predicting the impact of perturbations on salmon ( oncorhynchus spp.) communities: implications for...

10
Predicting the impact of perturbations on salmon (Oncorhynchus spp.) communities: implications for monitoring M.R. Arkoosh, L. Johnson, P.A. Rossignol, and T.K. Collier Abstract: Twenty-six salmon (Oncorhynchus spp.) stocks from the Pacific Northwest are listed as either threatened or endangered. A number of anthropogenic factors, likely including degradation of habitat by chemical contaminant expo- sure, have contributed to their decline. Techniques that can assess injury or judge the efficacy of regulatory actions on the recovery of this species are needed. We strive to understand why a population is changing by examining changes in their intrinsic birth rates, death rates, and (or) growth rates. However, salmon populations are influenced by other spe- cies in the community. To address this issue, we developed a parsimonious three-trophic-level community model con- sisting of prey, salmon, and parasites and examined the model’s response to one anthropogenic factor (contaminant exposure) using qualitative analysis. This community model may not only provide valuable insight into salmon survival but also may broaden the approaches available to elucidate direct and indirect effects. We demonstrate analytically that some community members, possibly salmon themselves, might be ambiguous or unreliable variables to monitor. We also demonstrate that other species in the community, such as parasites, may be more sensitive than salmon in monitor- ing the influence of anthropogenic factors such as contaminants. Résumé : Vingt-six stocks de saumons (Oncorhynchus spp.) de la région du nord-ouest pacifique d’Amérique du Nord sont menacés ou en danger. Plusieurs facteurs anthropiques, incluant vraisemblablement la dégradation des habitats par exposition aux contaminants chimiques, ont contribué à leur déclin. Il y a un besoin de techniques qui permettent d’évaluer les dommages et de juger de l’efficacité des mesures réglementaires sur la récupération de ces espèces. Nous cherchons généralement à comprendre les raisons des changements dans une population en mesurant les variations dans les taux intrinsèques de natalité, de mortalité et (ou) de croissance. Cependant, les populations de saumons sont affec- tées par les autres espèces de la communauté. Pour répondre à cette question, nous avons mis au point un modèle par- cimonieux de communauté à trois niveaux trophiques, soit les proies, les saumons et les parasites, et avons étudié, par analyse qualitative, les réactions du modèle à un facteur anthropique, l’exposition aux contaminants. Ce modèle de communauté fournit non seulement des informations précieuses sur la survie des saumons, mais il peut en plus procu- rer des nouvelles avenues pour élucider les effets directs et indirects. Notre analyse démontre que certains membres de la communauté, éventuellement les saumons eux-mêmes, peuvent s’avérer être des variables ambiguës ou incertaines à suivre durant la surveillance écologique. Elle démontre aussi que d’autres espèces dans la communauté, telles que les parasites, peuvent être plus sensibles que le saumon pour indiquer l’influence des facteurs anthropiques, tels que les contaminants. [Traduit par la Rédaction] Arkoosh et al. 1175 Introduction Human activities, such as overfishing, industrial and urban development, landscape alteration, and release of toxic chemicals into the environment (Kennish 1992), have dam- aged the health of North American fishery stocks. These fac- tors likely have brought a number of salmon populations in the Pacific Northwest close to extinction (Nehlsen et al. 1991; Quinn 1994; National Research Council 1996). To date, 26 salmon (Oncorhynchus spp.) stocks have been listed as either threatened or endangered in Washington, Oregon, Idaho, and California (National Marine Fisheries Service 2000). The need for effective ways of monitoring the health of salmon species is well recognized and is particularly highlighted by the requirement to monitor the status and re- covery of listed species. However, it is not always obvious in a community which species and endpoints will provide the most useful information on how environmental stressors Can. J. Fish. Aquat. Sci. 61: 1166–1175 (2004) doi: 10.1139/F04-068 © 2004 NRC Canada 1166 Received 12 May 2003. Accepted 23 December 2003. Published on the NRC Research Press Web site at http://cjfas.nrc.ca on 3 September 2004. J17523 M.R. Arkoosh. 1 Environmental Conservation Division, Northwest Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 2030 South Marine Science Drive, Newport, OR 97365, USA. L. Johnson and T.K. Collier. Environmental Conservation Division, Northwest Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 2725 Montlake Boulevard East, Seattle, WA 98112, USA. P.A. Rossignol. Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, OR 97331-3590, USA. 1 Corresponding author (e-mail: [email protected]).

Upload: independent

Post on 12-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Predicting the impact of perturbations on salmon(Oncorhynchus spp.) communities: implications formonitoring

M.R. Arkoosh, L. Johnson, P.A. Rossignol, and T.K. Collier

Abstract: Twenty-six salmon (Oncorhynchus spp.) stocks from the Pacific Northwest are listed as either threatened orendangered. A number of anthropogenic factors, likely including degradation of habitat by chemical contaminant expo-sure, have contributed to their decline. Techniques that can assess injury or judge the efficacy of regulatory actions onthe recovery of this species are needed. We strive to understand why a population is changing by examining changes intheir intrinsic birth rates, death rates, and (or) growth rates. However, salmon populations are influenced by other spe-cies in the community. To address this issue, we developed a parsimonious three-trophic-level community model con-sisting of prey, salmon, and parasites and examined the model’s response to one anthropogenic factor (contaminantexposure) using qualitative analysis. This community model may not only provide valuable insight into salmon survivalbut also may broaden the approaches available to elucidate direct and indirect effects. We demonstrate analytically thatsome community members, possibly salmon themselves, might be ambiguous or unreliable variables to monitor. Wealso demonstrate that other species in the community, such as parasites, may be more sensitive than salmon in monitor-ing the influence of anthropogenic factors such as contaminants.

Résumé : Vingt-six stocks de saumons (Oncorhynchus spp.) de la région du nord-ouest pacifique d’Amérique du Nordsont menacés ou en danger. Plusieurs facteurs anthropiques, incluant vraisemblablement la dégradation des habitats parexposition aux contaminants chimiques, ont contribué à leur déclin. Il y a un besoin de techniques qui permettentd’évaluer les dommages et de juger de l’efficacité des mesures réglementaires sur la récupération de ces espèces. Nouscherchons généralement à comprendre les raisons des changements dans une population en mesurant les variations dansles taux intrinsèques de natalité, de mortalité et (ou) de croissance. Cependant, les populations de saumons sont affec-tées par les autres espèces de la communauté. Pour répondre à cette question, nous avons mis au point un modèle par-cimonieux de communauté à trois niveaux trophiques, soit les proies, les saumons et les parasites, et avons étudié, paranalyse qualitative, les réactions du modèle à un facteur anthropique, l’exposition aux contaminants. Ce modèle decommunauté fournit non seulement des informations précieuses sur la survie des saumons, mais il peut en plus procu-rer des nouvelles avenues pour élucider les effets directs et indirects. Notre analyse démontre que certains membres dela communauté, éventuellement les saumons eux-mêmes, peuvent s’avérer être des variables ambiguës ou incertaines àsuivre durant la surveillance écologique. Elle démontre aussi que d’autres espèces dans la communauté, telles que lesparasites, peuvent être plus sensibles que le saumon pour indiquer l’influence des facteurs anthropiques, tels que lescontaminants.

[Traduit par la Rédaction] Arkoosh et al. 1175

Introduction

Human activities, such as overfishing, industrial and urbandevelopment, landscape alteration, and release of toxicchemicals into the environment (Kennish 1992), have dam-aged the health of North American fishery stocks. These fac-tors likely have brought a number of salmon populations inthe Pacific Northwest close to extinction (Nehlsen et al.1991; Quinn 1994; National Research Council 1996). To

date, 26 salmon (Oncorhynchus spp.) stocks have been listedas either threatened or endangered in Washington, Oregon,Idaho, and California (National Marine Fisheries Service2000). The need for effective ways of monitoring the healthof salmon species is well recognized and is particularlyhighlighted by the requirement to monitor the status and re-covery of listed species. However, it is not always obviousin a community which species and endpoints will providethe most useful information on how environmental stressors

Can. J. Fish. Aquat. Sci. 61: 1166–1175 (2004) doi: 10.1139/F04-068 © 2004 NRC Canada

1166

Received 12 May 2003. Accepted 23 December 2003. Published on the NRC Research Press Web site at http://cjfas.nrc.ca on3 September 2004.J17523

M.R. Arkoosh.1 Environmental Conservation Division, Northwest Fisheries Science Center, National Marine Fisheries Service,National Oceanic and Atmospheric Administration, 2030 South Marine Science Drive, Newport, OR 97365, USA.L. Johnson and T.K. Collier. Environmental Conservation Division, Northwest Fisheries Science Center, National Marine FisheriesService, National Oceanic and Atmospheric Administration, 2725 Montlake Boulevard East, Seattle, WA 98112, USA.P.A. Rossignol. Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, OR 97331-3590, USA.

1Corresponding author (e-mail: [email protected]).

impact community members. Nor is it obvious how manage-ment activities aimed at changing those stressors ultimatelyaffect the abundance of a target species such as salmon.

Numerous efforts to set environmental or regulatory crite-ria focus on protecting ecological resources at the populationlevel. They emphasize the measurement of endpoints such asgrowth, survival, and reproduction, which are clearly relatedto population growth rate. For example, the US Environmen-tal Protection Agency’s water quality criteria (US Environ-mental Protection Agency 1997, 1998) use this approach.Similarly, in ecological risk assessment, endpoints that arerelevant to population abundance are generally considereduseful because of their high ecological relevance (US Envi-ronmental Protection Agency 1997, 1998). Essentially, theseapproaches use a population- rather than community-levelassessment. In a population analysis, impacts arising fromthe community are summarized, or hidden, in the survivaland fertility of a single species. This analysis is common be-cause a life table approach is relatively straightforward. Theproblem is that interpretation can be ambiguous because somany factors are lumped into population parameters. In acommunity model, the interacting variables, such as species,are dissected out and impacts on all community memberstheoretically are analyzable. A difficulty arises from the taskof documenting and measuring all relationships. This prob-lem can be approached by taking a qualitative perspectivefor which a set of mathematical tools has been developed(Dambacher et al. 2002, 2003a, 2003b) and applied to fish-eries (Castillo et al. 2000). These two modeling approaches,which focus on the population and on the community, arebest considered as being complementary, not mutually ex-clusive.

For salmon recovery efforts, growth rate (λ) can be calcu-lated from a Leslie population matrix with data from directtracking of return rates; in turn, linking λ to environmentalvariables can be suggested as a means to devise a strategy(e.g., Kareiva et al. 2000). However, populations do not existindependently of each other but as a community of relation-ships with other species, their predators, prey, parasites,pathogens, or competitors (Preston 2002). Similarly, envi-ronmental stressors may not affect just a single target spe-cies but also affect the species with which they interact.Because of these indirect effects, the ultimate impact of astressor on target species abundance can be ambiguous andis not always easily predictable. In fact, it has been sug-gested that indirect effects may have a greater influence on apopulation than direct effects (Lampert et al. 1989; Menge1995). Simply monitoring changes in the target species pop-ulation may not always be the best way to assess injury or tojudge the efficacy of remedial or regulatory actions.

A tenet put forth by Karr (1998) is that the condition of awatershed is reflected by the health of the fish communitiesresiding in those waters. Fausch et al. (1990) outlined fourmain approaches that have been used when monitoring fishcommunities for determining water quality. Briefly, thesefour monitoring approaches are as follows: indicator taxa orguilds; species richness, diversity, and evenness; multivariatemethods; and the index of biotic integrity. Only one ap-proach specifically examines the proportion of individualswith disease, namely the index of biological integrity. How-ever, this assessment is only concerned with the presence of

externally evident disease and parasites of the fish and notthe presence of a parasite population in the environment intotal. Parasite assemblages of fish may be a more sensitiveindicator of environmental stressors, such as contaminants,than fish themselves because they reflect their host’s interac-tions with benthic, planktonic, and fish communities(Landsberg et al. 1998). In other words, a parasite respondsto all of its life cycle hosts. This interaction allows the para-site to be an indicator species of trophic interactions (Marco-gliese 2001).

Fish harbor both micro- and macro-parasites (May 1983).Microparasite models focus on prevalence, while macropara-site models are concerned with density and distribution. Ingeneral, toxicology studies have focused on macroparasitepopulations (e.g., Valtonen et al. 1997; MacRury and John-son 1999). Few studies have examined the effect of contami-nants on microparasites.

In this study, based on community theory, we propose ageneral modeling approach to improve monitoring activity.Specifically, we determine how different trophic levels of afood chain respond with different degrees of ambiguity to aperturbation that affects more than one level. As a specificexample, we examine a parsimonious three-trophic-levelcommunity consisting of prey, salmon, and parasites affectedby perturbation of exposure to chemical contaminants. Chem-ical contaminant exposure was chosen as the perturbation ofinterest because its effects on different components of thecommunity can be estimated from previous studies. Ouranalysis of this model is described in three sections. The ini-tial section provides a brief theoretical presentation on theuse of qualitative modeling in research. The second sectiondiscusses the proposed variables (benthic prey, salmon, andparasites) of our community matrix as well as how eachvariable is influenced by contaminant exposure. The thirdsection derives the predicted effects on a parasite–salmon–prey community after exposure to a defined stress, in thiscase contaminant exposure. These analyses suggest that in athree-variable chain, the top level demonstrates the only un-ambiguous change under a variety of scenarios. Thus, if cer-tain parasite populations occupy this niche, they would showpromise as a variable to monitor for detecting the impacts ofcontaminant-associated environmental degradation on asalmon community.

Qualitative community modeling

Qualitative models are typically constructed and analyzedin the following manner (Puccia and Levins 1985). A so-called signed digraph (directed graph) first is drawn to illus-trate the direct relationships between each variable. In thethree-trophic-level model or “signed digraph”, positive influ-ences that would tend to increase population abundance arepictorially represented with a line arrow. A line with a circleat the end of it represents a negative influence on the vari-able where the circle ends. The pairwise combination of apositive and negative arrow thus represents a predator–preyrelationship between two variables. The self-effect symbol, acircle within a circle, connects the variable to itself. A nega-tive self-effect is used to represent a population’s self-regulatory capacity. This type of effect can also representother complex and poorly characterized environmental fac-

© 2004 NRC Canada

Arkoosh et al. 1167

tors that limit the population but are not dealt with specifi-cally in the community matrix model. We provide a primeron analysis of complex communities or loop analysis (Ap-pendix A).

Perturbation experiments are considered to be of either a“pulse” or a “press” nature (Bender et al. 1984). A pulseperturbation is nonsustained; pulse experiments examine acommunity as it returns to equilibrium after it has been per-turbed temporarily. During a press (sustained) perturbationexperiment, species densities are continually altered by achange in mortality rate of these species. Its impact on allcommunity variables can be “predicted” theoretically(Bender et al. 1984; Yodzis 1988). Contaminant exposure ofa community is considered to act as a press perturbation(Yodzis 1988), although a pulse is clearly a possibility in mi-grating fish. Press perturbation experiments yield informa-tion on both direct and indirect effects resulting from theperturbation. Pulse perturbation experiments only provideinformation on direct effects (Bender et al. 1984) and havenot been studied extensively.

The inverse of the negative of the matrix predicts changesin the population density that arises from press perturba-tions. The procedures measuring the direct and indirect ef-fects in a community resulting from a press perturbation canbe done quantitatively, but since this type of data is virtuallyabsent from the ecological literature (Dambacher et al.2002), analyses have by necessity been qualitative. Untilrecently, this limitation was a serious impediment, since pre-dictions were all-or-none. New theoretical advances andmathematical algorithms have made qualitative analysis moreflexible and highly predictive (Dambacher and Rossignol2001; Wootton 2002; Dambacher et al. 2003a). Recent ex-perimental evidence indicates that qualitative (loop) analysisis the theoretical approach that is best predictive of the be-havior of a complex community following a press perturba-tion because it incorporates complex loops and nontrophicinteractions (Hulot et al. 2000).

Community structure

We propose that a predator–prey chain of three variables,essentially three trophic levels that parsimoniously repre-sents the community, would be the most clearly affected bycontaminant exposure in our salmon community. For thisspecific analysis, the variables in our community are benthicinvertebrates (prey), salmon (predator), and parasites (toppredator) (Figs. 1a and 1b). In our current model (Fig. 1),the prey positively influences the salmon population, and thesalmon population negatively influences the prey. Similarly,the salmon population positively influences parasites, whereparasites negatively influence salmon. Additionally, each vari-able has a negative self-effect, indicating that there is somedegree of negative feedback on the population owing to un-identified environmental influences, which could includeother predators and competitors, as well as physical environ-mental features.

The impact of contaminants on the variables within ourcommunity is a press perturbation, which is a sustainedchange in the strength of a parameter (Bender et al. 1984;Schmitz 1997). Intuitively, a press is a change in the birth ordeath rate of a variable. Based on our quantitative studies

and the literature, polychlorinated biphenyls (PCBs) andpolycyclic aromatic hydrocarbons (PAHs) negatively affector press salmon and their prey. Once exposed, salmon aremore susceptible to virulent microparasites (Arkoosh et al.1991, 1994). However, the press perturbation of PCBs andPAHs on parasite populations in the community can have anegative or positive affect on the population depending onthe kind of parasite.

As discussed above, the variables that we have chosen forour community are prey, salmon, and their parasites. Allthree of these variables interact with another variable in ourcommunity. A number of anthropogenic factors may affectsalmon. Our own studies have focused on the influence of aparticular anthropogenic factor, contaminants, on salmonhealth. We have determined through 15 years of study thatPCBs and PAHs alter salmon health (reviewed in Arkooshand Collier 2002). These studies have demonstrated that PCBsand PAHs can reduce growth rate and immune responsive-ness of juvenile salmon and increase their susceptibility toinfectious disease. All of these effects have the potential forreducing salmon populations. However, monitoring popula-tion and community impacts of anthropogenic influences isdifficult. At this point, we have not established a quantitativerelationship between contaminant impacts on salmon healthand salmon abundance. Nor have we considered these re-sponses in a community context or examined potential indi-

© 2004 NRC Canada

1168 Can. J. Fish. Aquat. Sci. Vol. 61, 2004

Fig. 1. (a) Community signed digraph, (b) community matrix,and (c) inverse of the negative of the community matrix. Thediagraph symbolizes the ecological relationship between popula-tions in a community. Each link (arrow) is represented as an ele-ment of the community matrix. Our salmonid communityconsists of parasites, salmon, and their prey. The symbols orlinks characterize the effects that a population variable or specieshas on other populations and corresponds to an element of thecommunity matrix. The qualitative inverse of the negative of thecommunity matrix tabulates the results of a positive press to aparticular variable on prey, salmon, and parasites and are readdown the appropriate column.

rect effects of PCBs and PAHs on salmon populationsmediated through the impacts of these contaminants on para-sites or prey. We put forward, based on a qualitative com-munity theory model, which variable(s) in the communityshould best be monitored to identify effects of contaminantson salmon populations.

SalmonPacific salmon have disappeared from approximately 40%

of their historical Pacific Northwest habitat (Nehlsen et al.1991). Many of the remaining stocks that were once abun-dant have now declined precipitously (Quinn 1994). The ma-jority of the human population in the Pacific Northwest ofthe United States and Canada puts a high value on salmon,and as a result, these fish have great symbolic as well aseconomic significance (National Research Council 1996).The salmon’s decline has led the US federal government tolist many “salmonid evolutionarily significant units” as ei-ther endangered or threatened to confer protection to thesedisappearing species. Consequently, it is important to under-stand the factors contributing to their decline and to monitorhow remedial actions contribute to their recovery.

Juvenile chinook salmon (Oncorhynchus tshawytscha) canbe exposed to contaminants, such as PCBs and PAHs, asthey migrate from their resident rivers to the ocean (McCainet al. 1990; Stein et al. 1995; Stehr et al. 2000). Juvenilesalmon exposed in both the field and the laboratory to PCBsand PAHs are immunosuppressed (Arkoosh et al. 1991, 1994)and more susceptible to disease (Arkoosh et al. 1998, 2001)than those not exposed to the contaminant. It has also beendetermined that PCB and PAH exposure (Casillas et al.1995a, 1995b) affects the growth rate of the young salmon.Since our data support the theory that contaminant exposurecould directly contribute to declines in salmon abundance bynegatively influencing salmon survival and reproductive po-tential, we examined the model with a negative press pertur-bation of contaminants on salmon.

PreyPrey species of juvenile salmon, primarily benthic and

epibenthic invertebrates, also accumulate substantial levelsof chlorinated hydrocarbons and aromatic hydrocarbons whenthey reside in contaminated sites. Several studies have docu-mented elevated concentrations of these contaminants instomach contents of juvenile salmon from Pacific Northwestestuaries (McCain et al. 1990; Collier et al. 1998; Stehr et al.2000), and additional field and laboratory studies confirmthe uptake of PCBs and PAHs by major juvenile salmon preyspecies such as amphipods, midges (Chironomidae) (Fry andFisher 1990; Hwang et al. 2001), mysids (Lester andMcIntosh 1994), and cladocerans (Dillon et al. 1990; Evanset al. 1991; Kucklick et al. 1995).

As with salmon, PCBs appear to impact survival and re-production of these animals. For example, Nebeker andPuglisi (1974) examined eight Aroclor mixtures and their ef-fects on survival and reproduction in Daphnia magna (waterflea), Gammarus pseudolimnaeus (amphipod), and Tanytarsusdissimilis (midge), which are all potential prey for salmon.Severe effects occurred at water concentrations in the low tosub-ppb range. Similarly, in a 10-day bioassay, Hwang et al.(2001) observed reduced survival, increased developmental

time, and reduced fecundity in Chironomus riparius (midge)fed a diet contaminated with certain PCB congeners. Chin etal. (1998) observed increased energy utilization for growth,molting, and oxygen consumption in mysids exposed to PCBs.Fewer studies exist that examine effects of PAHs on salmonprey species. However, Kemble et al. (2000) observed a sig-nificant negative correlation between the reproduction of theamphipod Hyalella azteca and the concentration of PAHs.

Alterations in benthic invertebrate community structurehave also been observed at sites with high concentrations ofPCBs in sediments. For example, Wright (1988) reported de-clines in chironomid species richness and abundance in Mis-souri streams contaminated with PCBs. Several studies reportdeclines in taxa richness, species diversity, and abundance ofbenthic invertebrates in sediments contaminated with PCBs(Wildhaber and Schmidt 1998; Bishop et al. 2000; Long2000) and PAHs (Poulton et al. 1997; Jewett et al. 1999;Peso-Aguiar et al. 2000). Accordingly, given the direct ef-fects of contaminants on the survival and reproduction ofsalmon prey species, we will examine the model with a neg-ative press perturbation of contaminants on prey.

ParasitesEffects of contaminants on parasite prevalence and inten-

sity are complex. In fish, responses of parasites to contami-nants appear to depend on whether the parasite is internal orexternal and whether the parasite has a direct or indirect lifecycle (D’Amelio and Gerasi 1997; Pukkinen and Valtonen1998). Pulkkinen and Valtonen (1998), for example, studiedparasites of whitefish (Coregonus lavaretus) in lakes con-taminated by pulp and paper mill effluents and found that, incomparison with fish from unpolluted lakes, those exposedto pulp mill effluent showed an increase in the prevalence ofinfection with cestodes and trematodes transmitted throughan intermediate host but a decrease in the prevalence of acrustacean ectoparasite with a direct life cycle. Parasite pop-ulation may be altered through effects on reproduction,transmission, and death (Lafferty and Kuris 1999).

Studies specifically examining effects of PCBs and PAHson parasites are scarce. However, a review by Overstreet(1993) examined effects of petroleum hydrocarbons onpathogens in relationship to the fish host. Overstreet (1993)found that petroleum hydrocarbons might cause either an in-crease or a decrease in a pathogen population. Several stud-ies report an overall decline in parasite species diversity andrichness and the disappearance of particularly sensitive para-site species at polluted sites (Dusek et al. 1998; Broeg et al.1999; Galli et al. 2001). However, at the same time, theremay be increases in prevalence and intensity of infectionwith certain other parasites. For example, increases in theprevalence and intensity of trichodinid infections in the gillsare observed in several fish species following exposure toorganochlorines and heavy metals (Broeg et al. 1999), sew-age treatment works effluent (Yeomans et al. 1997), andcrude oil (Khan 1990). Other studies report increases in cer-tain groups of parasites but declines in others. For example,Valtonen et al. (1997) studied parasite communities in fourstudy lakes in Finland and found reduced numbers ofdigeneans and myxosporeans but increased numbers of acan-thocephalans and monogeneans in a lake contaminated bypulp mill effluent as compared with two less-polluted lakes.

© 2004 NRC Canada

Arkoosh et al. 1169

Changes in the density of intermediate hosts, direct effectson ectoparasites, and impaired immune systems were re-garded as important mechanisms influencing parasite preva-lence and diversity in this system. Given the complexinfluence of contaminants on pathogen populations, we willexamine the model with either a negative or positive pressperturbation of contaminants on pathogens. We will also ex-amine the model with contaminants not having a direct ef-fect on parasites.

The model: interactions of prey, salmon,and parasites

Based on the above information, we propose the relation-ships between the variables in the community (parasites,salmon, and their prey) presented in Figs. 1a and 1b. Eachvariable is self-regulated, since they have a degree of inde-pendence from each other; the self-effects take into accountself-regulation and other trophic chains to which each vari-able is linked. A press to a variable is represented at the pop-ulation level; that is, a positive press can result in either anincrease in reproduction and (or) a decrease in death rate. Anegative press would be the reverse.

The signed digraph was entered into a computer programcalled Powerplay, which compiles the community matrix(available at www.jambrosi.com). The community matrix issimply a tabulation of interaction parameters between vari-ables (Puccia and Levins 1985). The generated communitymatrix was entered into another computer program, MAPLE,which assesses stability criteria (Dambacher et al. 2002). Fora system to be theoretically stable, that is, to be able to re-cover from a disturbance, feedback at all levels should benegative. Also, the feedback at higher levels cannot be greaterthan the feedback at lower levels. A stable community is onethat is resilient to perturbations and will exist through time(Li and Moyle 1981). Typical of a predatory–prey chain, oursystem exhibits strong stability (in its broad sense, namelyof persistence) as assessed by the Routh–Hurwitz criteria(May 1973; Logofet 1993; revised by Dambacher et al.2003b).

Given that the model of the community is stable, we canpredict changes in population levels following a press per-turbation (Fig. 1c). The procedure consists of taking the in-verse of the negative of the community matrix (Bender et al.1984; Dambacher et al. 2002).

Negative press of contaminants on salmon and preyThe contaminants of concern, namely PCBs and PAHs,

negatively affect salmon and prey populations (Fig. 2a). Ouranalysis predicts how this press will affect the community.These effects are shown pictorially as a digraph (Fig. 2b) ofthe inverse (prediction table, Fig. 2c) of the community ma-trix (as in Yodzis 1988). All of the effects of a PCB press atthe parasite and salmon level have a negative effect. Whentwo presses occur in a community, the effects are additive;when the effects are positive and negative, a degree of ambi-guity is introduced. Given a double press, the effect of thecontaminant on the prey level is qualitatively ambiguous be-cause both positive and negative effects affect the same vari-able. This ambiguity suggests that in a salmon communityexposed to contaminants, the most reliable variables to mon-

itor would be either the salmon population or the parasitepopulation. The effect of the contaminant on these popula-tions would be unambiguously negative, unlike the effect onthe prey population.

Negative press of contaminants on salmon, parasites,and prey

Contaminants can also influence parasitic populations, re-sulting in a triple press. However, depending on the parasite,the press may be either positive or negative. If the press isnegative (Fig. 3a), predictions are that all of the effects onthe parasite resulting from contaminants would be negative(Figs. 3b and 3c). However, the effects on the salmon andprey would either be negative or positive. This result makesthe effects on salmon and prey qualitatively ambiguous. Theonly unambiguous variable in the community would be theparasite population. This analysis suggests that, when thecontaminants are known to have a negative effect on the par-

© 2004 NRC Canada

1170 Can. J. Fish. Aquat. Sci. Vol. 61, 2004

Fig. 2. Negative press of contaminants on two variables (salmonand prey): (a) community signed digraph, (b) prediction signeddigraph, and (c) its inverse of the negative of the communitymatrix. The signed digraph represents the final effects of con-taminants negatively affecting salmon and prey in the communitywithout directly affecting the parasite population. If contaminantsnegatively affect prey and salmon in our community, there aretwo populations that can be monitored for unambiguous results.These are salmon and parasites (Figs. 2b and 2c). All of the ef-fects resulting from the press on salmon and parasites are nega-tive. Monitoring prey can potentially give ambiguous resultsbecause that population can either increase or decrease (Figs. 2band 2c). That is, the response of the prey population resultingfrom a press perturbation can either decrease or increase.

asite population as well as on the other variables, parasitesrather than salmon or prey populations would be the mostdesirable variable to monitor the effects of contaminants ona salmon community.

Negative press of contaminants on salmon and prey butpositive press on parasites

If the press of contaminants is positive on the parasitevariable but negative on the other variables (Fig. 4a), all ofthe effects on the salmon population resulting fom contami-nants would be negative. However, the effects on the preyand parasites would be either negative or positive (Figs. 4band 4c). Consequently, the effects on prey and parasites maybe too ambiguous to monitor. The only unambiguous vari-able in the community would be the salmon population. Thissuggests that when the contaminants are known to have apositive effect on the parasite population, the salmon popula-tion and not the prey or parasite populations would be the

most reliable variable to monitor when determining the ef-fects of contaminants on a salmon community.

Discussion

We demonstrate that the effect of press perturbations, asarising from a chemical contaminant, can have counterintuitiveresults on different trophic levels of a community. Further-more, if contaminants were to have compounded effects onthe community, that is, impact on more than one variable,the behavior of the system would be very complex. Taking aparsimonious structure of the system and analyzing it quali-tatively, we identify which variable(s) would yield unambig-uous or less ambiguous results. Monitoring a variable withouttaking community structure into account could possibly leadto misleading results and wasted resources.

The top trophic level, the parasite population in our spe-cific example, should be the target of a monitoring campaignbecause ambiguous predictions suggest the possibility that

© 2004 NRC Canada

Arkoosh et al. 1171

Fig. 3. Negative press of contaminants on three variables (para-sites, salmon, and prey): (a) community signed digraph, (b) pre-diction signed digraph, and (c) its inverse of the negative of thecommunity matrix (c). The signed digraph represents contami-nants negatively affecting salmon, prey, and parasite populationsin the community. If contaminants negatively affect parasites,salmon, and prey in our community, there is only one populationthat can be monitored for unambiguous results (Figs. 3b and 3c).The least ambiguous population to monitor would be the parasitepopulation. Monitoring prey and salmon can potentially give am-biguous results because these populations can either increase ordecrease (Figs. 3b and 3c).

Fig. 4. Negative and positive presses; negative press from contami-nants on salmon and prey but a positive contaminant press on par-asites: (a) community signed digraph, (b) prediction signeddigraph, and (c) its inverse of the negative of the community ma-trix. The signed digraph represents contaminants negatively affect-ing salmon and prey populations in the community but positivelyaffecting parasite populations. If contaminants negatively affectsalmon and prey but positively affect parasites in our community,there is only one population that can be monitored for unambigu-ous results (Figs. 4b and 4c). The least ambiguous population tomonitor would be the salmon population. Monitoring prey andparasites can potentially give ambiguous results because these pop-ulations can either increase or decrease (Figs. 4b and 4c).

no change or no significant change could be observed atlower trophic levels, salmon and their prey, following con-tamination. The cancellation and compensation between pos-itive and negative effects resulting from multiple pathwayslead to ambiguity at the lower levels. Monitoring these vari-ables may provide no signal of contamination or signals thatare ambiguous. Paradoxically, in the case with negative in-put on all of the variables, salmon abundance could increasefollowing contamination at the top trophic level. If this ef-fect is strong enough to overcome the consequences of thenegative impact on salmon itself and its prey, the overall re-sult is an increase in the level of salmon. In this case, moni-toring salmon would lead to a conclusion that thecommunity is healthy. In reality, highly deleterious inputmay be compensated by losses at the top trophic level,which eventually could lead to community destabilization. Arecent study documents the ability of migrating salmon todeliver pollutants into spawning grounds (Krummel et al.2003).

A recent study indicates that trophic levels respond differ-ently to a widely acting press perturbation, namely climatechange (Voigt et al. 2003). The authors document that “sen-sitivity increases significantly with increasing trophic level”.Our theoretical analysis above provides a possible explana-tion for this phenomenon. Any broad environmental stressorthat acts as a press perturbation on all levels of a trophicchain, climate change and toxicants being two examples, af-fects the top trophic level most clearly. This level thus ap-pears to be the most sensitive from a communityperspective.

Community matrix analysis is mostly qualitative becauseof the difficulty of quantifying relationships; indeed, thereappears to be a single totally specified community for whichthe predictions of the inverse matrix were tested (Schmitz1997). A recent qualitative study of a complex aquatic meso-cosm (containing eight variables) correctly predicted the re-sult of an experimental press perturbation (Hulot et al. 2000).A similar analysis of another aquatic community (10 vari-ables) also demonstrated that the long-term behavior of com-plex communities is highly predictable in a natural setting aswell (Dambacher et al. 2002).

Qualitative analysis now provides researchers with a pow-erful hypothesis-stating tool and can prevent serious miscal-culations in experimental design. Specifically, in our system,the most practical variable to monitor, prey, proves to be theleast reliable from a theoretical perspective and the analysismay prevent expenditure of valuable resources into an unre-warding project. The technique theoretically is prone to fail-ure if a particular relationship or input is an order ofmagnitude greater or less than the rest of the community(Bender et al. 1984). A more difficult task is identifying thestructure of the community. An aggregation, such as above,of all prey items or parasites into a single guild variable ulti-mately may prove too simplistic, in which case, more com-plex models can be devised but potentially with moreambiguity. Schmitz and Sokol-Hessner (2002) empiricallydemonstrated that guild level aggregation may lead to in-creased predictability.

Qualitative analysis also provides an avenue for not onlydetermining which population to examine in a community todetermine the effects of contaminants on endangered salmon

populations but also for determining the effects of combinedstressors on the population. Accordingly, Ruckelshaus et al.(2002) suggested that when examining the recovery ofsalmon, it is important to determine simultaneous modifica-tions in the four major identified anthropogenic factors thatinfluence salmon (habitat degradation, hydroelectric dams,harvest practices, and hatchery practices) and natural ecolog-ical factors. Our model could be expanded to incorporatethese four major classes of stressors.

Technical considerations will now direct future researchand monitoring. Given that parasites, pathogens, and othersalmon predators are a candidate for monitoring the impactof toxicants, we must give serious thought to this variable.In particular, how would we best devise indices of parasitelevel in the community? It may be that a controlled field orlaboratory experiment will be required to validate both thetheory and the monitoring technique. Furthermore, the actualdirect impact of toxicants on parasitic organisms and otherpredators clearly requires elucidation. Our model thereforeoffers a justification for laboratory and field testing of toxi-cants on organisms previously not considered in monitoring.

We conclude that a target population of interest, salmon inour example, may not always be the best indicator. Thus,while a toxicant may have a direct negative effect on salmonfitness, the reduction in fitness may not be translated intolower abundance because of countervailing reductions in theirparasites, and so on. An a priori theoretical analysis allowsus to anticipate such behavior and plan for alternative moni-toring methods.

Acknowledgements

We thank the organizers of the US Geological Survey work-shop, under the leadership of Dr. H. W. Li (Oregon Coopera-tive Fish and Wildlife Research Unit, US Geological Survey,Department of Fisheries and Wildlife, Oregon State University,Corvallis, OR 97331, USA), entitled “Application of qualitativemodeling to problems of ecosystem management”. The con-cept for this paper was developed at this workshop, which washeld at the H.J. Andrews Experimental Forest Lab Facilities,Blue River, Oregon, from 17 to 20 October, 2001. We thankDrs. J. Jorgensen (Jambrosi Inc., Corvallis, Oregon), H.W. Liand D. Marcogliese (St. Lawrence Centre, Environment Can-ada, Montreal, Quebec), and J. Meador, C. Green, and J.Spromberg (Environmental Conservation Division, NorthwestFisheries Science Center, Seattle, Washington) for their con-structive comments on the manuscript.

References

Arkoosh, M.R., and Collier, T.K. 2002. Ecological risk assessmentparadigm for salmon. Analyzing immune function to evaluaterisk. Hum. Ecol. Risk Assess. 8: 265–276.

Arkoosh, M.R., Casillas, E., Clemons, E., McCain, B.B., andVaranasi, U. 1991. Suppression of immunological memory in ju-venile chinook salmon (Oncorhynchus tshawytscha) from an ur-ban estuary. Fish Shellfish Immunol. 1: 261–277.

Arkoosh, M.R., Clemons, E., Myers, M., and Casillas, E. 1994.Suppression of B-cell mediated immunity in juvenile chinooksalmon (Oncorhynchus tshawytscha) after exposure to either apolycyclic aromatic hydrocarbon or to polychlorinated biphen-yls. Immunopharmacol. Immunotoxicol. 16: 293–314.

© 2004 NRC Canada

1172 Can. J. Fish. Aquat. Sci. Vol. 61, 2004

Arkoosh, M.R., Casillas, E., Huffman, P., Clemons, E., Evered, J.,Stein, J.E., and Varanasi, U. 1998. Increased susceptibility of ju-venile chinook salmon (Oncorhynchus tshawytscha) from a con-taminated estuary to the pathogen Vibrio anguillarum. Trans.Am. Fish. Soc. 13: 257–268.

Arkoosh, M.R., Clemons, E., Huffman, P., Kagley, A.N., Casillas,E., Adams, N., Sandborn, H.R., Collier, T.K., and Stein, J.E.2001. Increased susceptibility of juvenile chinook salmon tovibriosis after exposure to chlorinated and aromatic compoundsfound in contaminated urban estuaries. J. Aquat. Anim. Health,12: 257–268.

Bender, E.A., Case, T.J., and Gilpin, M.B. 1984. Perturbation ex-periments in community ecology: theory and practice. Ecology,65: 1–13.

Bishop, C.A., Struger, J., Barton, D.R., Shirose, L.J., Dunn, L.,Lang, A.L., and Sheperd, D. 2000. Contamination and wildlifecommunities in stormwater detention ponds in Guelph and theGreater Toronto area, Ontario, 1997 and 1998. Part 1. Wildlifecommunities. Water Qual. Res. J. Can. 35: 399–435.

Broeg, K., Zander, S., Diamant, A., Koerting, W., Kruener, G.,Paperna, I., and Westernhagen, H. 1999. The use of fish meta-bolic, pathological and parasitological indices in pollution moni-toring. 1. North Sea. Helgol. Mar. Res. 53: 171–194.

Casillas, E., Arkoosh, M.R., Clemons, E., Hom, T., Misitano, D.,Collier, T.K., Stein, J.E., and Varanasi, U. 1995a. Chemical con-taminant exposure and physiological effects in outmigrant juve-nile chinook salmon from urban estuaries of Puget Sound,Washington. In Proceedings from the 1994 Northwest PacificChinook and Coho Workshop: Salmon Ecosystem Restoration:Myth and Reality. Edited by M. Keefe. The Oregon Chapter ofthe American Fisheries Society, P.O. Box 722, Corvallis, Oregon.pp. 86–102.

Casillas, E., Arkoosh, M.R., Clemons, E., Hom, T., Misitano, D.,Collier, T.K., Stein, J.E., and Varanasi, U. 1995b. Chemical con-taminant exposure and physiological effects in outmigrant chi-nook salmon from urban estuaries of Puget Sound, Washington.Proceedings Puget Sound Research 95. Puget Sound WaterQuality Authority, P.O. Box 40900, Olympia, Wash. pp. 657–665.

Castillo, G., Li, H.W., and Rossignol, P.A. 2000. Absence of over-all feedback in a benthic estuarine community: a system poten-tially buffered from impacts of biological invasions. Estuaries,23: 275–291.

Chin, P., Shin, Y-K., and Joen, E-M. 1998. Effects of PCBs (poly-chlorinated biphenyls) on energy budget in mysid Neomysisawatschensis. 2. Effects of PCBs on energy budget in mysidNeomysis awatschensis. J. Korean Fish. Soc. 31: 104–108.

Collier, T.K., Johnson, L.L., Myers, M.S., Stehr, C.M., Krahn,M.M., and Stein, J.E. 1998. Fish injury in the Hylebos Water-way of Commencement Bay, Washington. NOAA Tech. Memo.NMFS-NWFSC-36.

Dambacher, J.M., and Rossignol, P.A. 2001. The golden rule ofcomplementary feedback. ACM SIGSAM Bull. 34: 1–9.

Dambacher, J.M., Li, H.W., and Rossignol, P.A. 2002. Relevanceof community structure in assessing indeterminacy of ecologicalpredictions. Ecology, 83: 1372–1385.

Dambacher, J.M., Li, H.W., and Rossignol, P.A. 2003a. Qualitativepredictions in model ecosystems. Ecol. Model. 161: 79–83.

Dambacher, J.M., Luh, H-K., Li, H.W., and Rossignol, P.A. 2003b.Qualitative stability and ambiguity in model ecosystems. Am.Nat. 161: 876–888.

D’Amelio, S., and Gerasi, L. 1997. Evaluation of environmentaldeterioration by analyzing fish parasite biodiversity and commu-nity structure. Parassitologia, 39: 237–241.

Dillon, T.M., Benson, W.H., Stackhouse, R.A., and Crider, A.M.1990. Effects of selected PCB congeners on survival, growth,and reproduction in Daphnia magna. Environ. Toxicol. Chem. 9:1317–1326.

Dusek, L., Gelnar, M., and Sebelova, S. 1998. Biodiversity of para-sites in a freshwater environment with respect to pollution: meta-zoan parasites of chub (Leuciscus cephalus L.) as a model forstatistical evaluation. Int. J. Parasitol. 28: 1555–1571.

Evans, M.S., Noguchi, G.E., and Rice, C.P. 1991. The biomagnifi-cation of polychlorinated biphenyls, toxaphene, and DDT com-pounds in a Lake Michigan offshore food web. Arch. Environ.Contam. Toxicol. 20: 87–93.

Fausch, K.D., Lyons, J., Karr, J.R., and Angermeier, P.L. 1990.Fish communities as indicators of environmental degradation.Am. Fish. Soc. Symp. 8: 123–144.

Fry, D.M., and Fisher, S.W. 1990. Effect of sediment contact and up-take mechanisms on accumulation of three chlorinated hydrocar-bons in the midge, Chironomus. Bull. Environ. Contam. Toxicol.44: 790–797.

Galli, P., Crosa, G., Mariniello, L., Ortis, M., and D’Amelios, S.2001. Water quality as a determinant of the composition of fishparasite communities. Hydrobiologia, 452: 173–179.

Hulot, F.D., Lacroix, G., Lescher-Moutoue, F., and Loreau, M.2000. Functional diversity governs ecosystem response to nutri-ent enrichment. Nature (Lond.), 405: 340–344.

Hwang, H., Fisher, S.W., and Landrum, P.F. 2001. Identifying bodyresidues of HCBP associated with 10-d mortality and partial lifecycle effects in the midge, Chironomus riparius. Aquat. Toxicol.52: 251–267.

Jewett, S.C., Dean, T.A., Smith, R.O., and Blanchard, A. 1999. Ex-xon Valdez oil spill: impacts and recovery in the soft-bottombenthic community in and adjacent to eelgrass beds. Mar. Ecol.Prog. Ser. 185: 59–83.

Kareiva, P., Marvier, M., and McClure, M. 2000. Recovery andmanagement options for spring/summer chinook salmon in theColumbia River Basin. Science (Wash., D.C.), 290: 977–979.

Karr, J.R. 1998. Rivers as sentinels: using the biology of rivers toguide landscape management. In River ecology and manage-ment. Lessons from the pacific coastal ecoregion. Edited by R.J.Naiman, R.E. Bilby, and S. Kantor. Springer-Verlag New York,Inc., New York. pp. 502–508.

Kemble, N.E., Hardesty, D.G., Ingersoll, C.G., Johnson, B.T., Dwyer,F.J., and MacDonald, D.D. 2000. Evaluation of the toxicity of con-taminated sediments from Waukegan Harbor, Illinois, followingremediation. Arch. Environ. Contam. Toxicol. 39: 452–461.

Kennish, J.J. 1992. Ecology of estuaries: anthropogenic effects.CRC Press, Inc., Boca Raton, Fla.

Khan, R.A. 1990. Parasitism in marine fish after chronic exposureto petroleum hydrocarbons in the laboratory and to the ExxonValdez oil spill. Bull. Environ. Contam. Toxicol. 44: 759–763.

Krummel, E.M., Macdonald, R.W., Kimpe, L.E., Gregory-Eaves, I.,Demers, M.J., Smol, J.P., Finney, B., and Blais, J.M. 2003. Deliv-ery of pollutants by spawning salmon. Nature (Lond.), 425: 255–256.

Kucklick, J., Baker, J., Ostrom, N., Ostrom, P., and Lee, D. 1995.Organochlorine trophodynamics in Lake Superior. Proceedingsof the 38th Conference of the International Association ofGreat Lakes Research. International Association for GreatLakes Research, 2000 Bonisteel Blvd., Ann Arbor, Mich.

Lafferty, K.D., and Kuris, A.M. 1999. How environmental stressaffects the impacts of parasites. Limnol. Oceanogr. 44: 925–931.

Lampert, W., Fleckner, W., Pott, E., Schober, U., and Storkey,K.U. 1989. Herbicide effects on planktonic systems of differ-ent complexity. Hydrobiologia, 188/189: 415–424.

© 2004 NRC Canada

Arkoosh et al. 1173

Landsberg, J.H., Blakesley, B.A., Reese, R.O., McRae, G., andForstchen, P.R. 1998. Parasites of fish as indicators of environ-mental stress. Environ. Monit. Assess. 51: 211–232.

Lester, D.C., and McIntosh, A. 1994. Accumulation of poly-chlorinated biphenyl congeners from Lake Champlain sedimentsby Mysis relicta. Environ. Toxicol. Chem. 13: 1825–1841.

Li, H.W., and Moyle, P.B. 1981. Ecological analysis of species intro-duction into aquatic systems. Trans. Am. Fish. Soc. 110: 772–782.

Logofet, D.O. 1993. Matrices and graphs. Stability problems inmathematical ecology. CRC Press, Inc., Boca Raton, Fla.

Long, E.R. 2000. Degraded sediment quality in U.S. estuaries: areview of magnitude and ecological implications. Ecol. Appl.10: 338–349.

MacRury, N.K., and Johnson, B.M. 1999. Sublethal responses oflargemouth bass to parasites and organochlorines. Environ. Toxicol.Chem. 18: 998–1006.

Marcogliese, D.J. 2001. Implications of climate change for parasit-ism of animals in the aquatic environment. Can. J. Zool. 79:1331–1352.

May, R.M. 1973. Qualitative stability in model ecosystems. Ecology,54: 638–641.

May, R.M. 1983. Parasitic infections as regulators of animal popu-lations. Am. Sci. 71: 36–45.

McCain, B.B., Malins, D.C., Krahn, M.M., Brown, D.W., Gronlund,W.D., Moore, L.K., and Chan, S.-L. 1990. Uptake of aromaticand chlorinated hydrocarbons by juvenile chinook salmon(Oncorhynchus tshawytscha) in an urban estuary. Arch. Environ.Contam. Toxicol. 19: 10–16.

Menge, B.A. 1995. Indirect effects in marine rocky intertidal inter-action webs: patterns and importance. Ecol. Monogr. 65: 21–74.

National Marine Fisheries Service. 2000. Northwest salmon recov-ery planning. Recovery planning for west coast salmon. Avail-able online from www.nwfsc.noaa.gov/cbd/trt/overview.htm.

National Research Council. 1996. Upstream salmon and society in thePacific Northwest. National Academy Press, Washington, D.C.

Nebeker, A.V., and Puglisi, F.A. 1974. Effect of polychlorinated bi-phenyls (PCB’s) on survival reproduction of Daphnia, Gammarus,and Tanytarsus. Trans. Am. Fish. Soc. 103: 722–728.

Nehlsen, W., Williams, J.E., and Lichatowich, J.A. 1991. Pacificsalmon at the crossroads: stocks at risk from California, Oregon,Idaho, and Washington. Fisheries, 16: 4–21.

Overstreet, R.M. 1993. Parasitic diseases of fishes and their relation-ship with toxicants and other environmental factors. In Advancesin fisheries science. Pathobiology of marine and estuarine organ-isms. Edited by J.A. Couch and J.W. Fournie. CRC Press, Inc.,Boca Raton, Fla. pp. 111–156.

Peso-Aguiar, M.C., Smith, D.H., Assis, C.F., Santa-Isabel, L.M.,Peixinho, S., Gouveia, E.P., Almeida, T.C.A., Andrade, W.S.,Carqueija, C.R.G., Kelmo, F., Carrozzo, G., Rodrigues, C.V.,Carvalho, G.C., and Jesus, A.C.S. 2000. Effects of petroleumand its derivatives in benthic communities at Baia de Todos osSantos/Todos os Santos Bay, Bahia, Brazil. Aquat. Ecosyst.Health Manag. 3: 459–470.

Poulton, B.C., Finger, S.E., and Humphrey, S.A. 1997. Effects of acrude oil spill on the benthic invertebrate community in the Gas-conade River, Missouri. Arch. Environ. Contam. Toxicol. 33:268–276.

Preston, B.L. 2002. Indirect effects in aquatic ecotoxicology: im-plications for ecological risk assessment. Environ. Manag. 29:311–323.

Puccia, C.J., and Levins, R. 1985. Qualitative modeling of complexsystems. An introduction to loop analysis and time averaging.Harvard University Press, Cambridge, Mass.

Pulkkinen, K., and Valtonen, E.T. 1998. The use of parasites as tags toelucidate differences between whitefish populations. Adv. Limnol.50: 257–271.

Quinn, T.P. 1994. Anthropogenic influences on fish populations ofthe Georgia Basin. Part I. Salmonids. Can. Tech. Rep. Fish.Aquat. Sci. No. 1948. pp. 219–229.

Ruckelshaus, M.H., Levin, P., Johnson, J.B., and Kareiva, P.M.2002. The Pacific salmon wars: what science brings to the chal-lenge of recovering species. Annu. Rev. Ecol. Syst. 33: 665–706.

Schmitz, O.J. 1997. Press perturbations and the predictability ofecological interactions in a food web. Ecology, 78: 55–69.

Schmitz, O.J., and Sokol-Hessner, L. 2002. Linearity in the aggre-gate effects of multiple predators in a food web. Ecol. Lett. 5:168–172.

Stehr, C.M., Brown, D.W., Hom, T., Anulacion, B.R., Reichert, W.L.,and Collier, T.K. 2000. Exposure of juvenile chinook and chumsalmon to chemical contaminants in the Hylebos Waterway ofCommencement Bay, Tacoma Washington. J. Aquat. Ecosyst.Stress Recovery, 7: 215–227.

Stein, J.E., Hom, T., Collier, T.K., Brown, D.W., and Varanasi, U.1995. Contaminant exposure and biochemical effects in outmigrantjuvenile chinook salmon from urban and non-urban estuaries ofPuget Sound, WA. Environ. Toxicol. Chem. 14: 1019–1029.

US Environmental Protection Agency. 1997. Priorities for ecolog-ical protection: an initial list and discussion document for EPA.EPA/600/S-97/002. Office of Research and Development, USEnvironmental Protection Agency, Washington, D.C.

US Environmental Protection Agency. 1998. Guidelines for eco-logical risk assessment. EPA/630/R-95/002F. Office of Researchand Development, US Environmental Protection Agency, Wash-ington, D.C.

Valtonen, E.T., Holmes, J.C., and Koskivaara, M. 1997. Eutro-phication, pollution and fragmentation: effects on the parasitecommunities in roach and perch in four lakes in central Finland.Parassitologia, 39: 233–236.

Voigt, W., Perner, J., Davis, A.J., Eggers, T., Schumacher, J.,Bahrmann, R., Fabian, B., Heinrich, W., Kohler, G., Lichter, D.,Marstaler, R., and Sander, F.W. 2003. Trophic levels are differ-entially sensitive to climate. Ecology, 84: 2444–2453.

Wildhaber, M.L., and Schmitt, C.J. 1998. Indices of benthic commu-nity tolerance in contaminated Great Lakes sediments: relationswith sediment contaminant concentrations, sediment toxicity, andthe sediment quality triad. Environ. Monit. Assess. 49: 23–49.

Wootton, J.T. 2002. Indirect effects in complex ecosystems: recentprogress and future challenges. J. Sea Res. 48: 157–172.

Wright, C.A. 1988. Utility of surface-floating Chironomidae pupalexuviae in assessing the impact of PCBs on two stream commu-nities. Ecol. Res. Ser. EPA 600/9-89/016. US EnvironmentalProtection Agency, Washington, D.C.

Yeomans, W.W., Chubb, J.C., and Sweeting, R.A. 1997. Use ofprotozoan communities for pollution monitoring. Parassitologia,39: 201–212.

Yodzis, P. 1988. The indeterminacy of ecological interactions as per-ceived through perturbation experiments. Ecology, 69: 508–515.

Appendix A. Primer on analysis of complexcommunities (loop analysis)

There are two common methods of representing commu-nities in the ecological literature. The first is as a signed di-graph. Each variable, most commonly a species, isrepresented as a large circle. Density-dependent interactions,or links, are represented as a pointed arrow or circled arrow,each being a positive or negative interaction, respectively.

© 2004 NRC Canada

1174 Can. J. Fish. Aquat. Sci. Vol. 61, 2004

The most common paired interaction is that of predator–prey, which is represented as a positive and negative arrow,as occurs below between N1/N2 and N2/N3. A “self-effect”represents intraspecific interactions, as on N1, such as aris-ing from logistic growth.

The second method is as a community matrix, which is atabulation of the density-dependent interactions, whereineach element of the matrix corresponds to a link in the di-graph. Elements of the matrix are generally subscripted andai,j can be read as “the direct effect to i from j”.

Two main analyses are carried out on a community ma-trix. The first is to assess stability. The procedure is to derivethe characteristic polynomial, the coefficients of which arethe feedbacks of the system. In order for a community to re-cover, all feedbacks must be negative. However, this condi-tion is not sufficient and the system must meet anothercondition that ensures that destabilizing overcorrection willnot occur. The procedure is more complex and nonintuitive.

These so-called Routh–Hurwitz criteria have recently beenrevised (Dambacher et al. 2003b).

Characteristic polynomial =

–λ3 – a1,1λ2 – (a1,2a2,1 + a2,3a3,2)λ1

– (a1,1a2,3a3,2)λ0

where λ are the eigenvalues (measures of recovery). The co-efficients are all negative and therefore the community isstable.

The second analysis consists of “predicting” new levels ofvariable density following a press perturbation. This is donesimply by applying Cramer’s rule, which consists of calcu-lating the inverse of the negative of the community matrix.Thus, the effects on all variables of a press perturbation tovariable N1 is read down the first column:

Further discussion and a description of computer programsare available in Dambacher et al. (2002).

ReferencesDambacher, J.M., Li, H.W., and Rossignol, P.A. 2002. Relevance

of community structure in assessing indeterminacy of ecologicalpredictions. Ecology, 83: 1372–1385.

Dambacher, J.M., Luh, H-K., Li, H.W., and Rossignol, P.A. 2003b.Qualitative stability and ambiguity in model ecosystems. Am.Nat. 161: 876–888.

© 2004 NRC Canada

Arkoosh et al. 1175

Prediction (inverse) matrix, (–A)–1

1a1 1,

0 a

a a1 2

1 1 3 2

,

, ,

0 0 −1

a3 2,

a

a a2 1

1 1 2 3

,

, ,

1

2 3a ,

a a

a a a1 2 2 1

1 1 2 3 3 2

, ,

, , ,