. U.S. EPA: Office of Water, Nr. EPA/823/R-04/002; 2004.

Park RA, Clough JS. AQUATOX (Release 3.1 Beta) Modeling Environmental Fate and Ecological Effects in Aquatic Ecosystems - DRAFT - Volume 2: Technical Documentation. U.S. EPA: Office of Water, Office of Science and Technology; 2010.

Park RA, Clough JS, Wellman MC. AQUATOX: Modeling environmental fate and ecological effects in aquatic ecosystems. Ecol Model 2008; 213: 1-15.

Pedersen ML, Sode A, Kaarup P, Bundgaard P. Habitat Quality in Danish Streams. Testing of Two Indices and Development of a National Physical Habitat Quality Index. Silkeborg: National Environmental Research Institute, Scientific Report Nr. 590; 2006 [In Danish].

Phillips JM, Russell MA, Walling DE. Time-integrated sampling of fluvial suspended sediment: a simple methodology for small catchments. Hydrol Process 2000; 14: 2589-2602.

Raimondo S, Vivian DN, Barron MG. Web-based Interspecies Correlation Estimation (Web-ICE) for Acute Toxicity: User Manual. Version 3.1. Gulf Breeze, FL: U.S. EPA, Office of Research and Development; 2010.

Rasmussen JJ, Baattrup-Pedersen A, Larsen SE, Kronvang B. Local physical habitat quality cloud the effect of predicted pesticide runoff from agricultural land in Danish streams. J Environ Monitor 2011a; 13: 943-50.

Rasmussen JJ, Baattrup-Pedersen A, McKnight US, Wiberg-Larsen P, Kronvang B. Buffer strip width and agricultural pesticide contamination in Danish lowland streams: Implications for stream and riparian management. Ecol Eng 2011b; 37: 1990-97.

Sánchez-Montoya MM, Vidal-Abarca MR, Suárez ML. Comparing the sensitivity of diverse macroinvertebrate metrics to a multiple stressor gradient in Mediterranean streams and its influence on the assessment of ecological status. Ecol Indic 2010; 10: 896-904.

Schäfer RB, Caquet T, Siimes K, Mueller R, Lagadic L, Liess M. Effects of pesticides on community structure and ecosystem functions in agricultural streams of three biogeographical regions of Europe. Sci Total Environ 2007; 382: 272-85.

Schäfer R, Pettigrove V, Rose G, Allinson G, Wightwick A, Von der Ohe PC, et al. Effects of pesticides monitored with three sampling methods in 24 sites on macroinvertebrates and microorganisms. Environ Sci Technol 2011; 45: 1665-72.

Schletterer M, Füreder L, Kuzovlev VV, Beketov MA. Testing the coherence of several macroinvertebrate indices and environmental factors in a large lowland river system (Volga River, Russia). Ecol Indic 2010; 10: 1083-92.

Skriver J, Friberg N, Kirkegaard J. Biological assessment of running waters in Denmark: introdution of the Danish Stream Fauna Index (DSFI). Verhandlungen des Internationalen Verein Limnologie 2000; 27: 1822-30.

Sourisseau S, Basseres A, Perie F, Caquet T. Calibration, validation and sensitivity analysis of an ecosystem model applied to artificial streams. Water Res 2008; 42: 1167-81.

Stoddard JL, Larsen DP, Hawkins CP, Johnson RK, Norris RH. Setting Expectations for the Ecological Condition of Streams: The Concept of Reference Condition. Ecol Appl 2006; 16:1267-76.

40

Styczen M, Wiberg-Larsen P, Aagaard A. Tag pulsen på pesticiderne i vandmiljøet. Vand & Jord 2003; 10(3): 84-7.

Theodoropoulos C, Iliopoulou-Georgudaki J. Response of biota to land use changes and water quality degradation in two medium-sized river basins in southwestern Greece. Ecol Indic 2010; 10: 1231-38.

Thrush SF, Hewitt JE, Hickey CW, Kelly S. Multiple stressor effects identified from species abundance distributions: Interactions between urban contaminants and species habitat relationships. J Exp Mar Bio Ecol 2008; 366: 160-68.

Tomlin CDS. The pesticide manual, a world compendium. Farnham, Surrey, UK: Crop Protection Publications; 2001.

Troldborg M, Lemming G, Binning PJ, Tuxen N, Bjerg PL. Risk assessment and prioritisation of contaminated sites on the catchment scale. J Contam Hydrol 2008; 101: 14-28.

U.S. EPA. Guidelines for Ecological Risk Assessment. Washington, DC: U.S. EPA, Risk Assessment Forum; 1998.

U.S. EPA. ECOTOX database. U.S. EPA; 2011. Vandenbrouck T, Jones OAH, Dom N, Griffin JL, De Coen W. Mixtures of similarly acting

compounds in Daphnia magna: From gene to metabolite and beyond. Environ Int 2010; 36: 254-68.

Von der Ohe PC, Prüss A, Schäfer RB, Liess M, de Deckere E, Brack W. Water quality indices across Europe—a comparison of the good ecological status of five river basins. J Environ Monitor 2007; 9: 970-78.

Von der Ohe PC, De Deckere E, Prüss A, Munoz I, Wolfram G, Villagrasa M, Hein M, Brack W. Toward an integrated assessment of the ecological and chemical status of European river basins. Integrated Environmental Assessment and Management 2009; 5: 50-61.

Von der Ohe PC, Dulio V, Slobodnik J, De Deckere E, Kühne R, Ebert RU, Ginebreda A, De Cooman W, Schüürmann G, Brack W. A new risk assessment approach for the prioritization of 500 classical and emerging organic microcontaminants as potential river basin specific pollutants under the European Water Framework Directive. Sci Total Environ 2011; 409: 2064-77.

Warne MSJ, Hawker DW. The number of components in a mixture determines whether synergistic and antagonistic or additive toxicity predominate: The Funnel hypothesis. Ecotox Environ Safe 1995; 31: 23-8.

Whiteman M, Brooks A, Skinner A, Hulme P. Determining significant damage to groundwater-dependent terrestrial ecosystems in England and Wales for use in implementation of the Water Framework Directive. Ecol Eng 2010; 36: 1118-25.

Wiberg-Larsen P, Windolf J, Baattrup-Pedersen A, Ovesen NB, Larsen SE, Thodsen H, Sode A, Kristensen E, Kjeldgaard A. Danish stream conditions, 2009. Danish EPA: Faglig raport fra DMU Nr. 804, Danmarks Miljkøundersøgelser; 2010 [In Danish].

Winslow SD, V. PB, Martin JJ, Hallberg GR, Munch DJ, Frebis CP, et al. Statistical procedures for determination and verification of minimum reporting levels for drinking water methods. Environ Sci Technol 2006; 40: 281-88.

Yamamoto K, Fukushima M, Kakutani N, Kuroda K. Volatile organic compounds in urban rivers and their estuaries in Osaka, Japan. Environ Pollut 1997; 95: 135-43.

Paper IX

Ah

UPa

b

c

a

ARRA

KSRICUHC

1

wwsdbmtWasqmg

UT

0d

Ecological Engineering 36 (2010) 1126–1137

Contents lists available at ScienceDirect

Ecological Engineering

journa l homepage: www.e lsev ier .com/ locate /eco leng

n integrated model for assessing the risk of TCE groundwater contamination touman receptors and surface water ecosystems

rsula S. McKnighta,c,∗, Simon G. Fundera, Jes J. Rasmussenb, Michael Finkel c, Philip J. Binninga,oul L. Bjerga

Department of Environmental Engineering, Technical University of Denmark, Miljøvej, Building 113, 2800 Kgs. Lyngby, DenmarkDepartment of Freshwater Ecology, National Environmental Research Institute, Århus University, Vejlsøvej 25, 8600 Silkeborg, DenmarkCenter for Applied Geoscience, University of Tübingen, Sigwartstrasse 10, 72076 Tübingen, Germany

r t i c l e i n f o

rticle history:eceived 28 August 2009eceived in revised form 11 January 2010ccepted 12 January 2010

eywords:

a b s t r a c t

The practical implementation of the European Water Framework Directive has resulted in an increasedfocus on the hyporheic zone. In this paper, an integrated model was developed for evaluating the impactof point sources in groundwater on human health and surface water ecosystems. This was accomplishedby coupling the system dynamics-based decision support system CARO-PLUS to the aquatic ecosystemmodel AQUATOX using an analytical volatilization model for the stream. The model was applied to a casestudy where a trichloroethylene (TCE)-contaminated groundwater plume is discharging to a stream. The

ystem dynamicsisk-based approaches

ntegrated modelinghlorinated solventsncertaintyyporheic zone

TCE source will not be depleted for many decades; however, measured and predicted TCE concentrationsin surface water were found to be below human health risk management targets. Volatilization rapidlyattenuates TCE concentrations in surface water. Thus, only a 30-m stream reach fails to meet surfacewater quality criteria. An ecological risk assessment found that the TCE contamination did not impactthe stream ecosystem. Uncertainty assessment revealed hydraulic conductivity to be the most importantsite-specific parameter. These results indicate that contaminant plumes with �g L−1 concentrations of

r syst

ts1afpEchR

m

ontaminated sites TCE entering surface wate

. Introduction

The practical implementation of the European Water Frame-ork Directive (WFD) has generated many new challenges forater managers across Europe. As global exploitation of both

tream water and groundwater increases, it is becoming more evi-ent that managers need to develop an awareness of the linkagesetween these two systems, the roles that these linkages play inaintaining water quality, and how human activities may impair

hem (Hancock, 2002). In recognition of this, implementation of theFD within the individual countries necessitates the evaluation of

ll types of contamination sources (e.g. point and diffuse) within apecific watershed in order to assess their direct impact on water
uality and ecosystem health. It is required that surface water musteet good water quality and minimum ecological criteria, and that
roundwater must have good chemical status.

∗ Corresponding author at: Department of Environmental Engineering, Technicalniversity of Denmark, Miljøvej, Building 113, 2800 Kgs. Lyngby, Denmark.el.: +45 45251412; fax: +45 45932850.

E-mail address: [email protected] (U.S. McKnight).

irhfTqsSfe

925-8574/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.ecoleng.2010.01.004

ems may not pose a significant risk.© 2010 Elsevier B.V. All rights reserved.

Chlorinated solvents, such as trichloroethylene (TCE), and pes-icides are among the most prevalent and serious contaminants ofurface and groundwater resources in the world (e.g. Winter et al.,998; Stroo et al., 2003; Guilbeault et al., 2005). In Denmark this ismajor problem because almost all drinking water comes directly

rom groundwater (Henriksen et al., 2008). And many of these com-ounds are either acknowledged or suspected carcinogens (U.S.PA, 2009b). Due to their widespread use, mobility and persistence,hlorinated volatile organic compounds (VOCs) are considered toave the greatest potential to discharge to surface waters (Ellis andivett, 2007).

Water management decisions are increasingly being based onodel studies (Scholten et al., 2007) and modeling tools are becom-

ng progressively more sophisticated, i.e. parameterized. Existingisk-based studies of coupled groundwater–surface water systemsave tended to focus explicitly on predicting diffuse source trans-

ers to surface waters (Heathwaite et al., 2005; Kannan et al., 2007).he use of these models, however, implies access to enough good
uality data in order to both calibrate and validate the physicalystem before these models can be used in a predictive capacity.impler modeling approaches and tools also exist, but here theocus has typically been on either the groundwater (e.g. Troldborgt al., 2008) or the surface water system (e.g. Ani et al., 2009).


http://www.elsevier.com/locate/ecoleng



U.S. McKnight et al. / Ecological Engineering 36 (2010) 1126–1137 1127

F the Lild umpo sion o

Ria

bscbttepti

biistm(sis(fera

2

its2wiLo

a

vwgspC

sfsTh((4omf

ctivts(t6scp(

tyaiBw

ig. 1. Propagation of the TCE contaminant plume and impacted stream section atots) and stream bed (blue dots), the location of the remediation well (B11) used for pf the references to color in this figure legend, the reader is referred to the web ver

isk-based approaches capable of predicting and quantifying thempacts of groundwater contamination on surface water resourcesnd ecosystem health are currently unavailable.

This paper aimed to bridge this gap by presenting a novel risk-ased, source–pathway–receptor methodology for analyzing pointource impacts to both human and ecological receptors, espe-ially for use at early decision levels. Uncertainty assessments muste incorporated into the decision-making process, with emphasishroughout the modeling process (Refsgaard et al., 2007). Thus,he proposed approach also endeavors to produce “usable” sci-ntific information by specifically addressing the role uncertaintylays on the decision-making process, particularly with respect tohe use of “effective” parameters in groundwater transport model-ng.

In this paper, we show how integrated modeling can supportoth human health and ecological risk assessments for evaluat-

ng surface waters impacted by point sources in groundwater. Thiss accomplished by coupling the system dynamics-based decisionupport system CARO-PLUS to the process-based aquatic ecosys-em model AQUATOX through a simple analytical volatilization

odel. The system dynamics approach implemented in CARO-PLUSSerapiglia et al., 2005; McKnight and Finkel, 2008) is particularlyuited for management issues regarding contaminated land sincet has the ability to incorporate past actions (e.g. previous remedialtrategies) that may have been undertaken to alleviate a probleme.g. groundwater contamination). AQUATOX (Park et al., 2008) wasound to be the most comprehensive of the few existing generalcological risk models, capable of representing the combined envi-onmental fate and effects of toxic chemicals and their impacts toquatic ecosystems.

. Case study site – Skensved stream

A TCE contaminant plume that is leaching from groundwaternto Skensved stream in Denmark is assessed with the new sys-em dynamics tool. The Skensved stream, located on the easternide of the island of Sjaelland in Denmark, has a catchment area of5 km2. Lille Skensved is located in an area with protected drinkingater interests. Lille Skensved Waterworks is situated approx-

mately 1.5 km northwest of the town. A second well field, foryngen Waterworks (Christensen and Raun, 2005), lies 3 km eastf Lille Skensved and immediately south of the Skensved stream.

The aquifer at Skensved is contaminated by TCE originating fromn auto lacquer shop in Lille Skensved which has used the sol-

m2l

o

le Skensved site. Also shown are the measurement locations in the stream (yellowand treat, and the location of the well in the hyporheic zone (3B). (For interpretationf the article.)

ent for degreasing metal parts since 1974. A leaking storage tankas found in 1993 where TCE had been seeping directly into the

round below. In 2003 it was determined that this storage tank wasolely responsible for the TCE contamination in the aquifer, with alume extending up to 1000 m from the source area (see Fig. 1;hristensen and Raun, 2005).

The geology beneath Lille Skensved and Skensved stream con-ists of 2–4 m of alternating layers of soil, gravel, sand and clayollowed by 8–10 m of bryozoan limestone (the Danien Lime-tone) that is underlain by a low permeability zone (GEO, 2009).he primary limestone aquifer is characterized by an effectiveydraulic conductivity of 19 m d−1, a hydraulic gradient of 0.00473Christensen and Raun, 2005), and an effective porosity of 0.02GEO, 2009), resulting in seepage velocities on the order of.5 m d−1. Since no information is available on the mass fraction ofrganic carbon, a typical literature value of 0.002 for Danish aquiferaterials was assumed (corresponding to a retardation value of 44

or TCE, Christensen et al., 1996).Although little data exist regarding the specific source zone

haracteristics (e.g. geometry), measured TCE concentrations (inhe mg L−1 range) in the primary aquifer below the sourcendicate the presence of separate phases of chlorinated sol-ents (Christensen and Raun, 2005). GEO (2009) estimateshat 150–240 tons of TCE have been used in the auto lacquerhop during the period from 1974 to 1999. Furthermore, GEO2009) conclude that the TCE plume is under hydraulic controlhrough the implementation of pump and treat. Approximately0,000–100,000 m3 yr−1 of water have been pumped from theource area, as well as 30,000–100,000 m3 yr−1 from well B11 (indi-ated in Fig. 1; Christensen and Raun, 2005; GEO, 2009). Currentlans are to continue with the pump and treat strategy until 2010GEO, 2009).

The average water flux in the Skensved stream was determinedo be 13,500 m3 d−1 (157 L s−1), based on data taken over the past 20ears, with large interannual and seasonal variation (Christensennd Raun, 2005). In 2005, the water flow decreased from 1200 L s−1

n early January to just 6.3 L s−1 in July and August (data not shown,ruun and Rose, 2005). The changing water flow directly affects theater levels in the stream so that, in 2005, the water depth rose to
ore than 50 cm during the winter months and dropped below
0 cm in the summer. In late September of 2005, water levels asow as 11 cm were observed.

In previous studies, conducted through the Technical Universityf Denmark in 2005 and 2008, the extent of groundwater–surface

1 l Engineering 36 (2010) 1126–1137

wthpsTaccs

ottsisoid1tg

issc1om

trmwobn2gsdt

Fwsc

Table 1Water chemistry and redox parameters measured at Skensved stream (Christensenand Raun, 2005).

Parameter Multilevelsamplers

Seepagemeter

Skensvedstream

pH [–] 6.7–8.1 6.9–7.0 7.6Temperature [◦C] 9.4–15.3 10.3–13.8 10.7–10.8Chloride [mg L−1] 50–278 71–104 24–26Oxygen [mg L−1] 0.2–3.5 0.3–0.5 8.5

3

atmmewwsief

msasothi

3

128 U.S. McKnight et al. / Ecologica

ater interaction at the site has been determined using piezome-ers and temperature measurements. Danish groundwater usuallyas a temperature of 10 ◦C, and so a local surface water tem-erature rise occurs in the winter when groundwater enters thetream, as the surface water is colder than the groundwater.he reverse occurs in the summer. The measurements, takenlong 4 km of the stream (indicated by yellow dots in Fig. 1),learly show a temperature change in the hyporheic zone indi-ating that groundwater influx to the stream occurs (data nothown).

Further evidence of groundwater–surface water interaction isbtained through groundwater monitoring data. The data showhat the TCE concentration in groundwater declines from the westo east side of the stream, as illustrated in Fig. 2, indicating that aignificant fraction of TCE (maximum concentration of 120 �g L−1

n borehole 3B) is entering the Skensved stream. Seepage meteramplers placed in the hyporheic zone show that concentrationsf up to 59 �g L−1 occur in groundwater entering the stream. Thenfiltration rates at three different locations along the stream wereetermined: 336 L m−2 d−1 at 1674 m, 84 L m−2 d−1 at 1765 m and70 L m−2 d−1 at 1943 m (compare distances with Fig. 6). Based onhese seepage meter measurements, the flux of contaminants fromroundwater into the stream could be determined.

In the surface water in August 2005, the TCE concentrationncreased from zero to a maximum of 17.4 �g L−1 along a 250 mtretch where the groundwater plume interacts with the Skensvedtream, and decreased thereafter to zero again. Thus, it was con-luded that the contaminant plume enters the stream between625 m and 1875 m. TCE was also observed in the stream on otherccasions, but annual maximums occurred during the summeronths.The bottom sediments are generally highly permeable and con-

ain insignificant amounts of organic matter. Water chemistry andedox parameters were measured in May–June 2005 using bothultilevel samplers and seepage meters for the influent ground-ater, as well as directly from the surface water (Table 1). Based

n these results, the hyporheic zone could be characterized aseing slightly aerobic and/or nitrate-reducing, thus preventing sig-ificant anaerobic dechlorination (Scheutz et al., 2008; Abe et al.,009). No apparent increase in 1,2-cis-DCE was observed in the
roundwater to the west of the stream (Figs. 1 and 2), or in thetream itself. And so it was concluded that no overall systematicegradation of TCE is taking place in the down-gradient portion ofhe plume.
ig. 2. TCE and 1,2-cis-DCE concentrations [�g L−1] in the groundwater movingest to east (left to right) from the source zone to the east bank of the Skensved

tream (Christensen and Raun, 2005). Well 3B was found to contain the highest TCEoncentrations, which was used for calibrating the source release model.

sptagshcsHemcPa

dtamcgaaM

Nitrate [mg N L−1] 0.4–6.4 1.8–2.0 1.8–1.9Dissolved Iron [mg L−1] <0.3–1.2 n.a. n.a.Sulphate [mg S L−1] 7–41 14–17 9–10

. Methodology: integrated modeling framework

The impacts of TCE on Skensved stream was assessed usingsource–pathway–receptor concept (see Fig. 3). The concep-

ual model was implemented by coupling the system dynamicsodel CARO-PLUS to the U.S. EPA ecological impact assessmentodel AQUATOX. This was necessary since CARO-PLUS is currently

quipped only for the analysis of risks to human health, specificallyhen the contamination plume at the receptor occurs in ground-ater (note however that the source zone can originate either in

oil, groundwater or both). Similarly, the U.S. EPA model AQUATOXs equipped only for the analysis of ecological effects on aquaticcosystems when the contaminants are already present in the sur-ace water.

In order to enable both a quantitative human health risk assess-ent and an ecological risk assessment of this site, the decision

upport system CARO-PLUS was coupled to the process-basedquatic ecosystem model AQUATOX through a simple analyticalurface water volatilization model. The model was constructed inrder to track the fate of TCE as it moves from the groundwa-er through the groundwater–surface water interaction zone (i.e.yporheic zone) into surface water. These steps will be described

n more detail in the following sections.

.1. System dynamics modeling: CARO-PLUS

CARO-PLUS, developed at the Center for Applied Geo-cience/University of Tübingen, is intended to be used forreliminary assessment as part of a tiered approach, and to allowhe user to simulate and optimize the effects of potential remedialctions including tackling the contaminant source and managingroundwater plumes (McKnight, 2009). It currently consists of aource release module, a contaminant transport module and auman health impact assessment module. Both mass release andontaminant transport in groundwater are quantified using tran-ient models that are based on analytical approaches (Sauty, 1980;untley and Beckett, 2002; Eberhardt and Grathwohl, 2002; Petert al., 2008). The existing contamination and its further develop-ent can be evaluated on the basis of contaminant mass fluxes,

oncentrations, and risk indices (carcinogenic/non-carcinogenic).ossible remedial actions causing mass flux changes over time arelso simulated (McKnight, 2009).

CARO-PLUS also contains a risk assessment module thatescribes exposure pathways, i.e. transport (transfer) of con-aminants from groundwater to the receptor “human being”, inccordance to the Multimedia Environmental Pollutant Assess-ent System developed by Strenge and Smith (2006). The model is

apable of analyzing pollutant behavior in various media (air, soil,roundwater and surface water) and estimating transport throughnd between media. The uncertainty inherent in both site-specificnd exposure parameters is explicitly taken into account usingonte-Carlo simulations.


recep

(stwpupaffitm(

3D

dbacfiDfl

J

watc

D

wm

aoatea

3

sflb

J

Fig. 3. Schematic illustration of the source–pathway–

CARO-PLUS is built on the system dynamics platform VensimVentana Systems, 2007). System dynamics uses an interlinkedystem of stocks (levels) and flows (rates). Water, as well as con-aminant mass can be thought of and measured using stock terms,hich describe the volume or mass present at a particular place andoint in time. Similarly, water and mass transfer can be calculatedsing flow terms, which describe the volume or mass reaching orassing a defined point or area in a given time period. The over-ll structure of the model as implemented in a systems dynamicsramework is described in Fig. 4. The stocks are represented in thegure using boxes and the flows by double arrows. The single arrowerms show the most important inputs required in calculations. The

odel has been tested for appropriateness and verified in McKnight2009).

.1.1. Source characterization: contaminant emission fromNAPL pool

The source model conceptually distinguishes between light andense NAPL contamination scenarios, as well as residual phase orlob zones. The organic contaminant phase is typically described asmixture of multiple compounds, but single compound scenarios

an also be modeled. Mass release from a DNAPL pool is quanti-ed using a model which describes the dispersive mass transfer ofNAPL into groundwater flowing across the pool, with the resultant
ux being given by (Eberhardt and Grathwohl, 2002):
dnapli

= Csatw,iBpLpne

√4Dvx

�Lp(1)

w[Lx

Fig. 4. Major structure for the DNAPL pool release scenario and transport alo

tor concept implemented in the program CARO-PLUS.

here Jdnapli

[M T−1] is the dissolution rate of compound i frompool with width Bp [L] and length Lp [L], and D [L2 T−1] is the

ransverse vertical hydrodynamic dispersion coefficient, which isalculated using:

= Dp + ˛tvvx (2)

here Dp [L2 T−1] is the pore diffusion coefficient (also approxi-ated as Daqne) and ˛tv [L] is the transverse vertical dispersivity.The values for transverse vertical dispersivity are estimated by

ssuming that dispersion can be represented as a linear functionf groundwater pore velocity (Klenk and Grathwohl, 2002). Otherssumptions inherent to this solution include constant concentra-ion at the NAPL–water interface where dissolution occurs, andquilibrium is assumed to have been reached between the NAPLnd a boundary layer of water (i.e. bordering the DNAPL pool).

.1.2. Groundwater contaminant transportThe mass transfer of the contaminant from the source to the

tream is modeled analytically, assuming steady-state (averaged)ow conditions, advective and diffusive transport, retardation andiodegradation (i.e. when applicable) by (Sauty, 1980):

R = JSE

2erfc

(Lx − vx,RtR

2√

˛Lvx,RtR

)e−�iLx/vx (3)

here JR [M T−1] is the estimated mass flux at the receptor, JSE

M T−1] is the mass flux at the down-gradient edge of the source,x [L] is the distance between the source and the receptor in the-direction, ˛L [L] is the longitudinal dispersivity, tR [T] is the sim-

ng the groundwater pathway, including retardation and degradation.

1 l Engi

ucpfd1

˛

3

fvon(

ef

V

wCgCc

C

Wb

C

wugo

3

3

tbtaiwrsdduq

wd

iear

(p

wcitte

(pd(itURbErccm2

wftwIs[ipvpu

3

130 U.S. McKnight et al. / Ecologica

lation time, �i [T−1] is the first-order biodegradation rate forompound i and vx [L T−1] is the groundwater pore velocity. Thearameter values employed are the average (or effective) valuesor the pathway between the source and receptor. The longitudinalispersivity is directly estimated in CARO-PLUS by (Xu and Eckstein,995):

L = 0.83[log(Lx)]2.414 (4)

.2. Analytical volatilization model

A stream transport model is developed to describe the trans-er of mass into the stream from groundwater and the subsequentolatilization of contaminant from the stream. The model is basedn the observation that volatilization is likely to be the domi-ant process affecting the concentration of VOCs in stream waterRathbun, 2000).

The model considers a stream of constant volume, where inflowquals the outflow. The change in concentration of a VOC over timeor a given volume of water is given by:

dC(t)dt

= QgwCgw − KvC(t)V (5)

here V [L3] is the volume of the water body, Qgw [L3 T−1] andgw [M L−3] are the flux and VOC concentration of the infiltratingroundwater, respectively, Kv [T−1] is the volatilization rate and(t) [M L−3] is the VOC concentration at time t [T]. If the initial VOConcentration is C(0) = 0, then the solution of (5) is given by:

(t) = CgwQgw

KvV− e−KvtCgwQgw

KvV(6)

hen the initial VOC concentration is not zero, the solutionecomes:

(t) = CgwQgw

KvV+ e−Kvt

(Ct−1 − CgwQgw

KvV

)(7)

here Ct−1 [M L−1] is the initial VOC concentration. Eq. (7) can besed for scenarios where the infiltration rates Qgw and/or the VOCroundwater concentrations Cgw are not constant over the distancef the modeled stream.

.3. Risk assessment

.3.1. Human health risk assessmentThe risk to human health due to direct or indirect exposure

o contaminated water (e.g. ingestion of water or leafy vegeta-les) can be assessed using a methodology that is applicableo both groundwater and surface water. In CARO-PLUS, the riskssessment is focused on the risks associated with pollutants orig-nating in groundwater. Here the human health risk assessment

as extended to also consider surface water contaminants. Theisk assessment begins with identified concentrations of the con-tituents of interest which are then converted to average dailyoses at the point of exposure (e.g. receptor location in Fig. 3). Theoses are then converted to risk values for either carcinogens (eval-ated as total risk level, RLtot,i [–]) or non-carcinogens (total hazarduotient, HQtot,i [–]) using:

RLtot,i =nE∑

k=1

LADDi × SFi

nE(8)

HQtot,i =∑k=1

CADDi

RfDi

here LADDi [M M−1 T−1] is the lifetime (cancer) average dailyose, SFi [M M−1 T−1]−1 is the corresponding (oral) slope factor, k

iacc

neering 36 (2010) 1126–1137

s the exposure pathway, i is the pollutant, nE [–] is the number ofxposure pathways considered, CADDi [M M−1 T−1] is the chronicverage daily dose and RfDi [M M−1 T−1] is the corresponding (oral)eference dose.

The risk values (HQtot,i; RLtot,i) are then summed over all theknown) compounds considered for a particular site in order toroduce a risk (or hazard) index:

RI =nC∑i=1

RLtot,i

HI =nC∑i=1

HQtot,i

(9)

here RI [–] is the risk index, nC [–] is the number of compoundsonsidered and HI [–] is the hazard index. Major assumptionsnclude that exposure to any amount of a carcinogen will increasehe cancer risk (i.e. no threshold dosage), risks are additive for mul-iple chemicals and (exposure) routes, and potential synergisticffects (between compounds) are not considered.

LADDi and CADDi result from exposure pathway functionsStrenge and Smith, 2006) that are based on a specific set ofarameters (see Eq. (10)). These parameters are typically set toefault values considered to be standard for a particular risk groupe.g. children) and toxicity assessment (e.g. carcinogens). The tox-city assessment is conducted using existing databases such ashe Integrated Risk Information System (IRIS), maintained by the.S. Environmental Protection Agency (U.S. EPA, 2009b), and theisk Assessment Information System (RAIS) that is maintainedy Oak Ridge National Laboratory for the U.S. Department ofnergy (RAIS, 2009). The goal is to provide an estimate of theelationship between the magnitude of exposure and severity (non-arcinogens) or likelihood (carcinogens) of adverse effects. Thus, forarcinogens and for each exposure pathway, LADDi can be deter-ined by a set of equations following the U.S. EPA (1991, 2001,

004) and ASTM (2004):

LADDi,ing = Ci,w

[IRw

EF × ED

BW × AT × 365 d yr−1

]

LADDi,inh = Ci,a

[IRa

EF × ED

BW × AT × 365 d yr−1

]K

LADDi,dc = Ci,w

[SA × ET

EF × ED

BW × AT × 365 d yr−1

]Kp × CF1

(10)

here LADDi,ing, LADDi,inh and LADDi,dc are the average daily dosesor the ingestion, inhalation and dermal contact pathways, respec-ively, Ci,w [mg L−1] and Ci,a [mg m−3] are the concentrations inater and air, respectively, IRw [L d−1] is the water ingestion rate,

Ra [m3 d−1] is the air inhalation rate, SA [m2] is the exposed skinurface, ET [hr d−1] is the exposure time for an outdoor activity, EFd yr−1] is the frequency of exposure to the contaminant, ED [yr]s the exposure duration, BW [kg] is the bodyweight, AT [yr] is theeriod over which the time is averaged, K [L m−3] is the Andelmanolatilization factor for chemical pollutants, Kp [cm hr−1] is the skinermeability coefficient for chemical i and CF1 [L m cm−1 m−3] is anit conversion factor.

.3.2. Ecological risk assessment
This paper extends the CARO-PLUS model to assess the ecolog-
cal impact of contaminants in surface water. An ecological riskssessment can be carried out in several ways. A rudimentary (i.e.onservative, screening level) method is to compare exposure pointoncentrations with ecotoxicity values for a set of ecological recep-


Table 2General (best estimate) parameter values used in the base case scenario, including aquifer, site and contamination parameters.

Estimate Reference

Aquifer parametersThickness [m] 10 GEO (2009)Hydraulic conductivity [m d−1] 19 Calibration parameter; Christensen and Raun (2005)Hydraulic gradient [–] 0.00473 Christensen and Raun (2005)Effective porosity [–] 0.02 GEO (2009)Seepage velocity [m d−1] 4.5 Calculated

Site parametersAquifer material organic carbon content [–] 0.002 Christensen et al. (1996)Depth to GW [m] 2 GEO (2009)Longitudinal dispersivity [m] 6.86 Calculated; Xu and Eckstein (1995)Transverse vertical dispersivity [m] 0.0001 Klenk and Grathwohl (2002)Distance from source to P&T well [m] 500 Christensen and Raun (2005)Distance from source to receptor [m] 750 Christensen and Raun (2005)Minimum (summer) surface water flux [m3 d−1] 544 Bruun and Rose (2005)

Contamination parametersPool width [m] 11 Calibration parameterPool length [m] 15 Calibration parameterThickness [m] 10 Calibration parameter

−1 .46

t

H

weiutrnr

hae2mTwapbflsit

3

ehiicutl

t

mtronCdio

4

4

tpicgel

sz–mb(zap

ofi

NAPL molecular weight [g mol ] 131NAPL density [kg L−1] 1.4Interfacial tension NAPL/water [dynes cm−1] 42Interfacial tension NAPL/air [dynes cm−1] 32

ors using (U.S. EPA, 1997a):

Qi = EECi

NOAECi(11)

here HQi [–] is the hazard quotient for pollutant i, EECi [M L3] is thestimated concentration at the exposure point and NOAECi [M L3]s the no-observed-adverse-effects-concentration. The NOAECi val-es were taken from the online database ECOTOX maintained byhe U.S. EPA (2009a). The U.S. EPA advises that these conservativeesults can be used to determine whether ecological threats areegligible or whether a more detailed ecological risk assessment isequired.

The simple ecological assessment approach described aboveas been criticized since NOAECi reference values exist only forfew ecological receptors and a high uncertainty exists for the

xtrapolation of these values to other organisms (Tannenbaum,005). A more comprehensive ecological risk assessment can beade using the freshwater ecosystem simulation model AQUA-

OX. AQUATOX combines the simulation of an aquatic ecosystemith the environmental fate and effect of various pollutants, such

s nutrients and organic chemicals. The model is capable of com-uting endpoint concentrations for pollutants in both water andottom sediments, as well as their effects on a variety of aquaticora and fauna. A detailed description of AQUATOX including thepecific calculations for ecosystem health parameters is includedn Park and Clough (2004), Park et al. (2008) and referencesherein.

.4. Uncertainty assessment

One of the goals of this paper is to carry out a quantitativexposure assessment, including an in-depth investigation intoow uncertainty influences management decisions. By segregat-

ng the uncertainties that are specific for the exposure parametersn a particular land use scenario from those that are site- andontaminant-specific, the investigator can work to reduce the
ncertainties associated with the site, while being able to recognizehe uncertainties inherent in the exposure process. This processeads to robust management strategies.
Uncertainty was investigated using Monte-Carlo simulation ofhe site-specific parameters within CARO-PLUS. A Latin hypercube

B1AF(

U.S. EPA (2009b)U.S. EPA (2009b)Mayer and Hassanizadeh (2005)Mayer and Hassanizadeh (2005)

ethod was used to sample parameter values from given distribu-ions. After generation of the Monte-Carlo output, the simulationesults were post-processed to remove model outcomes that lieutside a predefined acceptable range. This “model screening” is aecessary measure since parameter values inherent to the Monte-arlo process are drawn randomly from their individual probabilityistributions, and can therefore be combined in ways that are phys-

cally infeasible and their consideration would decrease the validityf the results.

. Results: application to Skensved stream

.1. Source characterization and groundwater transport model

The CARO-PLUS model was utilized to simulate the TCE flux fromhe source area to the hyporheic zone as a function of time. Table 2resents a summary of the general parameter values implemented

n the base case scenario. Hydraulic conductivity was used as aalibration parameter for (model) transport, based on the rangeiven in Christensen and Raun (2005). All further transport param-ters could be taken either from site-specific investigations or theiterature.

The historical conditions in the source zone are not known, ando observed concentrations measured down-gradient of the sourceone for two different points in time (1995 – pre-pump and treatand 2005) are used to calibrate the source zone geometry in theodel (see results in Table 2). Measurement information for cali-

ration of the transport model is taken from remediation well B11500 m down-gradient, Fig. 1) and well 3B located in the hyporheicone (750 m down-gradient, Figs. 1 and 2). The model calibrationssumes that TCE degradation is not occurring at this site, due to therevailing redox conditions and absence of degradation products.

Calibration of the model results in an initial volume and massf TCE of 33 m3 and 48.2 tons, respectively. The calibrated modelts well to concentrations measured in 1995 of 280 �g L−1 at well
11 (compared with 309 �g L−1 modeled – data not shown) and60 �g L−1 in well 3B (compared with 174 �g L−1 modeled, Fig. 5B).concentration of 120 �g L−1 was measured in well 3B in 2005 (see
ig. 2), again comparing well with the modeled value of 129 �g L−1

Fig. 5B).

1132 U.S. McKnight et al. / Ecological Engineering 36 (2010) 1126–1137

F oncena gets a

4c

fsHb1Atsgdnwrccwawm

pr2J(

TSa

oaE

p(w(im

gat1shmrptp1e

ig. 5. Output of the source zone and groundwater models: (A) mass depletion, (B) cnd (D) risk index. Also shown, as a dashed horizontal line, are the management tar

.2. Human health risk assessment based on groundwateroncentrations

The results from the groundwater transport model can be usedor an initial calculation of risk to human health. The cancer clas-ification for TCE is still under review by the U.S. EPA (2009b).owever, it is currently classified as a possible-probable carcinogeny the Agency for Toxic Substances and Disease Registry (ATSDR,997), and is listed as “medium priority” by the Internationalgency for Research on Cancer (IARC, 2008). Thus, it makes sense

o conduct the risk assessment for the case that TCE can be clas-ified as carcinogenic. In addition, the most sensitive target riskroup is chosen – children. Table 3 lists the exposure pathwaysetermined to be potentially relevant for this receptor. It should beoted that CARO-PLUS considers receptor endpoints such as surfaceater (i.e. fin fish ingestion, shoreline sediment ingestion) from a

isk perspective. Contaminant concentrations in surface water arealculated by a simple dilution model, i.e. through mixing of theontaminated groundwater with surface water. The risk scenarioas conducted for the worst-case, i.e. fish are assumed to ingestmixture of groundwater and stream water where the amountsere determined by the groundwater flow rate and the low (sum-er) observed surface water flux.Table 3 shows the results of the risk assessment: the exposure

athways considered and their associated maximum (calculated)
isk index. The maximum concentration was calculated to be47 �g L−1 in groundwater used for drinking (first occurring inuly 2021) and the risk index was determined to be 5.68 × 10−4
October 2021). These can be compared with target risk levels

able 3ummary of the human health assessment including initial set of exposure pathwaysnd their corresponding maximum RIi,k .

Exposure pathways Maximum RIi,k [–]

Drinking water ingestion 5.65E−04Fin fish ingestion 2.23E−06Shoreline sediment ingestion 3.63E−11Soil ingestion 3.75E−08Inhalation of re-suspended soil 2.71E−10Outdoor air inhalation 1.68E−07Soil dermal contact 6.25E−08Shoreline sediment dermal contact 1.21E−10Maximum RIi [–] 5.68E−04

4

mtctda

pTiattw

tration at stream receptor (750 m), (C) aqueous mass flux from DNAPL source zonessociated with (B) TCE concentration and (D) risk index.

f 10−6 and maximum allowable TCE concentrations of 10 �g L−1

nd 1 �g L−1 in surface and groundwater, respectively (Miljø-ognergiministeriet, 1996; Miljøstyrelsen, 1998).

Of the eight pathways considered for the receptor, only twoathways play a significant role in causing risk to the receptormarked in bold in Table 3). The drinking water ingestion path-ay assumes that groundwater is used directly for drinking water

e.g. from a private well) without treatment. The sum of the riskndices for the six least important pathways is 2.69 × 10−7, thereby

eeting the risk management target of 10−6.Fig. 5 shows overall results of the risk assessment based on

roundwater concentrations, including (A) mass depletion, (B)queous mass flux from NAPL source zone, (C) concentration athe receptor (750 m) and (D) risk index for a simulation period of00 years. The figure shows the impact of the pump and treat (P&T)ystem used for source control and the outcome if pump and treatad not been employed. The effect of the pump and treat contain-ent strategy can be seen in the decreasing concentrations and

isk between 1999 and 2009. The model shows that if the currentump and treat system is switched off in 2009, then the concen-ration and risk will rebound in 2014. This means that when theump and treat action is terminated, both management targets of�g L−1 in groundwater (Fig. 5B) and RI = 10−6 (Fig. 5D) will bexceeded in the base case scenario.

.3. Surface water transport model

The parameter values utilized in the analytical volatilizationodel are summarized in Table 4. Several of the parameters in

he model are known to be nonstationary (e.g. surface water flow,ross-sectional area), in part due to seasonal changes. To evaluatehe worst-case, the model was set up to simulate the conditionsuring the summer, which is characterized by low water fluxesnd water levels.

Using the volatilization and groundwater inflow rates as fittingarameters, the volatilization model was fitted to the measuredCE concentrations, as shown in Fig. 6. Three separate groundwater
nflow sections were identified, with the first occurring at 1625 mnd approximately 250 m long, the second only 75 m long, and thehird encompassing the remainder of the investigated stream sec-ion. The initial values for the first two groundwater inflow ratesere taken from the measured data (e.g. 336 L m−2 d−1 at 1674 m


Table 4Input parameters for the volatilization model.

Parameter Value Reference

TCE concentration in groundwater [�g L−1] 130 CARO-PLUS outputStream depth [m] 0.11 Minimum value; Bruun and Rose (2005)Stream width [m] 1.4 Average value; Bruun and Rose (2005)Surface water flux [m d−1] 3366 Bruun and Rose (2005)Volatilization rate [d−1] 19Groundwater inflow rate I [L m−2 d−1] 360Groundwater inflow rate II [L m−2 d−1] 30Groundwater inflow rate III [L m−2 d−1] 18

Ftag

aacta

wnonc

4c

thw

csiw

ctfuwtustfp

4

lDroraoDafo

TP

ig. 6. Analytical volatilization model (solid line) fitted to actual TCE concentra-ions (dashed line) measured in August 2005, by accounting for both volatilizationnd upstream inflows. The two “jumps” were fitted by including downstreamroundwater-to-surface water influx areas.

nd 84 L m−2 d−1 at 1765 m) and then adjusted. No initial data werevailable for the third rate. The figure shows two “jumps” in the TCEoncentration at 2125 m and 2300 m and these can be explained byhe model if it includes the increased groundwater infiltration ratest these points (see Fig. 6).

The model can be used to determine the stream reach overhich the surface water quality criteria for TCE has been exceeded,amely between 1700 m and 2000 m (Fig. 6). That is, only 300 mf the Skensved stream is actually affected by the TCE contami-ation (i.e. fails to meet surface water quality criteria) from theontaminated groundwater.

.4. Human health risk assessment based on surface wateroncentrations

The maximum modeled (and observed) TCE concentration inhe surface water of 17.4 �g L−1 was used to assess the humanealth risk for children. This second human health risk assessmentas conducted specifically for the potential exposure scenario of

aBclm

able 5arameter values used in the assessment of human health (i.e. children) exposure to cont

Parameter V

Exposure frequency, EF [d yr−1] 9Exposure time for outdoor activity, ET [hr d−1]Surface water ingestion rate, IRw [L d−1]Inhalation rate, IRa [m3 d−1]Body weight, BW [kg] 2Averaging time, AT [yr] 7Unit conversion factor, CF1 [L m cm−1 m−3] 1Andelman volatilization factor, K [L m−3]Exposure duration, ED [yr]Exposed skin surface area, SA [m2]Skin permeability coefficient, Kp [cm hr−1]Fraction of contaminant absorbed in the gastrointestinal tract, GI [–]Oral slope factor for ingestion pathway, SFo [kg d mg−1]Oral slope factor for inhalation pathway, SFi [kg d mg−1]Oral slope factor for dermal contact pathway, SFd [kg d mg−1]

Fitting parameterFitting parameterFitting parameterFitting parameter

hildren playing outdoors in the contaminated stretch of Skensvedtream. For a recreational land use scenario, risk from surface waters the sum of ingestion, inhalation and dermal contact with the

ater.Isolating LADDi in Eq. (8) and then solving for the respective

ontaminant concentrations from Eq. (10) allows the determina-ion of the maximum allowable concentrations in water and airor a specific risk target, e.g. RI = 10−6. Using the parameter val-es listed in Table 5, the maximum allowable concentrations inater for the ingestion and dermal contact pathways were found

o be 54.2 �g L−1 and 25.4 �g L−1, respectively. Both of these val-es are above the TCE concentration actually measured in thetream. Similarly, the maximum allowable concentration in air forhe inhalation pathway was found to be 0.7 �g L−1. No data existor TCE concentrations in the air in this vicinity for comparisonurposes.

.5. Preliminary ecological risk assessment in the stream

Indicator organisms were selected for the screening-level eco-ogical risk assessment (e.g. according to Eq. (11)), includingaphnia magna (aquatic crustaceans), mayfly and minnow (rep-

esenting fin fish). TCE toxicity data can be found for these threerganisms in the U.S. EPA ECOTOX database (U.S. EPA, 2009a). Theespective hazard quotients were calculated to be 0.013, 0.0004nd 0.001, respectively, and all were well below the target valuef 1. Neither D. magna nor minnow are typically found in smallanish streams (Dall and Lindegaard, 1995). However, D. magnand minnow were chosen to represent Gammarus pulex (commonreshwater shrimp) and (juvenile) Salmo trutta (brown trout), bothf which are common inhabitants of small Danish streams (Dall
nd Lindegaard, 1995). Furthermore, species of the mayfly family,aetidae, are frequently found in Danish streams. The chosen indi-ator organisms are thus assumed to represent different trophicevels in a simplified Danish stream ecosystem. NOAECs for D.
agna and G. pulex are comparably low in terms of other toxicant

aminated surface water.

alue Notes/references

0 Best estimate, recreational scenario0.75 Best estimate, recreational scenario0.05 U.S. EPA (1997b)7.5 ECETOC (2001)1.4 ECETOC (2001)5 ECETOC (2001)0 Default; U.S. EPA (1991)0.5 Default; U.S. EPA (1991)6 Default; U.S. EPA (1991)0.9052 Calculated as 423 cm2/kg; ECETOC (2001)0.0157 Chemical-specific; RAIS (2009)1.0 Chemical-specific; RAIS (2009)0.4 Chemical-specific; RAIS (2009)0.4 Chemical-specific; RAIS (2009)0.4 Chemical-specific; calculated as SFo/GI (U.S. EPA, 2004)


Table 6Parameter values used in AQUATOX for the assessment of ecological health.

Animal toxicity data Initial conditions [g m−2 dry] LC50 [�g L−1] LC50 experimental time [h] Lipid fraction [–] Average wet weight [%]

Stonefly 0.5a 70,000b 48b 0.05a 0.03a

Oligochaete 1a 132,000b 48b 0a 0a

Minnow 2a 52,000c 24c 0.047a 2a

Ostracod 0.65a 56,000b 48b 0.05a 0.002a

Chironomid 0.5a 42,000b 48b 0.06a 0.0006a

D. magna 0.03a 22,000a 24a 0.06a 0.0006a

Plant toxicity data Initial conditions [g m−2 dry] EC50 photo [�g L−1] EC50 experimental time [h] Lipid fraction [–]

Macrophytes 2a 0 0 0.02a

Bluegreens 1.2a 63,000a 192a 0.05a

Greens 0.05a 390,000c 96c 0.05a

96a 0.05a

gas

4

aspmauAaBthsisaepPb(mgLnbd

camIact

cmitt

Diatoms 1.2a 150,000a

a U.S. EPA (2009a).b Kegley et al. (2008).c Rippen (1995).

roups (e.g. pesticides, Møhlenberg et al., 2004), and we thereforessume that D. magna here is representative of G. pulex as the mostensitive organism at the TCE polluted stretch in Skensved stream.

.6. Comprehensive ecological risk assessment in the stream

In order to conduct a more comprehensive ecological riskssessment, the AQUATOX model was set up to model the 300 mtretch of the Skensved stream impacted by the TCE groundwaterlume. Although the Skensved site was extensively characterized,uch of the data needed for input to AQUATOX were not avail-

ble (e.g. lipid fraction for the various organisms modeled). Instead,nknown parameters were based on an existing case study inQUATOX, where the impact of an organic toxicant is simulatednd verified for a generic Ohio creek (see Table 6 for parameters).iomasses of the included autotrophic components were main-ained at low levels because of the small stream size and expectedigh level of human maintenance activities. Shading effects oftream bank vegetation increase proportionately with decreas-ng stream size (Allan, 1995). Furthermore, the vast majority ofmall Danish streams have lowered stream beds due to heavynd frequent maintenance which, additionally, increase shadingffects of bank vegetation. We therefore propose that primaryroduction is low at this impacted stretch of Skensved stream.lease note that some of the ecotoxicological parameters coulde either estimated directly from the parameters given in Table 6e.g. animal/plant K2 elimination rate constants), or were esti-

ated as suggested within the existing case study (e.g. animal EC50rowth/reproduction values were estimated using the D. magnaC50/EC50 ratio; see U.S. EPA, 2009c for more details). The “Man-ings Equation” method was selected for computing the waterody volume in AQUATOX, based on existing (dynamic) outflowata for 2008.

Fig. 7A shows the observed stream discharge and modeled TCEoncentration for 1 year. In July, the water volume reaches itsnnual minimum of 20 m3 (over 300 m, corresponding to 0.8 m3 per

stream), and the TCE concentration peaks at just over 16 �g L−1.n general, the pattern of peaks seen on both curves shows thatclear correlation exists between the water volume and the TCE

oncentration, with high water volumes reflecting low TCE concen-rations.

Fig. 7B illustrates the modeling results for the calculated bioac-
umulation factor for four species (D. magna, stonefly, mayfly andinnow) versus TCE concentration. A trend clearly exists, with an
ncreasing bioaccumulation factor from the lowest to the highestrophic level (i.e. moving from D. magna to minnow and bottom toop in Fig. 7B). It can also be seen that the bioaccumulation factor

Fig. 7. AQUATOX results for (A) stream discharge and (B) bioaccumulation factorversus TCE concentration for the 300 m groundwater-impacted stream stretch, and(C) half-lives for both TCE in water and sediment, as well as time required for 95%TCE loss in both water and sediment.


Fig. 8. Monte-Carlo realizations of the transport and risk assessment model for (A) concenrun is indicated in both figures as a dashed line. The model screening criteria (predicted cin (A) by the range allowed (vertical bar) in 2005.

Table 7Site-specific parameters varied in the Monte-Carlo sensitivity test.

(Input) Parameter Estimate Min Max Peak

iM

mmpofi1

lts(ftaSop

4

trpa

orv

Toot±2

mttcdtvctoicdd

vwoqto

TS

Hydraulic conductivity [m d−1] 19 9.5 28.5 –Transverse vertical dispersivity [m] 0.0001 0.00004 0.00015 –DNAPL pool length [m] 15 10 30 15

ncreases with TCE concentration in the summer months (i.e. fromay to September).D. magna was found by AQUATOX to be the most sensitive species

odeled (data not shown). However, a close comparison of the D.agna concentration between the perturbed and control (no TCEresent) scenarios indicates that there were no ecosystem effectsccurring. This low ecosystem impact is expected as the NOAECor this organism and TCE is 1384 �g L−1 (U.S. EPA, 2009a), whichs much higher than the observed surface water concentration of7.4 �g L−1.

The half-life of a contaminant provides a good indication of theength of time in which a stream will be impacted by the con-aminant. Fig. 7C depicts the modeled TCE half-life in water andediment (bottom two curves) and the time for 95% chemical lossupper two curves). Both parameters in water (versus sediment)ollow the trend for TCE concentration (depicted in Fig. 7B). In con-rast, both half-life and time for 95% chemical loss in sedimentre less affected by the dynamically changing TCE concentration.imulations were carried out for 5 years (2008–2012), and it wasbserved that the trends evident in Fig. 7 simply repeat themselves,roducing a yearly cycling pattern (data not shown).

.7. Uncertainty assessment

Monte-Carlo simulation was used to assess the uncertainty ofhe results from the groundwater transport model, focusing on theole of the highly uncertain or effective parameters on the decisionrocess. Table 7 lists the parameters and ranges applied for thenalysis. Input parameters were described using either triangular

wivr

able 8ummary statistics for the Monte-Carlo simulations for selected input and output param

Parameter Min Max

Input parametersHydraulic conductivity [m d−1] 10.96 20.69Transverse vertical dispersivity [m] 4.0E−05 1.5E−04DNAPL pool length [m] 10.95 29.41

Output parametersConcentration at receptor [mg L−1] 0.153 0.489RI [–] 3.52E−04 1.12E−03

tration at receptor and (B) risk index over time. Note that the base case (BC) modeloncentrations at receptor = observed concentration at receptor ± 50%) is indicated

r uniform distributions, where the degree of variation (i.e. spread)eflected the estimated uncertainty for a specific parameter (e.g.an Groenendaal and Kleijnen, 2002).

Simulations were done for an ensemble of 500 realizations.he outcome was then screened in order to generate a subsetf realizations that are consistent with actual site data. A subsetf 112 realizations was produced utilizing a screening criterionhat requires the predicted (model) concentrations to be within50% of the observed concentration measured at the receptor in005.

Uncertainties in input data (Table 7) lead to uncertainties inodel output. The simulation results show that TCE concentra-

ion and resulting risk to the receptor may be approximately threeimes higher or two times lower than was predicted for the basease (Fig. 8). Interestingly, two “trends” or sets of curves can beistinguished in Fig. 8. A deterministic sensitivity analysis usinghe allowable, i.e. screened parameter ranges showed that loweralues of hydraulic conductivity were responsible for the set ofurves appearing with a later breakthrough time, i.e. in 2000. Bothransverse vertical dispersivity and DNAPL pool length had nobservable effect on the time for the first breakthrough, but werenstead responsible for the resulting “spread” of both maximumoncentration and risk indices. This also plays a role in model pre-ictions in which contaminant concentrations can be expected toecrease as a result of the pump and treat strategy.

A synopsis of the results of the Monte-Carlo simulation is pro-ided in Table 8. Ranges of input parameters are narrow comparedith the previously defined ranges (see Table 7) as a consequence

f model screening. The coefficient of variation (CV) may be used touantify how the original uncertainties have “propagated” throughhe modeling system. The probability distributions of input andutput parameters appear to be very similar.

The results of the sensitivity analysis for the output variablesere also analyzed for the October 2021 (indicated as t = 10-2021

n Fig. 9) time step corresponding to the time of maximum riskalue, using histograms. For both concentration at receptor andisk index, the base case values fell into the ranges containing the

eters (at t = 10-2021).

Median Mean StDev CV

18.26 16.64 3.46 0.219.0E−05 9.3E−05 3.2E−05 0.35

17.04 17.89 4.43 0.25

0.273 0.279 0.065 0.236.26E−04 6.41E−04 1.49E−04 0.23


F at recs s.

lr

5

bwdmtmofap

ardtrm

cicmtotm5

mhtcvTtbmptb

cenhtiDfnciien

ametodt

isassla

A

G0mDtt

ig. 9. Histograms showing the probability distribution for (A) TCE concentrationpecific subset range containing the base case modeled value is indicated by arrow

argest probability with predicted probabilities of 24% and 22%,espectively.

. Conclusions

This paper has shown how integrated modeling can supportoth human health and ecological risk assessments for surfaceaters potentially impacted by point sources in groundwater. Theecision support system CARO-PLUS allows for a model-based,ulti-compartmental environmental impact assessment designed

o establish the suitability of management scenarios capable ofeeting predefined, site-specific compliance criteria. It is based

n a quantitative exposure assessment and determination of risksor potential receptors deemed relevant for a particular site. Thispproach is intended to evaluate and reduce the number of com-etitive options for any further detailed assessment.

The results of the screening tool at the Skensved site indicaten unsettling trend, with TCE concentrations (in groundwater)eaching a maximum of 250 �g L−1 that will hold steady for manyecades to come. The plume today is under hydraulic control, buthe results suggest that pump and treat, functioning either as aemediation or as a containment strategy, may not be a sustainableanagement solution for this site.For the human health risk assessment in surface water, no (car-

inogenic) risk was found for the developed worst-case scenario,.e. children, in a recreational setting. Measured and modeled TCEoncentrations in surface water were found to be below the esti-ated maximum allowable concentrations in water for meeting

he risk management target of 10−6. These results are dependentn the actual values chosen for the scenario. Risk was only foundo exist if the groundwater was to be used as drinking water, with

aximum modeled concentration and risk values of 247 �g L−1 and.68 × 10−4, respectively.

The results of the volatilization model could be used to deter-ine the stream reach over which surface water quality criteria

ave been exceeded. This corresponded to the 300 m stretch wherehe TCE plume actually infiltrates the Skensved stream. TCE con-entrations, as well as water levels were used for a preliminaryerification to support the quantitative ecological risk assessment.he applied ecosystem model AQUATOX was found to capture theserends well, when compared with actual data. Two major draw-
acks, however, were found to exist when using the AQUATOXodel for preliminary assessment: the extremely large number of
arameters needed to create a functioning ecosystem, and the facthat specific organisms not currently included in the model cannote added by the user.

pUbFt

eptor and (B) risk index at the expected time of maximum risk (t = 10-2021). The

For the ecological risk assessment, it was found that the TCEontamination does not have any significant effect on the streamcosystem. These results indicate that volatile organic solvents mayot pose a threat to surface water ecosystems. Caution is warranted,owever, since the only chemical investigated was TCE and ecosys-em effects were modeled based on the impact of TCE on threendicator organisms assumed to represent common species in smallanish streams. This is also reflected in the high NOAEC values

ound for all organisms investigated. Further modeling studies areeeded, including studies comparing ecosystem impacts for bothhronic and acute toxicants. We additionally propose that conduct-ng subsequent supplementary field studies is highly necessary tomprove the evaluation of modeling results, when ecosystem mod-ling input is restricted to only a few species which potentially areot present at the site in question.

Uncertainty with respect to transport modeling parameters waslso investigated. The hydraulic conductivity was found to be theost critical site-specific parameter, suggesting that additional

ffort should be directed towards it when prioritizing future inves-igation needs. Sensitivity analyses confirmed that an evaluationf uncertainties for site-specific parameters is critical to soundecision-making and must be taken into consideration in any quan-itative assessment.

The implementation of system dynamics to hydrogeologicalssues recognizes the increasing importance of interdisciplinaryystems research that relates policy assessment to resource man-gement options. The risk assessment tools developed in this papererve to integrate multiple issues, interest groups, disciplines andcales in order to address the wide-ranging impacts of control-able and uncontrollable socioeconomic, ecological, hydrologicalnd institutional drivers.

cknowledgements

The authors gratefully acknowledge the support of theerman BMBF research priority program KORA (contract no.2WN0382 and 02WN0300), the Helmholtz Centre for Environ-ental Research in Leipzig (contract no. UFZ-01/2007) and theanish Research Council (grant no. 2104-07-0035). Travel grants to

he HydroEco2009 conference in Vienna, Austria, were provided byhe Technical University of Denmark. The field data reported in this
aper are the results of several studies conducted at the Technicalniversity of Denmark by Stine Brok Christensen, Kristian Drags-aek Raun, Simon Bruun, Jonas Rose, Anna Juliane Clausen, Mettejendbo Petersen, Eva Mathilde Riis Hedegaard, and supported byheir advisors Poul L. Bjerg, Peter Bauer-Gottwein and Anders Baun.

l Engin

R

A

A

A

A

A

B

C

C

D

E

E

E

G

G

H

H

H

H

I

K

K

K

M

M

M

M

M

M

P

P

P

R

R

R

R

S

S

S

S

S

S

T

T

U

U

U

U

U

U

U

U

v

Vdate: July 4, 2007. Ventana Systems, Inc., Harvard (MA), USA.

U.S. McKnight et al. / Ecologica

eferences

be, Y., Aravena, R., Zopfi, J., Parker, B., Hunkeler, D., 2009. Evaluating the fateof chlorinated ethenes in streambed sediments by combining stable isotope,geochemical and microbial methods. J. Contam. Hydrol. 107 (1–2), 10–21.

llan, J.D., 1995. Stream Ecology. Structure and Function of Running Waters, 1stedition. Kluwer Academic Publishers, Holland.

ni, E.-C., Wallis, S., Kraslawski, A., Agachi, P.S., 2009. Development, calibration andevaluation of two mathematical models for pollutant transport in a small river.Environ. Modell. Softw. 24, 1139–1152.

STM, 2004. Standard Guide for Risk-based Corrective Action, E2081-00 (Re-approved). American Society for Testing and Materials, ASTM International,West Conshohocken, PA.

TSDR, 1997. Toxicological Profile for Trichloroethylene. U.S. Department of Healthand Human Services, Agency for Toxic Substances and Disease Registry, Divisionof Toxicology, Atlanta, GA.

ruun, S., Rose, J., 2005. Påvirkning af vandløb fra punktforurenet grundvand. Pre-Masters Thesis. Institute for Environmental Engineering, Technical Universityof Denmark, Kgs. Lyngby, Denmark.

hristensen, S.B., Raun, K.D., 2005. Udsivning af punktkildeforurenet grundvand tiloverfladevand i et Vandrammeperspektiv. Master’s Thesis. Institute for Environ-mental Engineering, Technical University of Denmark, Kgs. Lyngby, Denmark.

hristensen, T.H., Lehmann, N., Jackson, T., Holm, P.E., 1996. Cadmium and nickel dis-tribution coefficients for sandy aquifer materials. J. Contam. Hydrol. 24, 75–84.

all, P.C., Lindegaard, C., 1995. An Overview of Danish Freshwater Invertebrates forAssessment of Stream and Lake Contamination. Freshwater Biology Laboratory,Department of Biology, University of Copenhagen.

berhardt, C., Grathwohl, P., 2002. Time scales of organic contaminant dissolutionfrom complex source zones: coal tar pools vs. blobs. J. Contam. Hydrol. 59, 45–66.

CETOC, 2001. Exposure Factors Sourcebook for European Populations (with Focuson UK Data), Technical Report 79. European Centre for Ecotoxicology and Toxi-cology of Chemicals, Brussels, ISSN: 0773-8072-79.

llis, P.A., Rivett, M.O., 2007. Assessing the impact of VOC-contaminated groundwa-ter on surface water at the city scale. J. Contam. Hydrol. 91, 107–127.

EO, 2009. Lille Skensved, grundvandsforurening med chloreredeopløsningsmidler. Vurdering af afvaergeanlaeg, GEO projekt nr. 31817,Rapport 1, 2009-02-04.

uilbeault, M.A., Parker, B.L., Cherry, J.A., 2005. Mass and flux distributions fromDNAPL zones in sandy aquifers. Ground Water 43 (1), 70–86.

ancock, P.J., 2002. Human impacts on the stream–groundwater exchange zone.Environ. Manage. 29, 763–781.

eathwaite, A.L., Dils, R.M., Liu, S., Carvalho, L., Brazier, R.E., Pope, L., Hughes, M.,Phillips, G., May, L., 2005. A tiered risk-based approach for predicting diffuseand point source phosphorus losses in agricultural areas. Sci. Total Environ. 344,225–239.

enriksen, H.J., Troldborg, L., Hoejberg, A.L., Refsgaard, J.C., 2008. Assessment ofexploitable groundwater resources of Denmark by use of ensemble resourceindicators and a numerical groundwater–surface water model. J. Hydrol. 348,224–240.

untley, D., Beckett, G.D., 2002. Persistence of LNAPL sources: relationship betweenrisk reduction and LNAPL recovery. J. Contam. Hydrol. 59, 3–26.

ARC, 2008. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans.Internal Report 08/001, Report of the Advisory Group to Recommend Prioritiesfor IARC Monographs during 2010–2014. International Agency for Research onCancer, World Health Organization, Lyon, France.

annan, N., White, S.M., Whelan, M.J., 2007. Predicting diffuse-source transfers ofsurfactants to surface waters using SWAT. Chemosphere 66, 1336–1345.

egley, S.E., Hill, B.R., Orme, S., Choi, A.H., 2008. PAN Pesticide Database – Chem-icals. Pesticide Action Network, North America, San Francisco, CA, Available:http://www.pesticideinfo.org [August 11, 2009].

lenk, I.D., Grathwohl, P., 2002. Transverse vertical dispersion in groundwater andthe capillary fringe. J. Contam. Hydrol. 58, 111–128.

ayer, A.S., Hassanizadeh, M., 2005. Soil and Groundwater Contamination: Non-aqueous Phase Liquids. American Geophysical Union, Washington, DC.

cKnight, U.S., 2009. A System Dynamics Approach for the Integrative Assess-ment of Contaminated Land Management Options. PhD Thesis. TübingerGeowissenschaftliche Arbeiten (TGA), Reihe C: Hydro-, Ingenieur-, und Umwelt-geologie, Center for Applied Geoscience, Eberhard-Karls-Universität Tübingen,C114, Tübingen, Germany.

cKnight, U.S., Finkel, M., 2008. Model-based Preliminary Assessment of Mega-siteManagement Options: A New Approach to Improve Tiered Decision-Making.IAHS Red Book Series 324, pp. 499–506.

iljøstyrelsen, 1998. Oprydning fra forurenede lokaliteter. Vejledning fraMiljøstyrelsen, nr. 5 1998. Danish Environmental Protection Agency.

iljø-og Energiministeriet, 1996. Bekendtgørelse om kvalitetskrav for vandområder
og krav til udledning af visse farlige stoffer til vandløb, søer eller havet. BEK nr.921 af 08/10/1996 (Gaeldende). Danish Ministry for Environment and Energy.
øhlenberg, F., Kaas, H., Schlüter, L., Gustavson, K., Andersen, T.T., Forbes, V., Cold,A., Friberg, N., Larsen, S.E. Lauridsen, R.B., 2004. Effects of biocides and pesticideson flora and fauna in streams. Bekaempelsesmiddel-forskning fra MiljøstyrelsenNo. 82, Danish Ministry for Environment and Energy.

W

X

eering 36 (2010) 1126–1137 1137

ark, R.A., Clough, J.S., 2004. AQUATOX (Release 2): Modeling EnvironmentalFate and Ecological Effects in Aquatic Ecosystems. Technical Documentation,EPA/823/R-04/002. Office of Water, United States Environmental ProtectionAgency, Washington, DC.

ark, R.A., Clough, J.S., Wellman, M.C., 2008. AQUATOX: modeling environmentalfate and ecological effects in aquatic ecosystems. Ecol. Model. 213, 1–15.

eter, A., Miles, B., Teutsch, G., 2008. Estimation of emission from an LNAPL contam-inated zone considering groundwater recharge. Environ. Geol. 55, 321–337.

AIS, 2009. Chemical-specific Toxicity Values. Risk Assessment Information Sys-tem, Oak Ridge National Laboratory, U.S. Department of Energy, Available:http://rais.ornl.gov/tox/tox values.shtml [August 11, 2009].

athbun, R.E., 2000. Transport, behaviour, and fate of volatile organic compoundsin streams. Crit. Rev. Environ. Sci. Technol. 30 (2), 129–295.

efsgaard, J.C., van der Sluijs, J.P., Hojberg, A.L., Vanrolleghem, P.A., 2007. Uncertaintyin the environmental modelling process—a framework and guidance. Environ.Modell. Softw. 22 (11), 1543–1556.

ippen, G., 1995. Trichlorethylene, Ergänzungslieferung 31. 17/95. HandbuchUmwelt-Chemikalien: Stoffdaten, Prüfverfahren, Vorschriften. 3. Auflage mit71. Ergänzungslieferung. Band 1-9. andsberg/Lech, Ecomed Sicherheit. Verlags-gruppe Hüthig Jehle Rehm Gmbh, 1990–2005.

auty, J.P., 1980. An analysis of hydrodispersive transfer in aquifers. Water Resour.Res. 16 (3/4), 145–158.

cheutz, C., Durant, N.D., Dennis, P., Hansen, M.H., Jørgensen, T., Jakobsen, R., Cox,E.e., Bjerg, P.L., 2008. Concurrent ethene generation and growth of dehalo-coccoides containing vinyl chloride reductive dehalogenase genes during anenhanced reductive dechlorination field demonstration. Environ. Sci. Technol.42 (24), 9302–9309.

cholten, H., Kassahun, A., Refsgaard, J.C., Kargas, T., Gavardinas, C., Beulens, A.J.M.,2007. A methodology to support multidisciplinary model-based water manage-ment. Environ. Modell. Softw. 22, 743–759.

erapiglia, F., Sievers, J., Finkel, M., 2005. Risk-based Design of Cost-efficient Reme-diation Strategies at Megasites. IAHS Red Book Series 297, pp. 249–255.

trenge, D.L., Smith, M.A., 2006. Multimedia Environmental Pollutant AssessmentSystem (MEPAS): Human Health Impact Module Description. PNNL-16164.Pacific Northwest National Laboratory, Prepared for the U.S. Department ofEnergy, Contract DE-AC06-76RLO 1830.

troo, H.F., Unger, M., Ward, C.H., Kavanaugh, M.C., Vogel, C., Leeson, A., Marqusee,J.A., Smith, B.P., 2003. Remediating chlorinated solvent source zones. Environ.Sci. Technol. 37 (11), 224A–230A.

annenbaum, L.V., 2005. A critical assessment of the ecological risk assessmentprocess: a review of misapplied concepts. Integr. Environ. Assess. Manage. 1,66–72.

roldborg, M., Lemming, G., Binning, P.J., Tuxen, N., Bjerg, P.L., 2008. Risk assess-ment and prioritisation of contaminated sites on the catchment scale. J. Contam.Hydrol. 101, 14–28.

.S. EPA, 1991. Risk Assessment Guidance for Superfund: Volume I – Human HeathEvaluation Manual (Part B, Development of Risk-based Preliminary Remedia-tion Goals). EPA/540/R-92/003. United States Environmental Protection Agency,Office of Emergency and Remedial Response, Washington, DC.

.S. EPA, 1997a. Ecological Risk Assessment Guidance for Superfund: Process forDesigning and Conducting Ecological Risk Assessments. Interim Final. UnitedStates Environmental Protection Agency, Environmental Response Team, Edi-son, NJ.

.S. EPA, 1997b. Exposure Factors Handbook, Volumes I–III. EPA/600/P-95/002Fa,b,c. United States Environmental Protection Agency, Office of Researchand Development, Washington, DC.

.S. EPA, 2001. Appendix Q: Human Health Benchmarks. United States Environmen-tal Protection Agency, Office of Solid Waste, Washington, DC.

.S. EPA, 2004. Risk Assessment Guidance for Superfund: Volume I – Human HeathEvaluation Manual (Part E, Supplemental Guidance for Dermal Risk Assessment).OSWER 9285.702EP. United States Environmental Protection Agency, Office ofSuperfund Remediation and Technology Innovation, Washington, DC.

.S. EPA, 2009a. ECOTOX Database. United States Environmental Protection Agency,Available: http://www.epa.gov/med/Prods Pubs/ecotox.htm [August 11, 2009].

.S. EPA, 2009b. Integrated Risk Information System (IRIS). United StatesEnvironmental Protection Agency, Available: http://www.epa.gov/ncea/iris/help start.htm [August 11, 2009].

.S. EPA, 2009c. Water Quality Models and Tools. United States Environmental Pro-tection Agency, Available: http://www.epa.gov/waterscience/models/aquatox/[August 11, 2009].

an Groenendaal, W.J.H., Kleijnen, J.P.C., 2002. Deterministic versus stochastic sen-sitivity analysis in investment problems: an environmental case study. Eur. J.Oper. Res. 141, 8–20.

entana Systems, 2007. Vensim Simulation Environment – User’s Guide, Revision

inter, T.C., Harvey, J.W., Franke, O.L., Alley, W.M., 1998. Ground Water and SurfaceWater: A Single Resource. U.S. Geol. Surv. Circ. 1139.

u, M., Eckstein, Y., 1995. Use of weighted least-squares method in evaluationof the relationship between dispersivity and scale. Ground Water 33 (5),905–908.

http://www.pesticideinfo.org/

http://rais.ornl.gov/tox/tox_values.shtml

http://www.epa.gov/med/Prods_Pubs/ecotox.htm

http://www.epa.gov/ncea/iris/help_start.htm

http://www.epa.gov/ncea/iris/help_start.htm

http://www.epa.gov/waterscience/models/aquatox/

pesticide effects on the structure and function of stream ecosystems · anthropogenic stressors to...

Documents