@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix skos: . vivo:departmentOrSchool "Forestry, Faculty of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Chará-Serna, Ana M."@en ; dcterms:issued "2017-12-22T16:31:10Z"@en, "2017"@en ; vivo:relatedDegree "Doctor of Philosophy - PhD"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description "Agriculture is the primary cause of sedimentation, nutrient enrichment, and insecticide contamination of freshwater ecosystems. Despite the widespread co-occurrence of these ecological stressors, little is known about their potential interactive effects. I conducted three experiments manipulating combinations of these stressors in order to evaluate their cumulative effects on freshwater ecosystems at different scales of biological organization (community, ecosystem, meta-ecosystem). First, I evaluated stream invertebrate community responses to sedimentation, nutrient enrichment, and the insecticide chlorpyrifos using laboratory microcosms with distinct microhabitats. I demonstrated that chlorpyrifos can interact non-additively with fine sediment (reversal) and nutrients (antagonism), with potentially deleterious impacts on small-sized invertebrates. Furthermore, invertebrates in gravel microhabitats were more severely affected than those in leaf packs. Second, I manipulated levels of nutrients, sediment, and the insecticide imidacloprid in experimental pond ecosystems. I demonstrated these stressors had antagonistic effects on pelagic and benthic invertebrate diversity. Moreover, the results suggested imidacloprid increased ecosystem metabolism indirectly, through negative effects on invertebrate consumers. Finally, I explored processes at the scale of the river network meta-ecosystem. Using a network of experimental channels, I investigated how multiple-stressor interactions within tributaries affected downstream ecosystems. My results indicated that complex nutrient-sediment interactions within tributaries could strongly alter the flux of organisms from tributaries to downstream ecosystems. Furthermore, I observed that at small spatial scales, these alterations of within-network migration patterns could be more influential than the transport of the stressors from headwaters to recipient ecosystems. My research contributes novel evidence suggesting that complex interactions among nutrient enrichment, sedimentation, and insecticide contamination are frequent in freshwater ecosystems, and have distinct mechanisms operating at different scales. In particular, these findings underscore the importance of considering multiple-stressor interactions in insecticide environmental risk assessments; even at low concentrations, interactions with other stressors may result in unexpected negative effects for aquatic biota and ecosystem processes."@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/64138?expand=metadata"@en ; skos:note "CUMULATIVE EFFECTS OF MULTIPLEAGRICULTURAL STRESSORS ON FRESHWATERECOSYSTEMSbyAna M. Chara´-SernaM.Sc., University of Michigan, 2012A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORALSTUDIES(Forestry)The University of British Columbia(Vancouver)December 2017c© Ana M. Chara´-Serna, 2017AbstractAgriculture is the primary cause of sedimentation, nutrient enrichment, and in-secticide contamination of freshwater ecosystems. Despite the widespread co-occurrence of these ecological stressors, little is known about their potential in-teractive effects. I conducted three experiments manipulating combinations ofthese stressors in order to evaluate their cumulative effects on freshwater ecosys-tems at different scales of biological organization (community, ecosystem, meta-ecosystem). First, I evaluated stream invertebrate community responses to sedi-mentation, nutrient enrichment, and the insecticide chlorpyrifos using laboratorymicrocosms with distinct microhabitats. I demonstrated that chlorpyrifos can in-teract non-additively with fine sediment (reversal) and nutrients (antagonism), withpotentially deleterious impacts on small-sized invertebrates. Furthermore, inverte-brates in gravel microhabitats were more severely affected than those in leaf packs.Second, I manipulated levels of nutrients, sediment, and the insecticide imidaclo-prid in experimental pond ecosystems. I demonstrated these stressors had antag-onistic effects on pelagic and benthic invertebrate diversity. Moreover, the resultssuggested imidacloprid increased ecosystem metabolism indirectly, through neg-ative effects on invertebrate consumers. Finally, I explored processes at the scaleof the river network meta-ecosystem. Using a network of experimental channels,I investigated how multiple-stressor interactions within tributaries affected down-stream ecosystems. My results indicated that complex nutrient-sediment interac-tions within tributaries could strongly alter the flux of organisms from tributariesto downstream ecosystems. Furthermore, I observed that at small spatial scales,these alterations of within-network migration patterns could be more influentialthan the transport of the stressors from headwaters to recipient ecosystems. Myiiresearch contributes novel evidence suggesting that complex interactions amongnutrient enrichment, sedimentation, and insecticide contamination are problem-atic in freshwater ecosystems, and have distinct mechanisms operating at differ-ent scales. In particular, these findings underscore the importance of consideringmultiple-stressor interactions in insecticide environmental risk assessments; evenat low concentrations, interactions with other stressors may result in unexpectednegative effects for aquatic biota and ecosystem processes.iiiLay SummaryAgriculture is global driver of degradation of freshwater ecosystems, affectingthem through discharges of nutrients, sediment, and insecticides. Although thesedisturbances often occur simultaneously, little is known about their potential tointeract synergistically (i.e., intensify each other’s effects), which creates uncer-tainties in the prediction of their combined environmental impacts. To address thisresearch gap, I evaluated individual and combined effects of nutrients, sediment,and insecticides on experimental freshwater ecosystems. I found that insecticidetoxicity could be enhanced or mitigated by nutrient and sediment additions, withthe outcome depending on the characteristics of the system and the insecticide.Moreover, in river networks synergistic interactions within tributaries had surpris-ing impacts on recipient ecosystems downstream. My results suggest that complexinteractions among simultaneous disturbances are frequent and may have unex-pected, negative effects on aquatic life. Consideration of these potential interac-tions is important to protect freshwater ecosystems in agricultural landscapes.ivPrefaceChapter 2: Chlorpyrifos interacts with other agricultural stressors to alter streamcommunities in laboratory microcosmsAuthors: AM Chara´-Serna, JS RichardsonStatus: Accepted 19/09/2017Journal: Ecological Applications (in press)Comments: This study was conceptualized by AMCS and JSR. AMCS conductedthe experiment, collected the data, and performed laboratory work. AMCS ana-lyzed the data and wrote the paper with input from JSR.Chapter 3: Nutrients and sediment modify the impacts of a neonicotinoid in-secticide on experimental pond ecosystemsAuthors: AM Chara´-Serna, L Epele, Christy A. Morrissey, JS RichardsonStatus: In preparation (anticipated submission date: 15/1/2018)Comments: This study was conceptualized by AMCS and JSR. AMCS and JSRprepared the mesocosms and experimental set-up. AMCS and LE conducted theexperiment, collected the data, and performed laboratory work. CAM analyzed in-secticide water samples. AMCS analyzed the data and wrote the paper with inputfrom JSR. LE and CAM provided edits on the manuscript.Chapter 4: Multiple-stressor interactions in headwater streams and their im-pacts to downstream ecosystemsAuthors: AM Chara´-Serna, JS RichardsonStatus: In preparation (anticipated submission date: 27/2/2018)Comments: This study was conceptualized by AMCS and JSR. AMCS conductedvthe experiment, collected the data, and performed laboratory work. AMCS ana-lyzed the data and wrote the paper with input from JSR.viTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Conservation of freshwater ecosystems . . . . . . . . . . . . . . . 11.2 Impacts of agriculture on freshwater ecosystems . . . . . . . . . . 21.3 Cumulative effects . . . . . . . . . . . . . . . . . . . . . . . . . 51.3.1 Conceptual models to interpret cumulative effects . . . . . 61.4 Cumulative effects in freshwater ecology . . . . . . . . . . . . . . 71.5 Predicting ecosystem response to cumulative stressors . . . . . . . 101.6 Thesis objectives and overview . . . . . . . . . . . . . . . . . . . 11vii2 Chlorpyrifos interacts with other agricultural stressors to alter streamcommunities in laboratory microcosms . . . . . . . . . . . . . . . . 132.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.2.1 Experimental design . . . . . . . . . . . . . . . . . . . . 162.2.2 Stressor treatments . . . . . . . . . . . . . . . . . . . . . 172.2.3 Response parameters . . . . . . . . . . . . . . . . . . . . 182.2.4 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . 192.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.3.1 Water quality parameters . . . . . . . . . . . . . . . . . . 212.3.2 Invertebrate community characteristics . . . . . . . . . . 222.3.3 Effects on gravel invertebrates . . . . . . . . . . . . . . . 222.3.4 Effects on leaf pack invertebrates . . . . . . . . . . . . . 232.3.5 Effects on ecosystem processes . . . . . . . . . . . . . . 242.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.4.1 Single stressor effects . . . . . . . . . . . . . . . . . . . 252.4.2 Interactions between chlorpyrifos and other agricultural stres-sors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282.4.3 Implications . . . . . . . . . . . . . . . . . . . . . . . . 303 Nutrients and sediment modify the impacts of a neonicotinoid insec-ticide on experimental pond ecosystems . . . . . . . . . . . . . . . . 403.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433.2.1 Experimental design . . . . . . . . . . . . . . . . . . . . 433.2.2 Stressor treatments . . . . . . . . . . . . . . . . . . . . . 443.2.3 Response variables . . . . . . . . . . . . . . . . . . . . . 453.2.4 Ecosystem function . . . . . . . . . . . . . . . . . . . . . 473.2.5 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . 483.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503.3.1 Water quality and habitat characteristics . . . . . . . . . . 503.3.2 Pelagic community . . . . . . . . . . . . . . . . . . . . . 513.3.3 Benthic community . . . . . . . . . . . . . . . . . . . . . 52viii3.3.4 Biomass distribution among ecological compartments . . 533.3.5 Ecosystem function . . . . . . . . . . . . . . . . . . . . . 533.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543.4.1 Single stressors strongly affected ecosystem function andstructure . . . . . . . . . . . . . . . . . . . . . . . . . . . 543.4.2 Imidacloprid’s indirect effects on ecosystem function . . . 573.4.3 Frequent antagonistic interactions between imidacloprid andother agricultural stressors . . . . . . . . . . . . . . . . . 583.4.4 Implications . . . . . . . . . . . . . . . . . . . . . . . . 594 Multiple-stressor interactions in tributaries alter downstream ecosys-tems in stream mesocosm networks . . . . . . . . . . . . . . . . . . 734.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 764.2.1 Experimental design . . . . . . . . . . . . . . . . . . . . 764.2.2 Response variables . . . . . . . . . . . . . . . . . . . . . 774.2.3 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . 794.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 814.3.1 Stressor effects on first-order channels . . . . . . . . . . . 814.3.2 Stressor effects on second-order channels . . . . . . . . . 824.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.4.1 Stressors altered invertebrate communities and ecosystemfunction in first-order channels . . . . . . . . . . . . . . . 834.4.2 Tributaries influenced downstream ecosystems through dis-persal of sensitive taxa . . . . . . . . . . . . . . . . . . . 854.4.3 Implications . . . . . . . . . . . . . . . . . . . . . . . . 875 Conclusions: synthesis and implications . . . . . . . . . . . . . . . . 985.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 985.2 Single- and multiple-stressor responses across three freshwater meso-cosm manipulations . . . . . . . . . . . . . . . . . . . . . . . . . 995.2.1 Single stressor effects: the disproportionate impacts of sed-imentation . . . . . . . . . . . . . . . . . . . . . . . . . 99ix5.2.2 Cumulative effects: the importance of antagonistic and re-versal interactions . . . . . . . . . . . . . . . . . . . . . 1025.3 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110A Supporting Materials . . . . . . . . . . . . . . . . . . . . . . . . . . 133xList of TablesTable 2.1 ANOVA summary of linear fixed effects models evaluating im-pacts of stressor treatments on ecosystem functioning and in-vertebrate variables recorded on day 15 of the experiment. Sig-nificant effects are indicated in bold (P < 0.05). . . . . . . . . 32Table 3.1 ANOVA summary of linear models evaluating impacts of thestressor treatments on water quality and habitat characteristicsof the freshwater mesocosms. Significant effects (P < 0.05) areindicated in bold. . . . . . . . . . . . . . . . . . . . . . . . . 61Table 3.2 Average imidacloprid concentration measured in all tanks treatedwith the insecticide, and three randomly selected insecticidecontrols (tanks not treated with the insecticide), on days 1 and35 of the experiment. . . . . . . . . . . . . . . . . . . . . . . 62Table 3.3 ANOVA summary of linear models evaluating impacts of thestressor treatments on habitat, benthic invertebrates, zooplank-ton, and ecosystem function throughout the experiment. Signif-icant effects (P < 0.05) are indicated in bold. . . . . . . . . . . 63Table 3.4 Overview of hypotheses and study results . . . . . . . . . . . . 64Table 4.1 ANOVA summary of linear models evaluating impacts of thestressor treatments on response variables of the first-order chan-nels. Significant effects (P < 0.05) are indicated in bold. . . . . 89xiTable 4.2 Parameter estimates of the zero-inflated Poisson regression ex-plaining EPT drift flux from first-order channels on day 22. Themodel includes two components: a count model (Poisson withlog link), and a zero-inflation model (binomial with logit link).Significant effects (P < 0.05) are indicated in bold. . . . . . . . 90Table 4.3 Summary of independent t-test’s to compare response variablesin second-order channels fed by tributaries where the nutrientsand sediment were applied in combination (cumulative; tribu-tary treatment b in Figure 4.2) and second-order channels fedby tributaries where the stressors were applied separately (sin-gle; tributary treatment c in Figure 4.2). Significant effects(P < 0.05) are indicated in bold. . . . . . . . . . . . . . . . . 91Table 4.4 ANOVA summary of linear models evaluating the impact ofdisturbance level within the tributaries on response variables ofsecond-order channels. Significant effects (P < 0.05) are indi-cated in bold. . . . . . . . . . . . . . . . . . . . . . . . . . . . 92Table A.1 ANOVA summary of linear mixed effects models to test impactsof stressor treatments (fixed effects) and time (week: fixed ef-fect) on dissolved oxygen, temperature, pH, and conductivity,with individual microcosm treated as a random effect. Symbolsare used to represent significance of the effects according to theANOVAs: *** P < 0.001, **P < 0.01, *P < 0.05, ·P < 0.1. . . 134xiiList of FiguresFigure 2.1 Photo of the experimental set-up consisting in 32 laboratorymicrocosms located at the University of British Columbia, Van-couver, Canada (a). Detail of the microcosms (b). . . . . . . . 33Figure 2.2 Standardized effect size (Hedge’s d ± 95% CI) for significantmain and interactive effects of the stressors on gravel inver-tebrate communities. Letters in the y-axis represent main ef-fects (N, nutrients; S, sediment; I, insecticide) and interactionterms (N*S, nutrient x sediment; N*I, nutrient x insecticide;S*I, sediment x insecticide; N*S*I; nutrient x sediment x in-secticide). For main effects, significant (i.e. not overlappingzero) positive values indicate increases in the response variablewhereas negative values denote the opposite. For interactions,confidence intervals overlapping zero indicate additive inter-actions, positive values indicate synergies, and negative val-ues indicate antagonistic or reversal interactions (reversals aremarked with an R). Symbols are used to represent significanceof the effects according to the ANOVAs: *P < 0.05, ·P < 0.1. 34xiiiFigure 2.3 Bar plots illustrating significant interactive effects of the stres-sors on gravel (a-b) and leaf (c-e) invertebrate metrics, accord-ing to the ANOVAs (P = 0.05). All shown interactions wereclassified as antagonistic, except for panel (b) which was clas-sified as a reversal. Letter notation for the treatments is consis-tent with Figure 2.2. Bars represent the mean of four replicates(± SE). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Figure 2.4 Distance-based redundancy analysis on log-transformed abun-dance data of invertebrates collected in gravel. Only the mostabundant taxa are labeled. Letter notation for the treatmentsis consistent with Figure 2.2. Solid arrows indicate significantmain effects (P < 0.05) and dashed arrows indicate marginallysignificant effects (P = 0.059), according to a permutation testof the full model with 999 randomizations. The treatmentstogether explained 29% of the total variation on communitycomposition of gravel. Axes 1 and 2 represent 10 and 5% ofthe total variation, respectively. . . . . . . . . . . . . . . . . . 36Figure 2.5 Standardized effect size (Hedge’s d ±95% CI) for significantmain and interactive effects of the stressors on invertebratecommunities collected in leaf packs. Letter notation for thetreatments and interpretation of effects is consistent with Fig-ure 2.2. Symbols are used to represent significance of the ef-fects according to the ANOVAs: *P < 0.05, ·P < 0.1. . . . . 37Figure 2.6 Distance-based redundancy analysis on log-transformed abun-dance data of invertebrates collected in leaf packs. Only themost abundant taxa are labeled. Letter notation for the treat-ments is consistent with Figure 2.2. Solid arrows indicate sig-nificant main effects (P < 0.05) according to a permutationtest of the full model with 999 randomizations. The treatmentstogether explained 27% of the total variation on communitycomposition of leaf packs. Axes 1 and 2 represent 7 and 6% ofthe total variation, respectively. . . . . . . . . . . . . . . . . . 38xivFigure 2.7 Standardized effect size (Hedge’s d ±95% CI) for significantmain and interactive effects of the stressors on ecosystem pro-cesses. Letter notation for the treatments and interpretation ofeffects is consistent with Figure 2.2. Symbols are used to rep-resent significance of the effects according to the ANOVAs:**P < 0.01, *P < 0.05. . . . . . . . . . . . . . . . . . . . . 39Figure 3.1 Photo of the experimental set-up consisting in 32 outdoor fresh-water mesocosms located at the University of British Columbia’sexperimental pond facility, Vancouver, Canada. . . . . . . . . 65Figure 3.2 Temporal dynamics of phytoplankton biomass (a, b), zooplank-ton biomass (c, d), and net ecosystem production (e, f) in fresh-water mesocosms exposed to nutrient, sediment, and insecti-cide pulses. Each point represents the mean (± SE) for eachtreatment (n = 4). Letters are used to represent stressor treat-ments (N, nutrients; S, sediment; I, insecticide, N+S, nutrient+ sediment; N+I, nutrient + insecticide; S+I, sediment + in-secticide; N+S+I; nutrient + sediment + insecticide). Dashedvertical lines indicate the dates for nutrient and insecticide ad-ditions, while dotted vertical lines indicate sediment additions.Points are slightly jittered to ease interpretability . . . . . . . 66xvFigure 3.3 Standardized effect size (Hedge’s d ± 95% CI) for significantstressor main and interactive effects on zooplankton commu-nities. Letters in the y-axis represent main effects (N, nutri-ents; S, sediment; I, insecticide) and interaction terms (N*S,nutrient x sediment; N*I, nutrient x insecticide; S*I, sedimentx insecticide; N*S*I; nutrient x sediment x insecticide). Formain effects positive values indicate increases in the responsevariable whereas negative values indicate the opposite. For in-teractions, confidence intervals overlapping zero indicate addi-tive interactions, positive values indicate synergies, and nega-tive values are antagonisms or reversals (reversals are markedwith an R). Symbols are used to represent significance of theeffect according to the linear mixed effects models presentedin Table 3.3: ***P < 0.001, **P < 0.01, *P < 0.05, ·P < 0.1. 67Figure 3.4 Bar plots illustrating significant interactive effects of the stres-sors on zooplankton (a-c) and benthic invertebrate (d-e) met-rics. Letter notation for the treatments is consistent with Fig-ure 3.2. Dotted horizontal lines are used to represent the pre-dicted additive effect of the stressors. All shown interactionswere classified as antagonistic, except the S*I interaction onabundance of large-sized benthic invertebrates (d) which wasa reversal. Bars represent the mean of four replicates (± SE). 68Figure 3.5 Redundancy analysis plot showing the effects of the stressorpulses on zooplankton (a) and benthic invertebrate (b) commu-nity composition by the end of the experiment. Only the mostabundant taxa are labeled. Solid arrows indicate significantmain effects (P < 0.05) and dashed arrows indicate marginallysignificant effects (P < 0.1), following a permutation test ofthe full model with 999 randomizations. Letter notation forthe treatments is consistent with Figure 3.3. The treatments to-gether explained 35% of the variation on zooplankton compo-sition and 33% of variation on benthic invertebrate composition. 69xviFigure 3.6 Standardized effect size (Hedge’s d ± 95% CI) for significantstressor main and interactive effects on benthic invertebratecommunities. Letter notation for the treatments and interpre-tation of effects is consistent with Figure 3.3. Symbols areused to represent significance of the effect according to thelinear fixed effects models presented in Table 3.3: **P < 0.01,*P < 0.05, ·P < 0.1. . . . . . . . . . . . . . . . . . . . . . . 70Figure 3.7 Redundancy analysis plot showing the effects of the stressorpulses on the distribution of biomass among food web com-ponents in the mesocosms by the end of the experiment (day36). Solid arrows indicate significant main effects (P < 0.01),following a permutation test of the full model with 999 ran-domizations. Letter notation for the treatments is consistentwith Figure 3.3. The treatments together explained 27% of thevariation on biomass (a). Barplot comparing average biomass(n = 16) of food web components among nutrient treatmentlevels (b). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71Figure 3.8 Standardized effect size (Hedge’s d ± 95% CI) for significantstressor main and interactive effects on ecosystem function.Letter notation for the treatments and interpretation of effectsis consistent with Figure 3.3. Symbols are used to representsignificance of the effect according to the linear mixed effectsmodels presented in Table 3.3: ***P < 0.001, **P < 0.01,*P < 0.05, ·P < 0.1. . . . . . . . . . . . . . . . . . . . . . . 72Figure 4.1 Photo of the experimental channel-network set-up consistingin 24 first-order channels converging in pairs to feed 12 second-order channels (a). Detail of a pair of first-order channels andtheir second-order recipient (b). . . . . . . . . . . . . . . . . 93xviiFigure 4.2 Schematic representation of our experimental channel-networkset up. Each pair of first-order channels (represented as graybars) is a treatment for its second-order receptor (black bars).Initials are used to represent stressor additions in first-ordertreatments (N, nutrients; S, sediment; N+S, nutrient + sed-iment). Letters are used to represent second-order tributarytreatments (a, two control tributaries; b, one control tributaryand one tributary with both stressors; c, nutrients added in onetributary and sediment in the other; and d, nutrients and sedi-ment added in both tributaries). There were three replicates ofeach tributary treatment for a total of 12 second-order channelsand 24 first-order channels. . . . . . . . . . . . . . . . . . . . 94Figure 4.3 Standardized effect size (Hedge’s d ± 95% CI) for significantmain and interactive effects of the stressors on first-order chan-nels. Letters are used to represent main effects (N, nutrients;S, sediment) and interactions (N*S, nutrients x sediment). Formain effects positive values indicate increases in the responsevariable, whereas negative values indicate the opposite. Forinteractions confidence intervals overlapping zero indicate ad-ditive interactions, positive values denote synergies, and neg-ative values indicate antagonisms. Symbols are used to repre-sent significance according to the linear models presented inTable 4.1: ***P < 0.001, **P < 0.01, *P < 0.05, ·P < 0.1. . . 95Figure 4.4 Bar plot illustrating significant antagonistic effects of the stres-sors on leaf decomposition (as AFDM loss from 3 g leaf packs).Bars represent the mean of the treatments (± SE): control (n =9), sediment (S, n = 3), nutrients (N, n = 3), nutrients + sed-iment (N+S, n = 9). A dotted line represents the predictedadditive effect of the two stressors. Letters indicate significantdifferences among the treatments according to Tukey’s Hon-estly Significant Difference tests. . . . . . . . . . . . . . . . . 96xviiiFigure 4.5 Plots representing the impact of the stressor treatments on thefrequency of zeros (a) and log-transformed counts (b) of in-dividuals from the orders Ephemeroptera, Plecoptera, and Tri-choptera drifting out of the first-order channels on day 22. Let-ter notation for the treatments is consistent with Figure 4.4. . 96Figure 4.6 Bar plots illustrating the effect of tributary level of disturbanceon total sediment deposition (a), organic sediment deposition(b), and EPT density (c) in recipient second-order channels.Treatments in the x-axis represent: no disturbance in the trib-utaries (0, tributary treatment a in Figure 4.2), one tributarydisturbed (1, tributary treatment b in Figure 4.2), two tribu-taries disturbed (2, tributary treatment d in Figure 4.2). Barsrepresent the mean of each tributary treatment (±SE, n = 4).Letters indicate significant differences among the treatmentsaccording to Tukey’s Honestly Significant Difference tests. . . 97Figure 5.1 Frequency distribution of individual stressor effects (a), andmultiple-stressor interactions (b) on freshwater ecosystem re-sponses evaluated across the three experiments conducted forthis dissertation. Letters are used to represent individual stres-sor treatments (N, nutrients; S, sediment; I, insecticide) andinteractions (N*S, nutrient x sediment; N*I, nutrient x insecti-cide; S*I, sediment x insecticide; N*S*I; nutrient x sediment xinsecticide). The number of response variables analyzed foreach stressor (a) or stressor combination (b) is indicated inparentheses. . . . . . . . . . . . . . . . . . . . . . . . . . . . 108Figure 5.2 Distribution of significant multiple-stressor interactions involv-ing insecticides across the experiments. Letters are used to de-note the treatments: nutrients (N), sediment (S), imidacloprid(IMI), and chlorpyrifos (CPY). . . . . . . . . . . . . . . . . 109xixFigure A.1 Schematic of the stream microcosm set-up. Darker rectanglein the middle of the tank represents a horizontal glass plate in-troduced to partially divide the tank into an upper and a lowersection. A bubbler along with a plastic deflector placed in oneend of the tank promoted circular flow in the direction indi-cated by the blue arrows. Redrawn from Sanpera-Calbet et al.(2012). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135Figure A.2 Untransformed univariate responses of gravel invertebrates acrossthe treatments on day 15. Boxes are drawn around upper andlower quartiles with whiskers indicating maximum and mini-mum values, dark thick lines indicate the median, and pointsdenote outliers. Y-axis does not always start in zero. . . . . . 136Figure A.3 Untransformed univariate responses of leaf invertebrates acrossthe treatments on day 15. Boxes are drawn around upper andlower quartiles with whiskers indicating maximum and mini-mum values, dark thick lines indicate the median, and pointsdenote outliers. Y-axis does not always start in zero. . . . . . 137Figure A.4 Untransformed ecosystem processes across the treatments onday 15. Boxes are drawn around upper and lower quartileswith whiskers indicating maximum and minimum values, darkthick lines indicate the median, and points denote outliers. Y-axis does not always start in zero. . . . . . . . . . . . . . . . 138xxAcknowledgmentsI would like to express my gratitude to Dr. John Richardson, for being a wonder-ful supervisor and mentor. I am very thankful for all the time and resources hehas invested in me and my project, and for the confidence he placed in me fromthe beginning. I would also like to thank the members of my PhD committee, Dr.Sarah Gergel, Dr. Evgeny Pakhomov, and Dr. Marwan Hassan, for their supportand encouragement througouth this process. I gratefully acknowledge COLCIEN-CIAS (Beca Caldas), the Izaak Walton Killam Memorial Fund for Advanced Stud-ies (Killam Doctoral Scholarship), and the University of British Columbia (4 yearfellowship) for funding my doctoral studies.I am very grateful to the members of the Stream and Riparian Research lab fortheir invaluable emotional, intellectual, and logistic support throughout this PhD.Talking and doing science with this brilliant group of researchers has been one ofthe best learning opportunities I have had in my scientific career. In particular, Iwould like to thank Sean Naman, Liz Perkin, Alex Yeung, Liliana Garcı´a, BrianKielstra, Luis Epele, and Matt Wilson, for their help with data collection, analysis,and presentation at different stages of this research project.I would also like to thank my tribe of friends, who became my family in Van-couver. They have been a source of inspiration and emotional support throughoutthese years. Thank you for making my time in UBC so beautiful and memorable.Last but not least, I want to thank my family, Mom, Dad, Luis, Julia´n, and myhusband Santiago, for their love and support have made this journey possible fromthe very beginning. Thank you.xxiDedicationTo my lovely, brilliant husband Santiago Santacruz. From the times he pulled all-nighters with me to help with scholarship applications back in Colombia, to thesepast months of excruciating writing schedules –when he diligently kept me wellfed, hydrated, and sane– his unconditional support has been fundamental in everystep of this crazy journey.Thank you for being the best ropemate I could ever ask for.xxiiChapter 1Introduction1.1 Conservation of freshwater ecosystemsFreshwater ecosystems are the most threatened ecosystems worldwide (Saunderset al., 2002; Dudgeon et al., 2006; Abell et al., 2007; Nel et al., 2009; Strayer andDudgeon, 2010; Vo¨ro¨smarty et al., 2010). Due to the accelerated growth of humanpopulation and economy over the past century, the use of fresh waters has increasedrapidly, causing massive negative impacts on these ecosystems (Strayer and Dud-geon, 2010). Threats to fresh waters have been extensively documented, with habi-tat degradation and loss, water pollution, flow modification, land use change, over-exploitation, and introduction of non-native species, figuring as the most importantdrivers of the declining trends (Dudgeon et al., 2006; Allan and Castillo, 2007;Dyke, 2008). As of 2010, approximately 65% of the continental water dischargewas moderately to highly threatened by human activities, with 30 of the world’s47 largest rivers showing moderate to high incidence of threats (Vo¨ro¨smarty et al.,2010). These levels of degradation have resulted in alarming declines in freshwaterbiodiversity (Dudgeon et al., 2006; Strayer and Dudgeon, 2010), including the lossor imperilment of over 10,000 species (Strayer and Dudgeon, 2010). Moreover, thedegradation of fresh waters has direct impacts on human populations, who dependon water as their most essential natural resource. According to a recent global spa-1tial analysis, 80% of the world’s population lives in areas with low water security,which means that about 4.8 billion people do not have reliable access to potablewater (Vo¨ro¨smarty et al., 2010)Freshwater ecosystems have a high interface with the landscape, which makesthem extremely vulnerable to land-use alterations (Allan and Johnson, 1997; Al-lan, 2004). Anthropogenic land use affects these ecosystems through different pro-cesses that operate at multiple spatial scales (Allan and Johnson, 1997). Theseprocesses include alterations of habitat structure and organic matter inputs, whichare primarily governed by local conditions, and alterations to nutrient supply, sed-iment inputs, hydrology, and geomorphology, that are mainly influenced by condi-tions at the regional scale (Allan and Johnson, 1997). Furthermore, it has been longrecognized that human disturbances on fresh waters rarely occur in isolation, andthat certain land uses, such as agriculture and urbanization, tend to cause multiple,simultaneous impacts on these ecosystems (Allan, 2004; Townsend et al., 2008).However, to date we still lack a predictive understanding of the mechanisms bywhich these impacts alter ecological function and structure of fresh waters.1.2 Impacts of agriculture on freshwater ecosystemsAgriculture is the most important modifier of water-related ecosystem servicesaround the globe (MEA, 2005; Vo¨ro¨smarty et al., 2010; Smukler et al., 2012).Currently more than 15% of the global precipitation runs through cultivated land-scapes, and up to 70% of the global water withdrawals are used for agricultural ac-tivities (Smukler et al., 2012). As a result, agriculture is the main non-point sourceof excess nutrients, fine sediment, and toxic and organic pollutants to aquaticecosystems worldwide (Allan, 2004; MEA, 2005; Vo¨ro¨smarty et al., 2010). Fur-thermore, agricultural practices often result in degradation or loss of riparian forests,which in turn alters organic matter inputs, hydrology, flow regimes, and chan-nel morphology of natural water bodies (Allan, 2004; Johnson and Host, 2010).All these physical and chemical modifications ultimately impair habitat qualityand alter resource availability for biological communities, causing shifts in theirtrophic structure and composition (Allan, 2004; Diana et al., 2006; Johnson and2Host, 2010; Riseng et al., 2011).The nutrient pollution caused by the use of fertilizers in agricultural landscapeshas long been recognized as a global threat to freshwater and marine ecosystems(Vo¨ro¨smarty et al., 2010; Woodward et al., 2012b; Rosemond et al., 2015). Nutri-ent enrichment has been linked to alterations in allochthonous and autochthonousbasal resources of freshwater food webs (Woodward et al., 2012b; Rosemond et al.,2015). At low levels, nutrient inputs stimulate algae growth and accelerate leaf de-composition due to increased microbial processing (Rosemond et al., 2015; Garcı´aet al., 2017). Therefore, even though moderate nutrient inputs cause bottom-upeffects for the autochthonous component of freshwater food webs, they are alsoassociated with increased mineralization of organic carbon. This in turns resultsin release of CO2 and lower energy transfers to higher trophic levels for the al-lochthonous component of such food webs (Rosemond et al., 2015; Atwood et al.,2015). Furthermore, high levels of nutrient enrichment can be toxic to aquaticorganisms and are associated with hazardous algal blooms, and dissolved oxygendepletion in running waters and hypoxia in lentic and coastal ecosystems (Allan,2004; Riseng et al., 2011).Agricultural practices are also associated with increased fine sediment runoffdue to the replacement of the natural vegetation cover and livestock trampling(Allan, 2004). The term fine sediment is used henceforth to describe sedimentwith particles of less than 2 mm in size (Wood and Armitage, 1997), thus includ-ing very coarse sand (1000− 2000 µm), coarse sand (500− 1000 µm), mediumsand (250− 500 µm), fine sand (125− 250 µm), very fine sand (62− 125 µm),silt (4− 62 µm), and clay (< 4 µm), according to the Wentworth scale (Went-worth, 1922). Increased loads of fine sediment are documented to have partic-ularly harmful impacts on freshwater ecosystems (Waters, 1995; Wood and Ar-mitage, 1997; Allan, 2004). Several recent experimental manipulations have ob-served that compared to other agricultural impacts, such as nutrient enrichment,warming, and glyphosate contamination, sedimentation had the most detrimentaleffects on invertebrate communities and ecological function (Piggott et al., 2012;Magbanua et al., 2013; Piggott et al., 2015b). The entrance of fine sediment intofreshwaters increases water column turbidity, limiting light penetration and reduc-3ing primary production, with consequent negative impacts on higher trophic levels(Wood and Armitage, 1997; Wagenhoff et al., 2013). Once deposited, fine sedimentcan greatly impair habitat quality for aquatic organisms, smothering periphytonand biofilm, and clogging interstitial spaces that are refugia for invertebrates andgravel-spawning fish (Allan, 2004; Matthaei et al., 2006; Louhi et al., 2011). Fur-thermore, fine sediment is reported to induce invertebrate drift, and cover gills andrespiratory surfaces with deleterious effects on sensitive organisms (Allan, 2004;Piggott et al., 2015b).The widespread use of pesticides in agricultural catchments often results incontamination of freshwater ecosystems, usually through spray drift or runoff (Al-lan, 2004; Sa´nchez-Bayo, 2011; Morrissey et al., 2015). Even though it has beenlong recognized that pesticides may play a big role in ecological degradation,the concentration of agricultural pesticides is rarely measured in studies relat-ing agricultural land use and freshwater biota (Allan, 2004; Stehle and Schulz,2015; Scha¨fer et al., 2016; Gessner and Tlili, 2016). However, recent global es-timations indicate that pesticides are among some of the world’s predominantsources of freshwater pollution (Vo¨ro¨smarty et al., 2010; Scha¨fer et al., 2016), withcontinental- and country-scale surveys in Europe, North America, and Australiareporting frequent detection of pesticides at concentrations that exceed regulatoryrisk thresholds (Gilliom, 2007; Malaj et al., 2014; Stehle and Schulz, 2015; Mor-rissey et al., 2015; Scha¨fer et al., 2016; Sa´nchez-Bayo et al., 2016). The impactsof pesticides on freshwater ecosystems are diverse and largely depend on the modeof action of their active ingredients, their target organisms, and the magnitude andduration of exposures (Sa´nchez-Bayo, 2011). However, some recent large-scalesurveys indicate that in general, pesticide contamination in freshwater ecosystemsis associated with significant losses of fish and invertebrate diversity, as well asalterations in important ecosystem processes (Scha¨fer et al., 2012; Beketov et al.,2013; Malaj et al., 2014). Moreover, some of these studies registered such effects atconcentrations that are considered safe by European Union regulations, highlight-ing the urgent need to review current methodologies for pesticide risk assessment(Scha¨fer et al., 2012; Beketov et al., 2013).41.3 Cumulative effectsAnthropogenic stressors are physical, chemical, and biological factors that, as aresult of human activities, exceed their natural ranges of variation, causing mea-surable biological and ecological responses (Townsend et al., 2008; Crain et al.,2008; Statzner and Beˆche, 2010). The changes an ecosystem experiences due to thecombination of multiple stressors are herein denominated cumulative effects (Mac-donald, 2000; Crain et al., 2008; Seitz et al., 2011). Researchers have long beeninterested in cumulative effects because they are often difficult to predict on the ba-sis of single stressor responses. Increasing evidence highlights the importance ofcomplex interactions, in which the combination of two or more stressors results innegative effects that are amplified (synergistic interaction), mitigated (antagonisticinteraction), or reversed (reversal interaction), relative to what is expected fromthe stressors’ individual effects (Folt et al., 1999; Townsend et al., 2008; Matthaeiet al., 2010; Piggott et al., 2015c). For example, in their highly cited meta-analysisof cumulative effects on marine ecosystems, Crain et al. (2008) found that the over-all interaction effect across all studies was synergistic. Similarly, Holmstrup et al.(2010) found that synergistic interactions among natural stressors and toxicantswere common in 150 animal-focused studies. Moreover, a recent meta-analysis ofexperimental studies in freshwater ecosystems reported that 41% of the multiple-stressor interactions were antagonistic, while 28% were synergistic (Jackson et al.,2016).Synergistic interactions generally take place when the biological response toone or more of the stressors involved is non-linear (e.g. a certain impact has adisproportionate effect on the community), or when a combination of activitiestriggers secondary effects that enhance their impact on the ecosystem (Macdon-ald, 2000; Crain et al., 2008; Seitz et al., 2011). On the other hand, antagonisticinteractions are observed when the influence of one stressor dampens the impactof a second stressor (Folt et al., 1999; Crain et al., 2008). The importance of un-derstanding and predicting these phenomena was highlighted almost two decadesago (Breitburg et al., 1998; Folt et al., 1999), and is still considered one of themost pressing issues in ecology and conservation (Wagenhoff et al., 2013; Piggott5et al., 2015b; Jackson et al., 2016). Distinguishing categories of multiple-stressorinteractions is fundamental to improve our understanding of their mechanisms, ul-timately improving our ability to predict cumulative effects (Folt et al., 1999; Crainet al., 2008). Furthermore, disentangling multiple-stressor effects allows us to clas-sify stressors based on severity, providing us with strategies to better manage andmitigate their impacts on natural ecosystems (Townsend et al., 2008; Statzner andBeˆche, 2010).1.3.1 Conceptual models to interpret cumulative effectsEven though the terms “synergistic” and “antagonistic” have been used extensivelyin the scientific literature, their mathematical definitions vary according to differentconceptual frameworks. Folt et al. (1999) highlighted this lack of consensus anddescribed three broad conceptual models to categorize multiple stressor effects: thecomparative effects model, the additive model, and the multiplicative model.Comparative model: This model is appropriate when the combined effectof two stressors is equal to the effect of the single worst stressor (Folt et al., 1999).This model is useful in systems where a single stressor is the main driver of degra-dation. Once this dominant stressor affects a system, other lesser stressors have noadditional impact. In this model, combined effects that are greater or less than theeffect of the dominant stressor are synergistic or antagonistic, respectively (Foltet al., 1999).Additive model: This is the most commonly used model in theoretical andapplied research on multiple stressors (Folt et al., 1999; Crain et al., 2008; Dunne,2009; Piggott et al., 2015c). In this framework, the combined effect of multiplestressors is assumed to equal the sum of the individual stressor effects (Reid, 1993;Crain et al., 2008). Deviations that result in effects that are greater or less than thesum of the individual effects are considered synergistic and antagonistic, respec-tively. Despite its widespread use, some authors argue that this model cannot detectsynergistic interactions when the sum of the individual effects exceeds 100% (Pen-nings, 1996; Folt et al., 1999), which limits its usefulness to evaluate interactions6when individual stressor effects are high.Multiplicative model: This model assumes that the combined effects of mul-tiple stressors are equal to the product of the individual effects. Thus, antagonisticand synergistic effects are less or more than that expected from the multiplicativeinteractions of the stressors (Folt et al., 1999; Townsend et al., 2008).Because these three models mathematically result in different interpretations ofinteractions between cumulative stressors, Folt et al. (1999) emphasized the im-portance of specifying which model is used when conducting cumulative effectsresearch. Furthermore, after carrying out an empirical test comparing the applica-bility of the three models, they concluded that no single best model can be recom-mended for cumulative effects research. Instead, each researcher should choose theproper framework depending on the stress mechanisms of the system under study,and explicitly state why it was chosen.1.4 Cumulative effects in freshwater ecologyThe issue of cumulative effects in freshwater ecosystems has received increasingattention in the past few decades. Initially, with the recognition of the multi-stressortemplate in which freshwater ecosystems function, a number of studies evaluatedthe response of freshwater communities and habitats to multiple stressors acting si-multaneously (e.g., Swank and Bolstad, 1994; Greathouse et al., 2005; Alexanderand Culp, 2008; Dodds and Oakes, 2008; Aristi et al., 2012). Following this effort,freshwater research took a step forward to measure the relative effect of individ-ual stressors, as a means to distinguish the most important drivers of degradation(Comte et al., 2010; Riseng et al., 2010; Esselman and Allan, 2010; Wepener et al.,2011; Dama´sio et al., 2011; Wooster et al., 2012). Along with these studies, cumu-lative effects’ implications on management issues have received a lot of attention,with papers proposing methodologies to evaluate cumulative effects (e.g., Chen,1992; Bevenger and King, 1995; Smit and Spanling, 1995; Loftis et al., 2001; John-son et al., 2012), and others discussing their incorporation into management andenvironmental policies (e.g. Macdonald, 2000; Cooper and Sheate, 2002; Noble,72010; Seitz et al., 2011; Noble et al., 2011; Sheelanere et al., 2013).Even though the implications of cumulative effects are already being consid-ered in management frameworks, there is still much to learn empirically aboutcomplex interactions of some widespread ecological stressors (No˜ges et al., 2016).Only in the past couple of decades have researchers begun implementing fieldand experimental studies with the specific purpose of measuring the nature of theinteractions among different anthropogenic stressors (e.g., Wagner et al., 1997;Townsend et al., 2008; Kratina et al., 2012; Clements et al., 2013; Piggott et al.,2015a; Alexander et al., 2016). In a recent meta-analysis, Jackson et al. (2016)identified 88 papers examining multiple-stressor interactions on freshwater ecosys-tem receptors. These studies span the years 1995 to 2014, and have addressed dif-ferent stressor combinations (e.g., pH, UV radiation, warming, invasive species,ammonia, nutrients, fine sediment, heavy metals, water abstraction, predators, andpesticides), at different scales of biological organization, ranging from unicellu-lar organisms to whole ecosystems (Jackson et al., 2016). These investigationshave not only documented previously unknown cause-and-effect relationships be-tween stressors and biological responses, but have also underscored the prevalenceof complex multiple-stressor interactions in freshwater ecosystems. According toJackson et al. (2016), 84% of the interactions analyzed in those studies were non-additive, with antagonistic interactions representing 41%, followed by synergistic(28%), and reversal interactions (15%). These findings are particularly relevantto environmental management, as they suggest that additive environmental riskassessment frameworks may underestimate or overestimate cumulative effects ofcommon stressors on freshwater ecosystems.Despite these important advances, some stressors and ecosystems have beennot been sufficiently studied under the cumulative effects perspective. For instance,only 7 out of the 88 studies analyzed by Jackson et al. (2016) addressed interactiveeffects of pesticides. Understanding how the presence of additional stressors mayaffect the impacts of pesticides is a critical research need, especially after a recentinvestigation found that environmental stressors could increase individual toxicityto pesticides by a factor up to 100 (Liess et al., 2016). This issue is particularly im-portant for wetlands and shallow ponds in agricultural landscapes (Roessink et al.,82008; Main et al., 2014). These lentic environments, provide important ecosystemservices to agricultural production, and are habitat to waterbirds, amphibians, andinvertebrates (Main et al., 2014; Dodds and Whiles, 2010). Yet, they are frequentlyexposed to pesticide contamination, eutrophication, and sedimentation (Main et al.,2014; Skagen et al., 2008). Moreover, compared to other freshwater ecosystems,their hydraulic characteristics favour sediment deposition and long residence times(Luo et al., 1997; Skagen et al., 2008; Dodds and Whiles, 2010). Both conditionscan potentially increase the exposure of organisms to toxicants through longer per-sistence of the chemical in the water column, contact with contaminated with bedsediment, and resuspension of sediment particles (Warren et al., 2003; Roessinket al., 2008; Burton and Johnston, 2010). Research is needed to understand if thecombination of these factors ultimately enhances the detrimental impacts of pesti-cides on these important ecosystems.Studying cumulative effects on freshwater ecosystems becomes increasinglychallenging at larger spatial scales (Loftis et al., 2001). For example, at the scaleof river networks cumulative watershed effects (i.e. changes that involve water-shed processes and are influenced by multiple stressors) are important (Reid, 1993;Freeman et al., 2007). Yet, there are few empirical studies addressing the subjectof cumulative effects from the river network perspective (but see Tomscha et al.,2017). Thus, there few evaluations of how disturbances in multiple headwatersystems are transmitted down the river network, how complex interactions amongstressors in headwater ecosystems affect ecosystems downstream, or how the lossof headwaters affects river network function and resiliency (Freeman et al., 2007).Research to answer these type of questions is challenging as it may involve spa-tial and temporal lags in system responses, dilution of impacts, geographic decou-pling between cause and effect, and site-specific variations in impact expression(Reid, 1993, 1998; Gomi et al., 2002). However, it is fundamental to understandmechanisms of disturbance at the scale of river networks and formulate adequatestrategies of prevention and mitigation.91.5 Predicting ecosystem response to cumulativestressorsPredicting cumulative effects is extremely challenging due to the high complexityof the systems involved. The response of natural systems to multiple stressors isdetermined by characteristics of both the ecosystem and the stressors at play (Bre-itburg et al., 1998; Vinebrooke et al., 2004). Some ecosystem characteristics thatmodulate response to multiple stressors include: species diversity (especially thenumber of redundant species); openness of the system (increases ability to recoverfrom disturbance); temporal variability (i.e. successional processes); spatial pat-terns; and ecological and evolutionary history (Breitburg et al., 1998; Crain et al.,2008; Statzner and Beˆche, 2010). On the other hand, stressors’ specificity, theirmodes of action, their potential for interaction, the magnitude of their impacts, andtemporal patterns of occurrence (frequency or duration of the exposition, simulta-neous versus consecutive occurrence of stressors), are some stressor characteristicsthat will influence the outcome of the disturbance (Breitburg et al., 1998; Crainet al., 2008).Certain circumstances will result in cumulative effects that are easier to predictthan others. For example, it is reasonable to expect an inverse relationship betweenthe magnitude of the response to multiple stressors and the number of redundantspecies in the system (Breitburg et al., 1998). Diverse systems with many func-tionally similar species may be more stable due to differences in species’ toleranceto stressors that promote complementary responses, i.e. tolerant species increaseas sensitive species decline (Breitburg et al., 1998). In this case, an increase inthe number of stressors with independent modes of action, means that fewer tol-erant species will be able to survive and benefit from the disturbance (Breitburget al., 1998; Statzner and Beˆche, 2010). However, the relationship between themagnitude of the response and the number of stressors will change in communitieswith strongly dominant or keystone species. In the latter case, response to multiplestressors may be similar to the response to a single stressor, if that single stressoraffects the dominant or keystone species (Breitburg et al., 1998). Another casewhen the system may respond similarly to a single stressor as to multiple stressors,10is when the stressors have similar modes of action. Once the system responds toone stressor (e.g., by losing species sensitive to that stressor) no further losses willbe observed if the second stressor targets the same type of organisms, through thesame mechanisms (Breitburg et al., 1998).On the other hand, the relationship between functional redundancy and stabil-ity of processes at the ecosystem level is much less clear when the modes of actionof the different stressors are not independent. In the common case when the ac-tion of one stressor affects mechanistically or statistically the action of a secondstressor, is when we observe “ecological surprises” with unexpected antagonisticor synergistic effects (Breitburg et al., 1998). The cases illustrated above portrayonly a dimension of the complexity involved in the prediction of multiple effects.However challenging, enough empirical information would eventually allow forreasonable predictions of when and where certain interactions are expected to oc-cur, improving our ability to manage their impacts.1.6 Thesis objectives and overviewSedimentation, nutrient enrichment, and insecticide pollution are some of the mostpervasive freshwater ecosystem stressors associated with agricultural activities world-wide (MEA, 2005; Vo¨ro¨smarty et al., 2010). However, despite their obvious co-occurrence and the potential for complex interactions among them, there is rela-tively little empirical information about their pairwise combinations and virtuallyno information about their potential three-way interactions. The overarching objec-tive of my thesis was to improve our mechanistic understanding of the individualand cumulative impacts of these three stressors on freshwater ecosystems. Thus,I experimentally manipulated levels of sediment, nutrients, and insecticide con-tamination on mesocosms recreating different types of freshwater ecosystems, andtested hypotheses about impacts operating at different scales of biological organi-zation. My experiments included: a stream microcosm manipulation to test effectsat the scale of communities (Chapter 2), a pond mesocosm manipulation to testeffects at the scale of ecosystems (Chapter 3), and a stream mesocosm networkmanipulation to test effects at the scale of river meta-ecosystems (Chapter 4).11In Chapter 2, I evaluated in detail how benthic invertebrate communities re-sponded to low concentrations of chlorpyrifos when applied in combination withsedimentation and nutrient enrichment. Chlorpyrifos is a widely used organophos-phorus insecticide that is generally considered safe for non-target organisms. How-ever, to date its effects have not been evaluated in the context of other commonagricultural stressors. Using a fully-crossed factorial experiment in laboratory mi-crocosms, I tested the hypothesis that chlorpyrifos would interact non-additivelywith nutrient and sediment additions in two distinct microhabitats (leaf packs andgravel), and evaluated short-term indirect effects of the insecticide on two impor-tant ecosystem processes (leaf decomposition and primary production).In Chapter 3, I further explored potential non-additive interactions among agri-cultural stressors by testing cumulative impacts of imidacloprid, nutrient enrich-ment, and sedimentation on shallow pond ecosystems. Imidacloprid, a systemicinsecticide from the family of neonicotinoids, currently ranks amongst the mostwidely used agricultural insecticides in the world. Despite its popularity, relativelylittle is known about its potential to interact with other agricultural stressors and al-ter freshwater ecosystem functioning. I used a fully crossed factorial manipulationin outdoor pond mesocosms to test the hypothesis that imidacloprid would interactnon-additively with nutrient and fine sediment additions to alter benthic and plank-tonic communities. Further, I tested whether structural impacts on pelagic and ben-thic compartments of pond ecosystems would ultimately translate into ecosystem-wide alterations on metabolism and organic matter processing.In my final data chapter (Chapter 4), I used a network of outdoor stream meso-cosm channels to investigate potential consequences of multiple-stressor interac-tions at the scale of river networks. I manipulated sediment and nutrient levels inthe tributaries of second-order channels, to determine individual and combined ef-fects of the stressors at the tributaries and recipient channels. Specifically, I testedthe hypothesis that complex-multiple stressor interactions within tributaries wouldinfluence responses (benthic invertebrate density, invertebrate drift, and leaf de-composition) in downstream channels, and that increasing levels of disturbance inthe tributaries would cause proportional increases of disturbance on downstreammesocosms.12Chapter 2Chlorpyrifos interacts with otheragricultural stressors to alterstream communities inlaboratory microcosms2.1 IntroductionOver the past decades, agriculture has become one of the most important drivers offreshwater ecosystem degradation around the world (MEA, 2005; Smukler et al.,2012). Agricultural lands currently cover about 40% of the ice-free land surface(Ramankutty et al., 2010), receive more than 15% of the global precipitation, andare responsible for more than 70% of the water withdrawals (Smukler et al., 2012).The impacts of agriculture on freshwaters are context dependent and involve multi-ple ecological stressors interacting in space and time (Allan, 2004; Matthaei et al.,2010; Riseng et al., 2011; Chara´-Serna et al., 2015). The effects of some of theseecological stressors have been extensively documented. For example, it is well es-tablished that many agricultural practices increase inputs of fine sediment into wa-13ter bodies, generally causing negative effects on invertebrate communities and eco-logical processes (Wood and Armitage, 1997; Allan, 2004; Benoy et al., 2012; Bur-don et al., 2013). It is also widely acknowledged that the application of fertilizersincreases the input of nutrients into freshwater food webs, causing subsidy-stressresponses in biological communities and ecological function (Woodward et al.,2012a; Rosemond et al., 2015; Richardson and Wipfli, 2016). However, agricul-tural stressors rarely operate in isolation, and their cumulative effects are less wellunderstood. In the past decades, the potential complex interactions among agricul-tural stressors have received increasing attention, with several experimental studiesevaluating 2-way and 3-way combinations of stressors like fine sediment, nutri-ent enrichment, water abstraction, warming and glyphosate (e.g., Matthaei et al.,2010; Wagenhoff et al., 2012; Magbanua et al., 2013). Nevertheless, some perva-sive agricultural stressors, like insecticide contamination, have received relativelyless attention in the context of other anthropogenic stressors.The organophosphorus insecticide chlorpyrifos perfectly illustrates this situa-tion. Chlorpyrifos is one of the most widely applied insecticides in the world (Ge-bremariam, 2011; Giesy and Solomon, 2014). In the United States alone, about 3.2to 4.1 million kilograms of chlorpyrifos were applied per year in the last decade,for the control of insect pests and mites (Solomon et al., 2014). Not surprisingly,chlorpyrifos is frequently detected in agricultural water bodies around the world inconcentrations that sometimes exceed aquatic-life-protection criteria (Phillips andBode, 2004; Marino and Ronco, 2005; Williams et al., 2014). As a broad spectrumneurotoxic insecticide, chlorpyrifos is highly toxic to non-target aquatic inverte-brates and, to a lesser extent, to vertebrates (Giesy et al., 1999). Hence, a num-ber of experimental and field studies have been conducted to evaluate its effectson aquatic organisms and ecosystem processes (e.g., Brock et al., 1992b; Puseyet al., 1994; Daam et al., 2008). However, relatively few empirical studies haveaddressed potential complex interactions of the insecticide with other agriculturalstressors (but see Cuppen et al., 2002; Traas et al., 2004; Alexander et al., 2013),and to our knowledge no studies have explicitly tested three-way interactions be-tween chlorpyrifos, nutrient enrichment, and sedimentation.There are several plausible mechanisms by which the negative impacts of chlor-14pyrifos could be enhanced (synergistic interaction) or mitigated (antagonistic in-teraction) by sedimentation and nutrient enrichment. For instance, chlorpyrifosis moderately hydrophobic, so it tends to partition from the aqueous phase to bestrongly adsorbed by the sediment when it enters water bodies (Giesy et al., 1999;Gebremariam, 2011). This tendency to concentrate in the sediment may result inenhanced exposure of invertebrates in streams that are simultaneously affected bysedimentation, through contact and ingestion of contaminated sediment particles.Furthermore, a previous evaluation of interactions among an insecticide mixturethat contained chlorpyrifos, and different levels of nutrient enrichment, observedthat moderate levels of eutrophication mitigated the negative effects of the insec-ticide mixture, whereas high levels of eutrophication enhanced them (Alexanderet al., 2013).Here, we describe the results of a community-level, microcosm experimentdesigned to evaluate individual and cumulative effects of chlorpyrifos, sedimen-tation, and nutrient enrichment on stream invertebrate communities (abundance,biomass, richness, size structure, composition) and ecosystem processes (primaryproductivity and organic matter decomposition), at the scale of the microhabi-tat. We were particularly interested in testing potential non-additive interactionsamong the three agricultural stressors, and evaluating short-term effects of the in-secticide on ecosystem processes. Based on findings of previous investigations,we predicted that at the stressor levels tested in our study, we would observe: 1)positive individual effects of nutrients on periphyton biomass, leaf decomposition,and invertebrate biomass; 2) negative individual effects of sedimentation on mostinvertebrate community metrics and ecosystem processes; 3) negative individualeffects of chlorpyrifos on invertebrate abundance, richness, and biomass; 4) indi-rect negative effects of chlorpyrifos on leaf decomposition (through the inhibitionof shredders); 5) indirect positive effects of chlorpyrifos on periphyton biomass(through the inhibition of grazers); 6) synergistic interactions between chlorpyri-fos and sedimentation enhancing negative impacts on invertebrate communities;and 7) antagonistic interactions between chlorpyrifos and nutrient enrichment mit-igating negative effects on invertebrate communities.152.2 Methods2.2.1 Experimental designWe conducted a 15-day factorial manipulation of sedimentation, nutrient enrich-ment, and chlorpyrifos contamination in 32 laboratory microcosms, using two lev-els of each stressor (presence, absence) in a fully crossed factorial design with eighttreatments and four replicates of each treatment (Figure 2.1). Each microcosm con-sisted of a 38 L aquarium (25.4 x 50.8 x 30.5 cm) filled with 12 L of stream water.We modified the design used by Sanpera-Calbet et al. (2012), placing a 25 x 30 cmglass sheet parallel to the tank bottom to partially divide the tank into an upper anda lower section. We then installed a bubbler connected to an air pump at the bottomof one end of the tank, with a plastic deflector above to generate visible circularflow in the tanks (see Appendix Figure A.1 for a schematic). The lower sectionof each tank was stocked with 2 L of washed mixed gravels (0.5-2 cm grain sizerange), arranged for a mean substrate depth of 1.5 cm. Stream water was collectedfrom Spring Creek, a relatively non-perturbed, third-order stream located in theMalcolm Knapp Research Forest (British Columbia, Canada; 49◦ 16’ N, 122◦ 34’W), that also served as a source for aquatic invertebrates. This stream is charac-terized by high dissolved oxygen concentrations (near saturation), low suspendedsolids (0.4-2.2 mg L-1), and neutral to slightly acidic water (pH: 6.37-6.73). Themicrocosms were situated in a laboratory at the University of British Columbia un-der similar light (ViaVolt T5 high output fluorescent grow lights) and temperature(water temperature 20 ± 0.15◦C) conditions, and had similar values for dissolvedoxygen (9.0 ± 0.04 mg O2 L-1), pH (7.4 ± 0.02), and conductivity (0.02 ± 0.01mS cm-1) before the application of the treatments.Microcosms were inoculated with invertebrate densities 15% higher than den-sities found in Spring Creek, in order to offset for mortality due to transport tothe laboratory and manipulation. For every three microcosms, five Surber samples(Surber area = 0.09 m, mesh size = 500 µm) were collected, pooled, and subsam-pled into three equal portions. This procedure was repeated 11 times to obtain 32replicate samples that were randomly assigned to microcosms. Once inoculated16with aquatic invertebrates, and after deploying leaf packs and ceramic tiles (see be-low), the microcosms were allowed to equilibrate for one day before the applicationof treatments. Laboratory microcosms inoculated with natural invertebrate com-munities cannot incorporate all the complex dynamics of natural ecosystems butthey provide a useful tool to study short-term community responses to controlledmanipulations of multiple stressors. This approach offers sufficient replication forhypothesis testing and has been successfully employed in the past in a number ofstudies on invertebrate ecology and toxicology (e.g., Kiffney and Clements, 1996b;Clements et al., 2013).2.2.2 Stressor treatmentsStressor treatments were randomly assigned to the microcosms and applied once onday 1 of the experiment. For the sediment treatment, 0.8 L of sand (0.25 mm meangrain size, “fine sand” according to the Wentworth scale) were added as evenlyas possible to each sediment addition microcosm, resulting in 60 ± 2% (mean ±Standard Error [SE], visual estimation) of sediment coverage by streambed areaand 8.5 ± 2.1 mm (mean ± SE, measured with a ruler) of sediment deposited ontop of the original gravel substrate. These sediment additions are equivalent tovalues reported in rivers affected by farming practices (Townsend et al., 2008; Wa-genhoff et al., 2012), and are similar to those used in several previous experimentalassessments of the effects of sedimentation in stream ecosystems (e.g., Matthaeiet al., 2010; Piggott et al., 2015a; Louhi et al., 2017).Background nutrient concentrations (NH4-N: 19.8 ± 0.6 µg L-1, NO3-N: 60± 11.4 µg L-1, PO4-P: 2.5 ± 0.1 µg L-1, mean ± SE) were augmented in nutrientaddition tanks using potassium phosphate (KH2PO4: 83 µg L-1) and ammoniumnitrate (NH4NO3: 155 µg L-1). Analytical grade chlorpyrifos (>100%, PestanalrSigma-Aldrich) was used for the insecticide treatment. A stock solution of the in-secticide was prepared in 99.5% analytical grade ethanol and applied as a singlepulse to each treated tank using a micropipette for a nominal chlorpyrifos concen-tration of 0.3 µg L-1. Target chlorpyrifos concentration falls within the range ofchlorpyrifos concentrations that have been frequently reported in surface waters of17agricultural landscapes in countries like the United States (Kimbrough and Litke,1996; Williams et al., 2014) and Argentina (Jergentz et al., 2005). A clean mi-cropipette was used to apply ethanol to all non-insecticide treatment tanks to con-trol for potential solvent effects. The volume of ethanol applied in non-insecticidetanks was the same as the volume added with the insecticide application (400 µL).2.2.3 Response parametersWater samples collected 36 hours after treatment additions were used to determinechlorpyrifos concentrations using standard GC/MS methods (Price et al., 2009),subsequent water sampling three days after treatment application was conducted todetermine suspended solids, PO4-P, NH4-N, and NO3-N concentrations (APHA,2005; Hauer and Lamberti, 2007). Measures of dissolved oxygen concentration,conductivity, and pH were recorded weekly after the application of treatments us-ing hand-held probes.All the gravel from each microcosm was sampled for invertebrates on day 15.We washed the gravel in a 2 mm sieve stacked on top of a 250 µm sieve. Gravelretained in the 2 mm sieve was sorted for invertebrates at the time of collection, andthe material retained in the 250 µm sieve was stored in 80% ethanol and sorted laterunder the dissecting microscope. Half-decomposed invertebrates showing clearindications of being dead before the day of sample collection were discarded, re-maining individuals were enumerated for each microhabitat (leaf packs and gravel),identified to the lowest practical taxonomic level (usually genus), and measuredto determine dry mass from length-mass regressions (Smock, 1980; Benke et al.,1999; Johnston and Cunjak, 1999). Given the characteristics of our indoor micro-cosms and the short duration of the experiment, we can assume the experimentalinvertebrate communities were closed, with no reproduction, immigration or em-igration processes affecting invertebrate density (e.g. we did not observe insectemergence during the experiment). Thus, the reductions in density documentedby the end of the experiment likely reflect invertebrate mortality due to the treat-ments.18One day before treatment application (day 0), a coarse-mesh alder leaf pack(Alnus rubra, 1 g air-dry weight) was introduced to each microcosm to measureleaf decomposition. Leaves used for this study were collected after abscission inthe Malcolm Knapp Research Forest. On day 15, leaf packs were removed from themicrocosms, and stored at -18◦C until processing. Posterior processing involveddefrosting and rinsing over a 250 µm sieve that was sorted for invertebrates. Theremaining leaf material was dried at 60◦C for 5 days, weighed, ashed at 500◦C,and reweighed to calculate ash-free dry mass.Unglazed, 7 cm2 ceramic tiles were used to measure periphyton biomass inthe microcosms. Prior the beginning of the experiment, the tiles were incubatedfor 10 days in Spring Creek to promote algae colonization. Conditioned tiles wererefrigerated and transported to the laboratory, where one tile was introduced intoeach microcosm on day 0. On day 15, tiles were removed from the microcosmsand stored in the darkness at -18◦C for later processing. Periphyton biomass wasestimated as Chlorophyll-a using standard fluorometric methods (Arar and Collins,1997).2.2.4 Data analysisWe computed a total of 20 response variables, including two measures of ecosys-tem processes (leaf decomposition and periphyton biomass), and nine measuresof community structure that were calculated separately for invertebrates collectedin gravel and invertebrates collected in leaf packs (total invertebrate density, to-tal invertebrate biomass, total invertebrate richness, abundance of Ephemeroptera,Plecoptera, and Trichoptera [EPT] taxa, biomass of EPT taxa, richness of EPTtaxa, average body size, abundance of small [< 5 mm] individuals, and abundanceof large [> 5 mm] individuals). We used three-way, factorial ANOVAs to evaluatethe individual and combined effects of nutrient enrichment (N), sedimentation (S),and chlorpyrifos insecticide (I) on these univariate responses. For each responsevariable the linear model tested was: y = b0 +b1N +b2S+b3I +b4N∗S+b5N∗I +b6S∗I+b7N∗S∗I, where N is the nutrient treatment, S the sediment treatment, and Ithe insecticide treatment. Because two invertebrate samples were accidentally lost19during collection and processing, we used type-III sum of squares, which is robustto unbalanced designs (Quinn and Keough, 2002).Significance levels for all tests was P < 0.05. However, following the rec-ommendation of Nakagawa and Cuthill (2007) we present standardized effect sizeestimates for all findings with P ¡ 0.1, so readers can judge the biological impor-tance of the results. Hedges d estimates of effect size (Gurevitch and Hedges, 2006;Nakagawa and Cuthill, 2007) were calculated from the t values of our linear modelsusing the equations provided by Nakagawa and Cuthill (2007). In order to improvethe graphical representation of our results, we assigned the sign of significant maineffects size estimates to represent the direction of the response of manipulated ver-sus control microcosms (i.e. positive effect sizes indicate increases in the responsevariable, while negative effect sizes indicate the opposite). Further, we assignedthe signs of 2-way and 3-way interaction effect size estimates to represent the clas-sification of the interactions according to the framework proposed by Jackson et al.(2016). Thus, in our graphs positive interaction effect sizes represent synergisticinteractions (i.e. the combined effect of the stressors is greater than the sum of theirindividual effects), whereas negative effect sizes represent either antagonistic (i.e.the combined effect of the stressors is less than the sum of their individual effects)or reversal interactions (i.e. the combined effect of the stressors is in the oppositedirection than the sum of the individual effects).We used distance-based redundancy analysis (db-RDA) to evaluate the effectof the stressor treatments on multivariate taxa composition of benthic invertebratescollected in gravel and leaf packs. Db-RDA is based on redundancy analysis(RDA), a common form of direct gradient ordination that allows to test the asso-ciation between multivariate data and individual terms in a factorial experimentaldesign. Db-RDA is specially suited for community data because it runs the RDAfrom a matrix of dissimilarities that can be built from ecologically relevant indicesof species composition association (Legendre and Anderson, 1999; McArdle andAnderson, 2001). In our case, we used the Bray-Curtis dissimilarity index and log-transformed invertebrate abundance data prior analysis. The significance of eachtreatment and interaction was evaluated with a Monte Carlo permutation test of thefull linear model with 999 randomizations. Water quality measures recorded over20the two weeks of the experiment were analyzed with linear mixed effects models(LME). The week of the measurement and the different stressor treatments weretreated as fixed effects, while each individual microcosm was treated as a randomeffect. When necessary, square-root or fourth-root transformations were applied toimprove normality of positively skewed distributions of count variables, and log-transformations were used to improve normality in other types of variables (Quinnand Keough, 2002). All analyses were performed in R v. 3.3.0 (R Core Team,2016), using packages lme4 (Bates et al., 2015), car (Fox and Weisberg, 2011),and vegan (Oksanen et al., 2016).2.3 Results2.3.1 Water quality parametersIn agreement with our expectations, water samples collected three days after treat-ment application showed that nutrient additions augmented average nitrate (NO3-N: 90.83 ± 25.7 µg L-1, mean nutrient treatment; ANOVA nutrient effect: F1,15= 5.89, P = 0.029) and phosphate (PO4-P: 3.1 ± 0.7 µg L-1, mean nutrient treat-ment ± SE) concentrations. However, the change in phosphate was not statisti-cally significant at the time of collection (ANOVA nutrient effect: F1,15 = 5.89, P =0.677). Additionally, we observed an unexpected negative effect of the insecticideapplication on nitrate concentrations (ANOVA insecticide effect: F1,15 = 73.44,P < 0.0001). The sediment treatment caused a six-fold increase in total suspendedsolids concentration that was still detectable 3 days after the application (ANOVA,F1,24 = 8.11, P = 0.01) but was negligible by the end of the experiment (ANOVA,F1,24 = 0.0003, P = 0.98). A technical failure in the chlorpyrifos determination re-sulted in recoveries below 1% for the chemical analysis. Consequently, we cannotreport actual concentration values, but we note that despite low recoveries, chlor-pyrifos was detected in all microcosms treated with the insecticide, except for thosewhere sediment was simultaneously applied (sediment x insecticide treatment).Temperature, pH, conductivity, and dissolved oxygen changed over the course21of the experiment, but only conductivity and dissolved oxygen were significantlyaffected by the treatments (Appendix Table A.1). Conductivity was higher in sed-iment (LME: sediment effect, F1,24 = 245.5, P < 0.001) and nutrient microcosms(LME: nutrient effect, F1,24 = 5.3, P = 0.031); and we observed a significant inter-action between chlorpyrifos and nutrients for dissolved oxygen (LME: insecticidex nutrients effect, F1,24 = 10.98, P = 0.003). This interaction was classified asantagonistic; while both stressors tended to increase dissolved oxygen indepen-dently, their combination had no effect, resulting in average oxygen concentrationsthat were 1.2% lower than expected if the effect of the stressors was additive.2.3.2 Invertebrate community characteristicsInvertebrate communities in control microcosms were diverse, averaging 26.6 (±11.2,SE) taxa and 126 (± 19.5, SE) individuals per microcosm by the end of the experi-ment. On average, 84% of the individuals were collected in gravel (0.04 individualsper cm3 of gravel, mean, n = 4), with the remaining 16% collected in leaf packs(15 individuals per gram of leaf litter, mean, n = 4) in control microcosms. In leafpacks of the control treatments, the Chironomidae subfamilies Orthocladiinae andChironominae were the most abundant taxa, comprising 44% and 21% of the leafinvertebrate abundance, respectively. Ecclisomyia and Capnia were also an impor-tant component of the leaf pack community with 13% and 5% of the abundance,respectively. The most abundant taxa in the gravel of the control treatments wereChironominae (36%), Ecclisomyia (13%), Heterlimnius (11%), and Orthocladiinae(10%).2.3.3 Effects on gravel invertebratesBoxplots for all the biological variables evaluated in this study are presented in Ap-pendix Figures A.2-A.4. We detected strong negative effects of sedimentation onthe abundance and diversity of gravel invertebrate communities (Table 2.1, Figure2.2). Microcosms treated with sediment had on average 26% fewer individuals and17% less taxa than microcosms without sediment. According to our analyses these22significant effects were mostly due to negative impacts on sensitive EPT taxa (46%reduction of EPT abundance, 21% reduction of EPT richness), and large individu-als (42% reduction in abundance). Further, we detected a significant antagonisticinteraction between nutrients and insecticide affecting gravel invertebrate richness.Nutrient x insecticide microcosms had approximately 1.6 times more taxa than ex-pected if the cumulative effect of two stressors was additive (Figure 2.3a). Wealso detected a significant interaction between sedimentation and insecticide forthe abundance of small-sized invertebrates. This interaction was classified as areversal; while the two stressors independently had weak positive effects on num-bers of small-sized individuals, their cumulative effects on this metric were neg-ative. In fact, the abundance of small individuals in sediment x insecticide tankswas approximately 2.4 times lower than expected if the two stressors were additive(Figure 2.3b).According to the db-RDA, the treatments together explained 29% of the to-tal variation on community composition of gravel invertebrates (Figure 2.4). Notsurprisingly, sedimentation explained 8% of the variation, and was the treatmentwith the most significant effects on composition of gravel invertebrates (db-RDA,sediment effect: F1,22 = 2.0, P= 0.007); causing substantial reductions in the abun-dance of Orthocladiinae (65% reduction), Ecclisomyia (55% reduction), Paralep-tophlebia (48 reduction), and Serratella (42% reduction). Nutrient additions alsoaffected the composition of gravel invertebrates, explaining 6% of the variationin composition, but the effect was only marginally significant (db-RDA, nutrienteffect: F1,22 = 1.5,P = 0.071). The nutrient treatment was negatively associatedwith the abundance of Ecclisomyia (48% reduction). Insecticide contaminationexplained 4% of the variation in taxa composition but was not deemed significantaccording to the Monte Carlo permutation test.2.3.4 Effects on leaf pack invertebratesWe observed less severe effects of the stressors on invertebrate communities col-lected in leaf packs (Table 2.1, Figure 2.5). According to our linear models, theabundance of small-sized invertebrates was the only variable showing significant23effects of the stressors, with a significant sediment x insecticide interaction term.This interaction was classified as antagonistic, because the combined effects ofthe two stressors were 3 times less negative than predicted by additivity (Figure2.3c). In fact, when applied as single stressors, the sediment and the insecticidetreatments caused 48% and 70% reductions in the number of small individualscollected in leaf packs, respectively.The distance-based redundancy analysis (db-RDA) detected significant inter-active impacts of the treatments on the composition of leaf pack invertebrates(Figure 2.6). The treatments together explained 27% of the total variation incommunity composition in leaf packs, with the sediment x insecticide interac-tion explaining 7.3% of the total variation (db-RDA, sediment x insecticide effect:F1,22 = 1.6,P = 0.027), and the sediment x nutrient x insecticide interaction ex-plaining 7.2% of the variation in the composition (db-RDA, nutrient x sediment xinsecticide effect: F1,22 = 1.6,P = 0.04). A closer examination of Chironominae,an abundant subfamily strongly associated with the sediment x insecticide interac-tion term, suggested an antagonistic interaction. Sediment x insecticide tanks hadon average 6.3 times higher abundance of Chironominae than predicted by addi-tivity (Figure 2.3d). Similarly, the subfamily Orthocladiinae, which was stronglyassociated with the nutrient x sediment x insecticide term, showed a significant an-tagonistic interaction among the three stressors, as tanks with the three combinedstressors had approximately the same abundance of Orthocladiinae as control tanks(Figure 2.3e).2.3.5 Effects on ecosystem processesChlorpyrifos had strong negative effects on leaf decomposition, causing an aver-age 21% reduction in the amount of mass lost from the leaf packs by the end ofthe experiment (Figure 2.7, Table 2.1). Sediment additions also depressed leafdecomposition in the microcosms, reducing mass loss by 12%. In addition, thesediment treatment caused strong negative effects on periphyton; there was 61%less periphyton biomass in microcosms treated with sediment. Contrary to ourexpectations, we did not detect significant effects of nutrient enrichment on leaf24decomposition or periphyton biomass in our experiment.2.4 DiscussionOur results support the hypothesis that low concentrations of chlorpyrifos have thepotential to alter freshwater ecosystem processes and interact with environmentallyrelevant levels of sedimentation and nutrient enrichment. Our observations are con-sistent with a growing body of research highlighting the importance of consideringmultiple-stressor interactions, and indirect effects on ecosystem processes, whenevaluating the impacts of organic toxicants on freshwater ecosystems (Alexanderet al., 2016; Gessner and Tlili, 2016; Scha¨fer et al., 2016).2.4.1 Single stressor effectsBecause the N:P ratio of Spring Creek (the stream we used as source of water, al-gae, and invertebrates) suggested P-limitation of the system, we predicted in ourfirst hypothesis that a moderate pulse of phosphorus and nitrogen would have sig-nificant positive effects on functional and structural variables (Woodward et al.,2012b; Rosemond et al., 2015). Contrary to our expectations, our nutrient treat-ment was not sufficient to cause a response that would be detectable by the end ofthe experiment. However, some of our results suggest the enrichment may havehad a short-term effect on the experimental systems. For example, despite thenutrient addition, there was no evidence of higher concentration of inorganic phos-phorus in nutrient tanks three days after the application, which might indicate thatphosphorus was quickly taken up. Consistent with this observation, nutrient en-richment tanks showed a 16% increase in periphyton biomass, but the main effectwas not significant due to large variation in periphyton response in enriched tanks.These observations may suggest that the phosphorous-limited periphyton commu-nities quickly assimilated the extra phosphorous but the subsequent increase inbiomass was rapidly transferred to consumers. Modest nutrient pulses have beenpreviously documented to have little effects on the P-limited stream communitiesof coastal British Columbia (Mallory and Richardson, 2005), suggesting that these25communities tend to become P-limited again after nutrient pulses. Another fac-tor that may have contributed to the mild response of the periphyton communityis the light limitation of our study systems. Previous investigations have demon-strated that periphyton communities in headwater streams of the Malcolm KnappResearch Forest (including Spring Creek) are strongly light-limited (Kiffney et al.,2003, 2004). Even though we did not directly measure light levels in our micro-cosms, the average Chlorophyll-a values we observed in control microcosms ofour study (0.14 ± 0.1 µg cm-2, average ± SD) closely matched values reported inthe field for undisturbed headwater streams of the Malcolm Knapp Research Forestin the fall season (0.1 ± 0.1 µg cm-2), suggesting similar levels of light-limitation(Kiffney et al., 2003). Thus, even if the response of the periphyton was limited bylight availability in our systems, such a pattern likely recreates natural conditionsof the communities under study.Consistent with our second hypothesis, sedimentation was the most influen-tial of the three stressors evaluated in this study; it had strong negative effects onmost gravel invertebrate community metrics and affected all ecosystem processesmeasured in the microcosms. These observations agree with an extensive bodyof literature reporting negative effects of sedimentation on experimental and nat-ural stream ecosystems (Wood and Armitage, 1997; Matthaei et al., 2006; Piggottet al., 2015b; Louhi et al., 2017). Furthermore, our results indicate that sedimentadditions had particularly strong effects on the pollution-sensitive EPT taxa andlarge-sized individuals in our experimental invertebrate communities, consistentwith patterns reported in previous manipulative studies (Matthaei et al., 2006; Pig-gott et al., 2015b). Increased inputs of fine sediment on stream ecosystems affectsensitive invertebrates directly by reducing habitat availability, coating gills andrespiratory surfaces, and impairing food quality and quantity (Wood and Armitage,1997; Allan, 2004). On the other hand, sedimentation reduces leaf breakdown ratesby altering microbial colonization and consumption by detritivorous invertebrates(Lecerf and Richardson, 2010), and limits periphyton growth by reducing lightavailability, while impairing substrate for recruitment (Wood and Armitage, 1997;Allan, 2004).Contrary to our third prediction, the chlorpyrifos exposure evaluated in this26study was not sufficient to cause significant lethal effects on the aquatic inverte-brate community independently. The only invertebrate metric that showed signifi-cant reductions due to the insecticide was the number of small-sized individuals inleaf packs, but this effect was only observed in absence of nutrient enrichment andsedimentation. Interestingly, even though the insecticide pulse did not cause sig-nificant invertebrate mortality in the microcosms, it still altered ecosystem functionby lowering leaf decomposition rates, supporting our fourth prediction. Becausechlorpyrifos is not reported to affect microbial communities at environmentallyrelevant concentrations (Giesy and Solomon, 2014), we attribute this result to theinhibition of invertebrate-mediated leaf decomposition. Our observations are con-sistent with this hypothesis, as the insecticide tended to reduce the biomass of leafpack invertebrates in our microcosms. This reduction suggests the possibility offeeding inhibition of invertebrate shredders, a sublethal effect that has been re-ported in the past for other insecticides (Kreutzweiser et al., 2008, 2009; Pestanaet al., 2009). This finding generally agrees with previous studies suggesting indi-rect negative effects of chlorpyrifos on leaf decomposition in experimental fresh-water ecosystems. For example, Brock et al. (1992b) and Van den Brink et al.(1996) reported in different mesocosm experiments that chlorpyrifos substantiallyreduced the number of arthropod detritivores in outdoor experimental ditches. Fur-thermore, Cuppen et al. (1995) detected reduced decomposition rates due to de-tritivore mortality in similar freshwater mesocosms. More generally, these resultscontribute to a growing body of literature reporting indirect impacts of insecticidecontamination on leaf breakdown in field and mesocosm studies (Lauridsen et al.,2006; Kreutzweiser et al., 2008; Pestana et al., 2009; Scha¨fer et al., 2012; Brosedet al., 2016).On the other hand, our fifth hypothesis predicting positive effects of chlor-pyrifos on primary production was not supported by our observations. This resultcontrasts with a number of studies reporting that chlorpyrifos increases periphytonand phytoplankton biomass through top-down effects cascading from planktonic orbenthic invertebrate grazers (Ward et al., 1995; Van den Brink et al., 2009; Williamset al., 2014). However, we do not find it entirely surprising, as periphyton commu-nities in our microcosms seemed to be nutrient and light-limited (see above).272.4.2 Interactions between chlorpyrifos and other agriculturalstressorsEven though we could not obtain a reliable estimation of chlorpyrifos concentra-tions in the water column of our microcosms, we observed that 36 h after theapplication the chemical was detected in all spiked microcosms except for thosesimultaneously treated with sediment. This qualitative observation agrees withan extensive body of literature documenting quick sorption of the chemical intoaqueous sediment and suggests accumulation and longer persistence of the pesti-cide in the sediment (Brock et al., 1992b; Giesy et al., 1999; Pablo et al., 2008;Gebremariam, 2011). Our hypothesis 6 predicted that this accumulation would en-hance the negative effects of the pesticide by increasing the contact exposure ofinvertebrates to the toxicant. We found support for this prediction in the form ofa significant reversal interaction for the abundance of small gravel invertebrates.Reversal interactions have been also defined as ”mitigating synergisms” (Piggottet al., 2015c) and represent one of the most extreme and interesting cases of non-additive effects; when the observed cumulative effects of two stressors are in theopposite direction than the predicted additive effects (Piggott et al., 2015c; Jacksonet al., 2016). In our study, such interaction suggests that even though the effectsof the insecticide alone were not significant, the presence of fine sediment in thesubstrate increased the exposure enough to affect small-sized invertebrates nega-tively. This result suggests that risk assessments assuming that invertebrates arenot susceptible to chlorpyrifos when it is adsorbed to sediment could potentiallyunderestimate its negative effects on sensitive components of the invertebrate as-semblage, and highlights the importance of measuring toxicity of sediment-boundpesticides on aquatic organisms, a risk factor that is often overlooked (Warren et al.,2003).Furthermore, our results are in keeping with previous reports of size-dependentsensitivity of aquatic invertebrates to pollutants (Kiffney and Clements, 1996b,a),since the reversal interaction was only observed on the smaller size class of ourexperimental invertebrate communities. Smaller invertebrates are more sensitive topollutants due to larger surface area to volume ratios, higher mass-specific metabolism28that potentially accelerates uptake of the toxicant, and frequent molting duringearly larval stages (Kiffney and Clements, 1996b; Townsend and Thompson, 2007).Interestingly, our findings indicate that for leaf pack invertebrates the presence offine sediment in the substrate had the opposite effect, mitigating the negative im-pacts of the insecticide on the abundance of small-sized invertebrates and the rich-ness of sensitive taxa. This observation may indicate that with added fine sedimentin the substrate, there was less chlorpyrifos exposure in the leaf pack microhabitat,which mitigated the negative effects of the toxicant on the leaf community. Pre-vious empirical research on chlorpyrifos has also reported that physical character-istics of the ecosystem, such as the presence or absence of dominant macrophytescan substantially modify the exposure of organisms to chlorpyrifos thus modifyingthe nature and magnitude of the impacts of the insecticide (Brock et al., 1992b).In agreement with hypothesis 7, we detected a significant antagonistic interac-tion between nutrient enrichment and chlorpyrifos contamination for gravel inver-tebrate richness. Even though both stressors tended to reduce gravel invertebraterichness, their combination did not produce additive taxa losses in the system.Instead, both stressors tended to eliminate the same sensitive taxa (e.g., Amele-tus, Cinygmula, Baetidae), thus producing less losses than predicted by additivity.This type of antagonistic interaction suggests that the invertebrate community un-der study presents positive species co-tolerance to eutrophication and chlorpyrifos(Vinebrooke et al., 2004). These observations contrast with a previous investigationreporting additive effects of chlorpyrifos and eutrophication in indoor microcosmsmodelled after Dutch drainage systems (Traas et al., 2004). However, a mesocosmstudy evaluating interactions between a tertiary mixture of insecticides (includingchlorpyrifos) and nutrient enrichment, also reported mitigating interactions at mod-erate levels of nutrient enrichment and sublethal levels of the insecticide mixture(Alexander et al., 2013). Furthermore, they detected that such mitigating inter-actions turned into synergistic after a threshold of nutrient enrichment that wasspecies-specific (Alexander et al., 2013). These contrasting lines of evidence fur-ther highlight that the outcome of multiple stressor interactions is strongly context-dependent (Clements et al., 2016), and deserve further investigation to unravel themechanisms and determine thresholds where eutrophication levels start interact-29ing synergistically with insecticide concentrations that are deemed safe for aquaticecosystems.2.4.3 ImplicationsCommunity-level microcosm and mesocosm experiments can incorporate enoughecological complexity to test indirect effects of the stressors, while providing rela-tively controlled conditions to isolate relevant variables, as well as sufficient repli-cation for rigorous hypothesis testing (Culp et al., 2000; Clements, 2004). How-ever, the extrapolation of our experimental results to field scenarios should be donewith care. Several simplifications imposed by the logistical constraints of our ex-periment, such as the recirculating flow, the hydraulic conditions in the tanks, therelatively short duration of the test, and the closed community population dynamics(without immigration or emigration), may result in overestimation or underestima-tion of the negative impacts of the stressor pulses. For instance, the duration of theinsecticide exposure caused by a single pulse of chlorpyrifos, is likely to be longerin recirculating microcosms than in natural environments, where the insecticide isquickly transported downstream. Hence, the effects of the pesticide on a singlehabitat patch may be overestimated. However, with these caveats in mind, ourstudy still offers some interesting insights into the single and interactive impacts ofchlorpyrifos at the microhabitat scale, revealing indirect effects and complex mul-tiple stressor interactions that deserve further investigation. Our results support abody of evidence showing that multiple stressor interactions should be explicitlyconsidered when conducting environmental risk assessments for insecticides, aseven at low insecticide concentrations, synergistic interactions with other commonagricultural stressors may result in unexpected negative effects for aquatic inver-tebrate communities (Alexander and Culp, 2008; Alexander et al., 2016). Fur-thermore, the differential effects of the insecticide on habitat patches, highlightthe importance of habitat structure in modulating the impacts of organic toxicantson non-target species, keeping with a growing body of literature suggesting thatin-stream habitat conditions must be taken into account when conducting envi-ronmental risk assessments for pesticides (Rasmussen et al., 2011, 2012; Scha¨fer30et al., 2016). Additionally, our detailed study at the scale of microhabitats allowedus to document chlorpyrifos-induced effects on invertebrate communities inhab-iting leaf packs, along with lower leaf decomposition rates. These results are inkeeping with field studies correlating leaf breakdown inhibition with pesticide ex-posures (Scha¨fer et al., 2007, 2012; Rasmussen et al., 2012; Brosed et al., 2016),and offer a mechanistic explanation for such impacts, underscoring the indirecteffects of insecticides through invertebrate decomposers at sublethal concentra-tions. Moreover, our results strongly support the notion that leaf decomposition isa useful early indicator of ecological impairment in stream ecosystems and shouldcomplement the suite of structural measurements already employed in biomonitor-ing (Gessner and Chauvet, 2002; Pestana et al., 2009; Scha¨fer et al., 2012; Brosedet al., 2016).31Table 2.1: ANOVA summary of linear fixed effects models evaluating impacts of stressor treatments on ecosystemfunctioning and invertebrate variables recorded on day 15 of the experiment. Significant effects are indicated inbold (P < 0.05).N S I N * S N * I S * I N * S *IResponse variables df F P F P F P F P F P F P F PEcosystem responsesLeaf decomposition 1, 24 2.881 0.103 4.592 0.042 8.864 0.007 0.648 0.429 1.273 0.270 0.035 0.854 0.955 0.338Periphyton biomass 1, 24 0.207 0.653 8.215 0.009 2.516 0.126 0.665 0.423 0.301 0.588 1.697 0.205 0.147 0.704Gravel invertebratesAbundance 1, 22 2.999 0.097 4.645 0.042 0.094 0.762 0.001 0.975 1.476 0.237 3.622 0.070 0.784 0.386Biomass 1, 22 0.018 0.893 1.298 0.267 0.167 0.687 0.105 0.749 0.649 0.429 0.064 0.803 0.047 0.831Richness 1, 22 3.935 0.060 6.504 0.018 1.002 0.328 0.437 0.515 5.939 0.023 0.389 0.539 0.017 0.896EPT abundance 1, 22 2.999 0.097 4.645 0.042 0.094 0.762 0.001 0.975 1.476 0.237 3.622 0.070 0.784 0.386EPT biomass 1, 22 0.000 0.986 0.565 0.460 1.134 0.298 0.064 0.802 0.227 0.638 0.797 0.382 0.004 0.951EPT richness 1, 22 1.955 0.176 4.399 0.048 2.225 0.150 0.009 0.927 4.399 0.048 0.313 0.582 0.002 0.963Average body size 1, 22 2.726 0.113 1.876 0.185 0.191 0.667 0.217 0.646 1.665 0.210 2.876 0.104 0.157 0.696Abundance small size 1, 22 3.766 0.065 3.196 0.088 0.203 0.657 0.011 0.917 0.900 0.353 4.706 0.041 0.977 0.334Abundance large size 1, 22 0.001 0.980 4.798 0.039 0.132 0.720 0.057 0.814 2.072 0.164 0.010 0.921 0.001 0.980Leaf invertebratesAbundance 1, 22 0.050 0.825 0.002 0.968 0.940 0.343 0.368 0.551 2.611 0.120 4.300 0.050 1.341 0.259Biomass 1, 22 0.005 0.942 0.117 0.736 2.993 0.098 0.203 0.657 0.435 0.517 0.753 0.395 0.017 0.899Richness 1, 22 0.174 0.681 0.113 0.740 0.763 0.392 0.157 0.696 1.038 0.319 0.139 0.713 3.254 0.085EPT abundance 1, 22 0.889 0.356 0.013 0.910 0.978 0.333 0.887 0.356 0.138 0.714 1.016 0.325 1.785 0.195EPT biomass 1, 22 2.535 0.126 0.626 0.437 2.050 0.166 1.165 0.292 0.743 0.398 1.333 0.261 0.099 0.756EPT richness 1, 22 1.202 0.285 0.079 0.781 0.934 0.344 1.455 0.241 0.091 0.766 2.817 0.107 4.005 0.058Average body size 1, 22 0.876 0.360 0.016 0.900 1.546 0.227 0.000 0.983 0.006 0.938 0.063 0.804 1.107 0.304Abundance small size 1, 22 0.299 0.590 0.004 0.948 0.913 0.350 0.913 0.350 3.458 0.076 5.276 0.032 1.212 0.283Abundance large size 1, 22 1.123 0.301 2.201 0.152 0.719 0.406 0.719 0.406 0.404 0.531 0.045 0.834 0.180 0.67632a.b.Figure 2.1: Photo of the experimental set-up consisting in 32 laboratory mi-crocosms located at the University of British Columbia, Vancouver,Canada (a). Detail of the microcosms (b).33Figure 2.2: Standardized effect size (Hedge’s d ± 95% CI) for significantmain and interactive effects of the stressors on gravel invertebrate com-munities. Letters in the y-axis represent main effects (N, nutrients; S,sediment; I, insecticide) and interaction terms (N*S, nutrient x sedi-ment; N*I, nutrient x insecticide; S*I, sediment x insecticide; N*S*I;nutrient x sediment x insecticide). For main effects, significant (i.e. notoverlapping zero) positive values indicate increases in the response vari-able whereas negative values denote the opposite. For interactions, con-fidence intervals overlapping zero indicate additive interactions, posi-tive values indicate synergies, and negative values indicate antagonisticor reversal interactions (reversals are marked with an R). Symbols areused to represent significance of the effects according to the ANOVAs:*P < 0.05, ·P < 0.1.34Figure 2.3: Bar plots illustrating significant interactive effects of the stres-sors on gravel (a-b) and leaf (c-e) invertebrate metrics, according to theANOVAs (P = 0.05). All shown interactions were classified as antag-onistic, except for panel (b) which was classified as a reversal. Letternotation for the treatments is consistent with Figure 2.2. Bars representthe mean of four replicates (± SE).35Figure 2.4: Distance-based redundancy analysis on log-transformed abun-dance data of invertebrates collected in gravel. Only the most abundanttaxa are labeled. Letter notation for the treatments is consistent withFigure 2.2. Solid arrows indicate significant main effects (P < 0.05)and dashed arrows indicate marginally significant effects (P = 0.059),according to a permutation test of the full model with 999 randomiza-tions. The treatments together explained 29% of the total variation oncommunity composition of gravel. Axes 1 and 2 represent 10 and 5%of the total variation, respectively.36Figure 2.5: Standardized effect size (Hedge’s d ±95% CI) for significantmain and interactive effects of the stressors on invertebrate commu-nities collected in leaf packs. Letter notation for the treatments andinterpretation of effects is consistent with Figure 2.2. Symbols areused to represent significance of the effects according to the ANOVAs:*P < 0.05, ·P < 0.1.37Figure 2.6: Distance-based redundancy analysis on log-transformed abun-dance data of invertebrates collected in leaf packs. Only the mostabundant taxa are labeled. Letter notation for the treatments is con-sistent with Figure 2.2. Solid arrows indicate significant main effects(P < 0.05) according to a permutation test of the full model with 999randomizations. The treatments together explained 27% of the totalvariation on community composition of leaf packs. Axes 1 and 2 repre-sent 7 and 6% of the total variation, respectively.38Figure 2.7: Standardized effect size (Hedge’s d ±95% CI) for significantmain and interactive effects of the stressors on ecosystem processes.Letter notation for the treatments and interpretation of effects is consis-tent with Figure 2.2. Symbols are used to represent significance of theeffects according to the ANOVAs: **P < 0.01, *P < 0.05.39Chapter 3Nutrients and sediment modifythe impacts of a neonicotinoidinsecticide on experimental pondecosystems3.1 IntroductionFreshwater ecosystems are amongst the most threatened ecosystems in the world(Dudgeon et al., 2006; Vo¨ro¨smarty et al., 2010; Strayer and Dudgeon, 2010).Threats to freshwaters have been extensively documented, with habitat degrada-tion and loss, nutrient enrichment, flow modification, overexploitation, introduc-tion of non-native species, and chemical pollution figuring as some of the stressorsdriving the declining trends (Dyke, 2008; Vo¨ro¨smarty et al., 2010; Scha¨fer et al.,2016). With the increasing recognition that these stressors rarely operate in isola-tion, the last decade has seen growing interest in the experimental assessment oftheir cumulative effects. This experimental approach has improved our mechanis-tic understanding of the impacts of stressors like nutrient enrichment, sedimenta-40tion, temperature, acidification, and water abstraction (e.g., Matthaei et al., 2010;Piggott et al., 2015b; Alexander et al., 2016). Most importantly, this approachhas demonstrated that the combined action of these stressors frequently results inimpacts that cannot be predicted from the addition of the stressors’ individual ef-fects, highlighting the importance of this type of experimental work to understandand predict complex multiple-stressor interactions (Wagenhoff et al., 2012; Shurinet al., 2012; Piggott et al., 2015b; Jackson et al., 2016).Despite these advances, our understanding about cumulative effects of someimportant freshwater ecosystem stressors is still relatively limited. Neonicotinoidinsecticides are a clear example of this situation. Neonicotinoid insecticides wereintroduced as systemic pest controls in the early 1990s, but their large-scale ap-plication started around 2004 and has rapidly increased to the point that they arecurrently the most widely used insecticide family in the world (Main et al., 2014;Simon-Delso et al., 2015). They have been the object of great controversy in thelast few years, as they have been associated with the colony collapse disorder caus-ing alarming honey bee declines around the world, and have been attributed to toxiceffects on mammals (Ko¨hler and Triebskorn, 2013). These controversial findingshave raised interest in these insecticides, and a number of studies have addressedtheir potential negative effects on non-target organisms, including some freshwa-ter species (e.g. Kreutzweiser et al., 2008; Langer-Jaesrich et al., 2010; Churchelet al., 2011; Bo¨ttger et al., 2013). However, to date there is relatively little empir-ical information about interactions between neonicotinoids and other agriculturalstressors in freshwater ecosystems (but see Alexander et al., 2013, 2016).Neonicotinoid insecticides are frequently detected in puddles, irrigation chan-nels, streams, rivers, and wetlands in agricultural lands of different countries (Mainet al., 2014; Morrissey et al., 2015; Sa´nchez-Bayo et al., 2016). In these waterbodies, freshwater communities are simultaneously exposed to other agriculturalstressors that may affect their sensitivity to the neonicotinoid insecticides. Forexample, some agricultural practices result in increased inputs of fine sedimentinto water bodies (Allan, 2004). Elevated inputs of fine sediment increase waterturbidity reducing light penetration and primary production, impair substrata forinvertebrate habitat and periphyton growth, and smother respiratory organs of sen-41sitive invertebrate taxa (Knowlton and Jones, 1995; Wood and Armitage, 1997;Matthaei et al., 2006). On the other hand, the application of fertilizers in agricul-tural landscapes often causes nutrient enrichment of freshwater ecosystems (Allan,2004). At moderate levels this enrichment has positive bottom-up effects in fresh-water food webs, but at high levels they might lead to extensive algae growth andconsequent decreases in dissolved oxygen that ultimately result in loss of sensi-tive species (Allan, 2004; Wagenhoff et al., 2012; Garcı´a et al., 2017). There aremultiple mechanisms by which neonicotinoids may interact non-additively withsedimentation and nutrient enrichment. For example, imidacloprid, the most pop-ular insecticide in this family, is relatively soluble in water (610 mg L-1 in 20◦CH2O; log Kow = 0.57), but it tends to persist longer in sediment than in water (waterhalf-life under light conditions: 4 h, sediment half-life under anaerobic conditions:27 days; CCME, 2007). Thus, imidacloprid may interact synergistically with sedi-mentation in agricultural landscapes, if its longer persistence in sediment increasescontact exposure of benthic organisms to the insecticide. Furthermore, a recentexperiment reported that high levels of nutrient enrichment were able to mask thenegative effects of imidacloprid on benthic invertebrate communities (Alexanderet al., 2016).In addition to potential multiple-stressor interactions, some experiments havesuggested the possibility of indirect impacts of sublethal levels of imidaclopridcontamination on important freshwater ecosystem processes. Using single-speciesmicrocosms and outdoor stream mesocosms, respectively, Kreutzweiser et al. (2008)and Pestana et al. (2009) demonstrated that imidacloprid contamination reducedleaf breakdown rates through feeding inhibition of detritivorous invertebrates. More-over, Alexander et al. (2016) observed in their stream mesocosm experiment thatimidacloprid contamination enhanced the positive indirect impacts of predation onperiphyton biomass. These findings underscore that even sublethal concentrationsof this common insecticide may alter primary productivity and organic matter dy-namics in freshwater ecosystems, and highlight the need for more empirical studiesexplicitly assessing the effects of the insecticide on integrative measurements ofecosystem functioning, such as ecosystem metabolism.Here, we present the results of an ecosystem-scale, mesocosm experiment de-42signed to evaluate individual and combined effects of imidacloprid, sedimentation,and nutrient enrichment on structure and function of pond ecosystems. To ourknowledge, this is the first time these three stressors have been manipulated simul-taneously, thus we were particularly interested in testing whether their combinedeffects could be predicted from the addition of their individual effects. Addition-ally, we evaluated indirect impacts of the insecticide on ecosystem metabolism andorganic matter decomposition. We predicted that at the stressor levels we testedin our study, we would observe: 1) positive individual effects of nutrients on leafdecomposition, primary production, and invertebrate biomass; 2) negative individ-ual effects of sedimentation on most community metrics and ecosystem processes;3) negative effects of imidacloprid on benthic invertebrate density; 4) no effects ofimidacloprid on zooplankton density; 5) indirect positive effects of imidacloprid onnet ecosystem productivity through the inhibition of grazers; 6) indirect negativeeffects of imidacloprid on leaf decomposition through the inhibition of shredders;7) antagonistic interactions between imidacloprid and nutrient enrichment mitigat-ing negative effects on invertebrate communities; and 8) synergistic interactionsbetween imidacloprid and sedimentation exacerbating negative impacts on inverte-brate communities.3.2 Methods3.2.1 Experimental designWe conducted a 36-day factorial manipulation of fine sediment, nutrients, and imi-dacloprid concentration in 32 outdoor freshwater mesocosms (1136 L plastic tanks,0.6 m deep, 1.4 m in diameter; Rubbermaidr, Atlanta, GA, USA), located at theUniversity of British Columbia’s experimental pond facility, Vancouver, Canada(49◦ 14’ 52.1” N, 123◦ 13’ 55.9” W; Figure 3.1). Four months before the on-set of the experiment, tanks were filled with 616 L of municipal water and left todechlorinate by degassing for three weeks. Then they were stocked with substrateand organic matter to provide suitable habitat for invertebrates, this included: a 3cm layer of medium sand (< 5 mm grain size) and washed mixed gravels (0.5-243cm grain size range), 25 g of rabbit pellets, and 50 g of air-dried red alder (Alnusrubra) leaves. Three months prior to treatment application, each mesocosm wasinoculated with 10 L of unfiltered lake water and a 1 L aliquot of concentratedlive plankton (collected using a 64 µm mesh conical tow net) from local lakes,to provide colonists of zooplankton and phytoplankton. Similarly, 1 L of benthicsediments from local lakes were applied to each mesocosm to provide colonists ofbenthic invertebrates. All collected sediment, water, and zooplankton was mixedthoroughly before addition to ensure mesocosms were receiving similar planktonicand benthic communities. Additionally, three weeks after inoculation, 40 L ofwater were exchanged between each tank and each of the six tanks closest to it,in order to homogenize planktonic communities. All tanks were left uncoveredthroughout the experiment to allow natural colonization and emigration of aquaticinvertebrates. Water levels were maintained by natural precipitation and additionof equal volumes of dechlorinated municipal water to each tank once a week. Ac-cording to handheld probe measurements conducted one day before the applicationof the treatments, there were no significant differences (P> 0.05) in dissolved oxy-gen (8.81 ± 0.1 mg L-1, mean ± SE, n = 32), conductivity (35.4 ± 0.42 µS cm-1),and pH (7.7 ± 0.06), among tanks assigned to different stressor treatments.3.2.2 Stressor treatmentsOur experimental manipulation was conducted in the summer season, from June 4to July 11 of 2015. The experimental design involved two levels of each stressor(added, ambient) in a fully crossed factorial design (2 x 2 x 2), with eight treat-ment combinations and four replicates per treatment. Treatments were randomlyassigned to mesocosms and designed to simulate pulsed exposures, similar to thoseexperienced by pond ecosystems in agricultural landscapes, due to fluctuation inrainfall and runoff events, seasonal application of fertilizers and pesticides, and ac-cidental spills. Thus, fine sediment was added weekly, starting on June 4 (day 1),while nutrients and insecticide were applied only twice throughout the experiment(days 1 and 23).We used powdered kaolin (Ward’sr Kaolin, 74 µm mesh size) for the sedi-44ment treatment. Kaolin is an inorganic clay with similar optical and physical prop-erties to naturally-occurring suspended silts (Boube´e et al., 1997), which is oftenused in studies of suspended solids in aquatic ecosystems (e.g., Boube´e et al., 1997;Sanpera-Calbet et al., 2012). It has a pH of 7.5, an average organic carbon contentof 14%, and a volume-specific mass of 2.6 g cm-3. We added 90 g of kaolin ineach sediment-treatment tank once a week. To promote particle suspension, priorto addition we mixed the kaolin for each tank in 1 L of water from the same tankin a separate container. The water-kaolin mix was then added as evenly as possibleonto the surface of the treated tank, and the water column was stirred. For stirring,we used a 40 cm-long PVC pipe (3/4” diameter), which was immersed in the waterhalf-length and rotated clockwise three times without touching the tank walls. Tocontrol for a potential confounding effect of the stirring process, we also stirred thenon-sediment treatment tanks on the days of sediment additions. Each tank had itsown PVC mixer to avoid cross-contamination.Background nutrient concentrations (NH4-N: 44.8± 1.2 µg L-1, NO3-N: 100.5± 46.3 µg L-1, PO4-P: 4.9 ± 0.5 µg L-1) were augmented by adding potassiumphosphate (target P concentration: 50 µg L-1 above background concentration)and ammonium nitrate (target N concentration: 1500 µg L-1 above backgroundconcentration) on days 1 and 23 of the experiment. A stock solution of 1.2 g L-1imidacloprid was prepared by dissolving analytical grade imidacloprid (≥100%,Pestanalr, Sigma-Aldrich) in deionized water. We applied 1.8 mL of the stockinsecticide solution to each insecticide tank using a micropipette on days 1 and 23,for target imidacloprid pulses of 3.5 µg L-1. Our target imidacloprid pulses fallwithin the range of concentration values reported for neonicotinoid insecticides insurface waters of different agricultural regions in North America (Giroux, 2003;Starner and Goh, 2012; Anderson et al., 2013).3.2.3 Response variablesWe measured variables reflecting the effect of the treatments on habitat condition,pelagic and benthic communities, and ecosystem function. Water samples werecollected 3 hours following treatment additions on day 1 and again on day 36 to de-45termine imidacloprid and nutrient concentrations (PO4-P, NH4-N, and NO3-N) inthe tanks. Nutrients were analyzed by the Analytical Chemistry Laboratory of theBritish Columbia Ministry of Environment, Victoria, BC, using standard methods(APHA, 2005). Water samples for imidacloprid determination were analyzed at theNational Hydrology Research Centre, Environment and Climate Change Canada,Saskatoon, SK, using the methods described in Main et al. (2014). Measures ofdissolved oxygen concentration (DO), conductivity, and pH were recorded weeklyusing hand-held probes after the application of treatments. Weekly water samples(100 mL) were transported to the laboratory to measure turbidity using a digitaldesktop turbidity meter. Sedimentation rates were measured in the mesocosms byfixing three uncapped Falcon tubes vertically in different random points of the bot-tom of each tank. Falcon tubes were recovered on day 36, capped, and transportedto the laboratory for drying (40◦C for 5 d) and weighing.We sampled pelagic communities weekly to estimate phytoplankton and zoo-plankton biomass. Phytoplankton biomass was estimated by collecting 2 L samplesfrom the water column of each tank in opaque plastic containers. Containers weretransported to the lab, where the water was filtered onto precombusted glass fiberfilters. Filters were transferred to 15 mL Falcon tubes and covered with 90% ace-tone during 24 h in the dark (at 4◦C), where they were homogenized by agitationthree times during the steeping period. Once the steeping period was completed,samples were analyzed for chlorophyll-a in a Turner Designs TD-700 laboratoryfluorometer according to standard fluorometric methods (Arar and Collins, 1997).Zooplankton biomass was estimated from one composite sample per tank collectedwith a small conical tow net (26.6 cm opening diameter, 64 µm mesh). Each com-posite sample consisted of two random 30 cm-deep tows that were immediatelyfixed in 70% ethanol and transported to the laboratory. In the laboratory, sampleswere divided in half using a Folsom plankton splitter. One half of the sample wasinspected under the dissecting microscope to remove all pieces of organic mat-ter that were not zooplankton (exuviae, winged insects, seeds, etc.) and filteredonto a glass microfiber filter (Whatmanr GF/F). Filters were dried at 60◦C for 5days, weighed, ashed at 500◦C, and reweighed to estimate zooplankton ash-freedry mass (AFDM). For days 26 and 33 the remaining half of the sample was used46for zooplankton counting and identification, using a Bogorov counting chamberunder a dissecting microscope. A 5 mL Hensen-Stempel pipette was used to ob-tain subsamples until at least 200 individuals were identified and enumerated, inorder to estimate zooplankton densities. Zooplankters were identified to the lowestpractical taxonomic resolution, usually genus.We sampled benthic communities only once, at the end of the experiment (day36). To estimate periphyton biomass we introduced three unglazed, 84 cm2 ce-ramic tiles in each mesocosm on day 0 and collected them on day 36. Tiles werestored in the dark at -18◦C until processing for chlorophyll-a estimation throughfluorometry following the methods described above. Benthic invertebrates weresampled by introducing one cylindrical benthic trap (15 cm diameter, 10 cm high)in each mesocosm two weeks before the application of the treatments. Each trapconsisted of a circular plastic bottom with coarse-mesh plastic walls (square meshfence, 1 cm mesh size), filled with a 3 cm layer of mixed gravels and 1 g of air-dried red alder leaves. On day 36, we extracted the traps with the help of a D-net(250 µm mesh size) to prevent the loss of invertebrates and washed their contentsthrough a 2 mm sieve stacked on top of a 250 µm sieve. Gravel retained in the2 mm sieve was sorted for invertebrates at the time of collection, and the mate-rial retained in the 250 µm sieve was stored in 70% ethanol and sorted later underthe dissecting microscope. Invertebrates were enumerated, identified to the lowestpractical taxonomic level (usually genus), and measured to determine size and drymass from length-mass regressions (Smock, 1980; Benke et al., 1999; Johnstonand Cunjak, 1999). Finally, invertebrates were classified into three size categories:small (< 2.5 mm), medium (2.5 - 5 mm), and large (> 5 mm).3.2.4 Ecosystem functionOne day before treatment application (day 0), we introduced three leaf packs ofred alder leaves in coarse-mesh bags (3 g air-dry weight; 10 mm mesh) into eachmesocosm to measure leaf decomposition. On day 36, leaf packs were removedfrom the mesocosms and stored at -18◦C until processed in the laboratory to es-timate remaining AFDM, according to standard methods (Hauer and Lamberti,472007). Posterior processing involved defrosting, rinsing, and drying at 60◦C for 5days. Remaining leaf material was then weighed, ashed at 500◦C, and reweighedto calculate AFDM.Net ecosystem production (NEP) was measured four times throughout the ex-periment (days 7, 22, 29, and 35) using gas-exchange methods. NEP was estimatedas the difference in DO concentration between dusk and dawn, reflecting the rateof change in DO between the hours with maximum (dusk) and minimum (dawn)concentrations. DO measurements were recorded hourly for the full 24-h cycleon the first sampling date, in order to determine the hours of the day at whichmaximum and minimum DO concentrations were detected in the tanks. For theremaining sampling dates, tanks were sampled only at such hours when maximumand minimum values were recorded; 17:00 and 7:00, respectively.3.2.5 Data analysisWe used linear models to quantify the individual and combined effects of nutrientenrichment, sedimentation, and insecticide contamination on univariate responses.Variables with one measurement per mesocosm were analyzed with fixed effectsmodels, including the three stressor treatments and all the possible two-way andthree-way combinations among them. Variables with multiple measurements permesocosm were analyzed with linear mixed effects models (LME) that also in-cluded the individual mesocosm and the sampling date as random effects. UsingLMEs for response variables with repeated measurements, allowed us to focus oneffects that were significant across the duration of the experiment, while facilitatingthe collective interpretation of responses measured at different time intervals.Significance levels for all our tests was P< 0.05 and was evaluated with ANOVAtype III sum of squares. Following the recommendation of Nakagawa and Cuthill(2007) we present standardized effect size estimates for all findings with P < 0.1,so readers can judge the biological importance of the results. Hedges d estimatesof effect size (Gurevitch and Hedges, 2006) were calculated from the t values ofour linear models using the formulas provided by Nakagawa and Cuthill (2007).48In order to improve the graphical representation of our results, we coded signifi-cant main effects to represent the direction of the response of manipulated versuscontrol mesocosms (i.e. positive effect sizes indicate increases in the response vari-able, while negative effect sizes indicate the opposite). Further, we coded 2-wayand 3-way interaction effect sizes to represent the classification of the interactionaccording to the framework proposed by Jackson et al. (2016). Thus, positive in-teraction effect sizes represent synergistic interactions (i.e. the combined effect ofthe stressors is greater than the sum of their individual effects), whereas negativeeffect sizes represent either antagonistic interactions (i.e. the combined effect ofthe stressors is less than the sum of their individual effects) or reversal interactions(i.e. the combined effect of the stressors is in the opposite direction than the sumof the individual effects).We used redundancy analysis (RDA) to evaluate the effect of the stressor treat-ments on multivariate taxa composition of zooplankton and benthic invertebrates.Zooplankton density and benthic invertebrate abundance were Helliger-transformedto reduce the influence of rare taxa (Legendre and Gallagher, 2001). Significanceof main effects and interactions was assessed with Monte Carlo permutations ofthe full model with 999 randomizations. We also used redundancy analysis (RDA)to evaluate if the treatments affected the distribution of biomass across the pelagicand benthic food webs in the ponds. In order to facilitate comparisons, biomassunits from all compartments were converted into g C (g C m3 for pelagic organ-isms, g C m2 for benthic organisms). We assumed a carbon biomass to chlorophyllbiomass ratio of 40:1 for phytoplankton and 50:1 for periphyton (Shurin et al.,2012). For zooplankton we had AFDM, so we assumed an average 10% of ashcontent (Waters, 1977), and an average carbon content of 48% of dry mass (Ander-sen and Hessen, 1991). For benthic invertebrates we assumed an average carboncontent 48.3% of dry mass (Evans-White et al., 2005).When necessary, root transformations were applied to improve normality ofcount variables with positively skewed distributions. Log-transformations wereused to improve normality and homoscedasticity in other types of variables (Quinnand Keough, 2002). All analyses were performed in R v. 3.3.0 (R Core Team,2016), using packages lme4 (Bates et al., 2015), car (Fox and Weisberg, 2011),49and vegan (Oksanen et al. 2016).3.3 Results3.3.1 Water quality and habitat characteristicsWater samples collected on day one, three hours after treatment application, showedthat nutrient additions caused significant increases in phosphate (609% increase,PO4-P: 35.1 ± 0.6 µg L-1, mean nutrient treatment ± SE, n = 16, Table 3.1)and ammonia nitrogen concentrations (939% increase, NH4-N: 465.6 ± 10.9 µgL-1). Nutrient additions also increased mean nitrate concentrations (125% increase,NO3-N: 226.4 ± 75.5 µg L-1) but the effect was not statistically significant, prob-ably due to the large variation in nitrate levels recorded in nutrient addition tanks.On the other hand, the insecticide treatment had a significant effect on nitrate,causing a 76% reduction in measured concentration values. Nutrient additions alsotended to reduce DIN:DIP ratio (29% reduction) but again, the effect was not sig-nificant at the time of collection (DIN:DIP: 20.1± 2.7). Three hours after the firstinsecticide application, average imidacloprid concentration in the water was 1.97± 0.01 µg L-1 in insecticide tanks (Table 3.2). Even though this exposure levelwas lower than our target concentration (3.5 µg L-1), it was consistent across alltanks treated with insecticide. Average imidacloprid concentration on day 35 (12days after the second insecticide pulse) was 0.071 ± 0.01 µg L-1, indicating thatdespite the expected decrease in imidacloprid concentration after each pulse, therewas some level of insecticide exposure in the treated tanks throughout the durationof the experiment.According to biweekly turbidity measurements, the sediment addition causedan average six-fold increase in turbidity across the duration of the experiment (Ta-ble 3.1). The sediment treatment also resulted in significantly higher sedimentationrates; sediment tanks had on average 2.4 times more sediment accumulated in thesubstrate by the end of the experiment. According to weekly daytime measure-ments, water in nutrient-enriched mesocosms had 14% more dissolved oxygen and508% higher pH than the rest of the mesocosms (P < 0.05). Insecticide applicationsalso caused significant increases in water pH (4% increase). Conductivity was notsignificantly affected by any of the treatments throughout the experiment.3.3.2 Pelagic communityOur weekly biomass measurements showed that the stressors had strong indepen-dent and interactive effects on the biomass of the pelagic community, however,their effects varied through time (Figure 3.2a-d). Nutrient enrichment was the mostinfluential stressor, causing increases in zooplankton and phytoplankton biomassstarting one week after the second nutrient pulse in the case of phytoplankton, andtwo weeks after the first nutrient pulse in the case of zooplankton. Linear mixed ef-fects models controlling for time (Figure 3.3, Table 3.3), showed that across the du-ration of the experiment, nutrient enrichment tripled phytoplankton biomass (231%increase, Figure 3.3a) and doubled zooplankton biomass (102% increase), whereassedimentation reduced the latter by 28% (Figure 3.3b). Furthermore, density anddiversity recorded in the last two weeks of the experiment indicated that insecticidepulses reduced zooplankton density by 63% (Figure 3.3c), while sediment pulseshad negative effects on richness (12% reduction, Figure 3.3d).Our results indicate that the effects of the three stressors on the zooplanktoncommunity were not independent. We detected significant antagonistic interac-tions between sediment and insecticide, and between the three stressors affectingzooplankton richness in the ponds (Figure 3.3d). Zooplankton richness in sedimentx insecticide mesocosms was 2 times higher than expected if the two stressorswere independent. Similarly, richness in mesocosms receiving the three stressortreatments was on average 2.4 times higher than expected if the three stressorswere additive (Figure 3.4a). The Shannon-Wiener and the Evenness diversity in-dices also showed significant 3-way antagonistic interactions; the three stressorscombined produced Shannon-Wiener diversity 3.15 times higher and Evenness 1.5times higher than expected if their impacts were additive (Figure 3.4b, c).The redundancy analysis (RDA) showed that sediment and insecticide addi-51tions significantly altered the composition of zooplankton communities in the meso-cosms by the end of the experiment (Figure 3.5a). Insecticide pulses were stronglycorrelated with lower densities of Bosmina and higher densities of Chydoridae andDiaphanosoma; while sediment additions increased Diaphanosoma and reducedDaphnia densities.3.3.3 Benthic communityBenthic invertebrate samples collected at the end of the experiment showed that thesediment pulses caused a 32% reduction in invertebrate abundance (Figure 3.6a),mainly due to negative impacts on the number of small (61% reduction, Figure3.6b) and medium-sized individuals (35% reduction, Figure 3.6c). However, de-spite having stronger impacts on the smaller size categories, the application of sed-iment did not alter the average size of the benthic invertebrate community (Table3.3). Sediment pulses also had marginally significant (P = 0.07) impacts on inver-tebrate taxa richness (10% reduction, Figure 3.6e), but did not significantly affectShannon-Wiener diversity or Evenness. Nutrient additions had positive effects onthe benthic invertebrate community, increasing the number of invertebrate taxa inthe mesocosms by 20% (Figure 3.6e). We did not detect strong independent effectsof the insecticide pulses on benthic invertebrate communities. However, imidaclo-prid caused a marginally significant reduction of the abundance of medium-sizedindividuals (35% reduction, Figure 3.6c), and interacted non-additively with sedi-mentation and nutrient enrichment (Figure 3.6d). We observed a significant sedi-ment x insecticide reversal interaction affecting the abundance of large individuals;while the predicted additive effect of the two stressors was negative, the observedeffect of the stressor combination was positive (abundance of large individuals insediment x insecticide mesocosms was 1.7 times higher than predicted by addi-tivity, Figure 3.4d). Furthermore, imidacloprid attenuated the positive effect ofnutrient enrichment on benthic invertebrate richness; nutrient x insecticide tankshad 19% lower taxa richness than expected if the effects of the two stressors wereindependent (antagonistic interaction, Figure 3.4e).RDA ordination of benthic invertebrates collected the last day of the experi-52ment indicated that nutrient pulses significantly altered composition of benthic in-vertebrate communities in the ponds (RDA nutrient term: P < 0.001, Figure 3.5b),mainly by reducing the abundance of Orthocladiinae midge larvae, while increas-ing the abundance of dragonfly larvae from the genus Sympetrum and midge lar-vae from the subfamily Chironominae. According to the RDA, sedimentation andinsecticide had only marginally significant effects on benthic invertebrate com-position, which were associated with lower densities of Procloeon mayflies andtanypodine midges, as well as higher abundance of Chironominae and Sympetrum(Figure 3.5b).3.3.4 Biomass distribution among ecological compartmentsWe used redundancy analysis to examine the impact of the stressors on the dis-tribution of biomass among the trophic compartments we measured (benthic andpelagic) on the last day of the experiment (Figure 3.7a). Unsurprisingly, nutri-ent enrichment was the only stressor that significantly altered the distribution ofbiomass among compartments, explaining 27% of the variation in biomass (RDAnutrient term, P < 0.01). According to the RDA ordination, nutrient pulses werestrongly correlated with increases in zooplankton and phytoplankton biomass, andreductions of benthic invertebrate biomass in the mesocosms. Plots comparingbiomass in mesocosms with and without nutrient additions revealed that zooplank-ton was disproportionally favoured by nutrient enrichment, showing an average68% increase in biomass (Figure 3.7b). Phytoplankton, which had much lessbiomass relative to zooplankton in all mesocosms, was also enhanced by nutrientenrichment (83% increase). On the other hand, consumers in the benthic habitatcompartment were negatively affected by the enrichment (26% reduction), whilebenthic primary producers were not strongly affected.3.3.5 Ecosystem functionThe stressor treatments had strong independent effects on net ecosystem produc-tion (NEP) of the mesocosms throughout the experiment. Nutrient enrichment pro-53duced the quickest and strongest response on NEP, with significant positive effectsfrom the first week after the application of the treatment that became the strongestone week after the second nutrient addition (Figure 3.2e, f). We also observed neg-ative effects of sedimentation starting after the fourth sediment addition (day 22)until the end of the experiment. LME models of the whole time series indicated thatoverall, nutrient enrichment increased NEP by 50%, while sedimentation decreasedit by 14% (Table 3.3, Figure 3.8a). Insecticide pulses also increased average NEPthroughout the experiment, but their effects were not statistically significant (9%increase, P < 0.1). Nutrient enrichment was the only treatment with significanteffects on biochemical oxygen demand (BOD), increasing it by 14% (Figure 3.8b).Contrary to our expectations we found no evidence of significant effects of any ofthe treatments on leaf decomposition by the end of the experiment (Table 3.3).3.4 DiscussionOur findings support the hypothesis that environmentally relevant concentrations ofimidacloprid, a widely used systemic insecticide, may interact non-additively withnutrient enrichment and sedimentation, two of the most widespread stressors inagricultural landscapes (Allan, 2004). These results are consistent with a growingbody of evidence suggesting that additive frameworks may underestimate or over-estimate cumulative effects of common stressors on freshwater ecosystems (Shurinet al., 2012; Piggott et al., 2015b; Wagenhoff et al., 2016). Furthermore, our ob-servations suggest that imidacloprid has the pontential to alter freshwater ecosys-tem metabolism, underscoring the importance of monitoring ecosystem function inecological impact assessments (Young et al., 2008).3.4.1 Single stressors strongly affected ecosystem function andstructureTable 3.4 presents a summary of our initial hypothesis and their respective out-comes. In agreement with our first hypothesis, the nutrient enrichment had strongpositive effects on the pelagic food web and enhanced net ecosystem production54and biochemical oxygen demand in the mesocosms. However, contrasting withour notion that nutrients would have positive effects across all ecological compart-ments, we observed mostly negative effects of the enrichment on benthic food websin our ponds. In benthic food webs, nutrient additions increased taxa richness butshifted community composition towards more tolerant taxa, and reduced benthicinvertebrate biomass. We attribute this tendency towards negative effects on thebenthic food web to the strong response of the pelagic compartment to the nutrientsubsidy. The rapid increase in phytoplankton biomass may have reduced light pen-etration, thereby limiting periphyton growth and negating the positive effects of thesubsidy in the benthos (Scheffer et al., 1993). Other mesocosm and field studieshave reported similar responses to nutrient enrichment in lentic ecosystems, wherethe rapid response of the phytoplankton shades periphyton growth, dampening itsresponse (Brock et al., 1992b). However, these findings contrast with results ofa similar mesocosm experiment of longer duration (16 months), where nutrientadditions enhanced biomass across all trophic levels (Shurin et al., 2012).Our results largely supported our second hypothesis predicting negative effectsof sedimentation in most ecological compartments of the pond ecosystem. Weeklysediment additions significantly reduced zooplankton biomass and richness, low-ered benthic invertebrate abundance, and limited net ecosystem production in theponds. These observations are consistent with a large body of literature reportingstrong negative effects of fine sediment additions in lotic and lentic ecosystems(Knowlton and Jones, 1995; Wood and Armitage, 1997; Horppila and Liljendahl-Nurminen, 2005; Matthaei et al., 2006). Negative effects of sedimentation on pri-mary production can be due to reduced light penetration, physical abrasion, andimpairment of substrate for periphyton recruitment (Davies-Colley et al., 1992;Wood and Armitage, 1997). Deleterious effects on benthic invertebrates are gen-erally related to reduced habitat availability, and physical damage of gills and res-piratory surfaces (Wood and Armitage, 1997; Allan, 2004). On the other hand,negative effects on zooplankton biomass can be attributed to decreased feedingrates and ingestion of suspended sediment particles that reduces assimilation effi-ciency of planktonic filter feeders (Hart, 1988; Horppila and Liljendahl-Nurminen,2005; Rellstab and Spaak, 2007).55Contrary to our third prediction, we did not observe significantly lower den-sities of benthic invertebrates in mesocosms treated with imidacloprid. However,imidacloprid did cause significant compositional changes on benthic invertebratecommunities. Imidacloprid additions were associated with reduced densities ofmidge larvae from the subfamily Tanypodinae and lower densities of Procloeon,the most abundant Ephemeroptera genus in our pond ecosystems. These changeswere similar to those reported by previous outdoor experiments and generally agreewith toxicity tests showing that Ephemeroptera, Trichoptera, and Chironomidaespecies are particularly sensitive to neonicotinoids (Morrissey et al., 2015). For ex-ample, Mohr et al. (2012) also reported that imidacloprid pulses in outdoor streammesocosms reduced densities of Tanypodinae, as well as densities of Baetidae, thedominant Ephemeroptera family in their stream mesocosms. Similarly, Colomboand Mohr (2013) reported that imidacloprid reduced densities of Tanypodinae andEphemeroptera in outdoor lentic microcosms.Also contrary to our initial predictions, our results suggested strong negative ef-fects of imidacloprid on zooplankton density. We did not expect lower zooplanktondensities at the concentration levels we tested because our zooplankton communi-ties were mostly comprised by cladoceran crustaceans, which are generally consid-ered tolerant to neonicotinoid insecticides (Sa´nchez-Bayo et al., 2016). For exam-ple, a laboratory bioassay reported median lethal concentrations (LC50) between65000-133000 µg L-1 after 48 h of static exposure for two common cladoceranspecies (Sa´nchez-Bayo and Goka, 2006). We observed significant reductions inzooplankton density, and significant changes in composition after two short pulsesof imidacloprid of approximately 1.97 µg L-1 in our mesocosms. Interestingly, thisis not the first mesocosm study reporting effects on zooplankton at levels that aremuch lower than those suggested by laboratory toxicity tests. Sa´nchez-Bayo et al.(2016) noted in their review paper that mesocosm studies consistently report pop-ulation and community effects of neonicotinoids at concentrations lower than theLC50 of the species under study. In this regard, neonicotinoid insecticides differfrom other pesticides, which regularly show lower toxicity in realistic field sce-narios than in acute toxicity tests (Sa´nchez-Bayo et al., 2016). This discrepancyis attributed to the irreversible neurotoxicological effects of neonicotinoids that re-56sult in impacts that cumulate over time, which amplifies their effects with repeatedexposures (Sa´nchez-Bayo et al., 2016).3.4.2 Imidacloprid’s indirect effects on ecosystem functionThis is one of few studies evaluating ecosystem-scale functional responses to neon-icotinoid insecticides (Pestana et al., 2009; Sa´nchez-Bayo et al., 2016). In agree-ment with our fifth prediction, our results suggest potential indirect effects of im-idacloprid on whole-ecosystem primary production, as the insecticide tended toincrease NEP. However, this result was slightly surprising, because we did notobserve strong effects of imidacloprid on periphyton and phytoplankton biomassseparately. We think this result may be attributed to aggregate indirect impacts ofthe insecticide on all primary producers in the ponds. Disconnections between pat-terns observed in algae biomass and NEP have been reported by previous studies(Young et al., 2008), and highlight the complexity of the interactions between dif-ferent structural components of aquatic ecosystems. More research on impacts ofimidacloprid on net ecosystem productivity is desirable to get a better understand-ing of other potential indirect effects.On the other hand, our results did not support our sixth hypothesis predict-ing significant effects of imidacloprid on leaf decomposition. This prediction wasbased on previous mesocosm and microcosm studies reporting negative indirect ef-fects of the insecticide on leaf decomposition, due to the inhibition of invertebrateshredders (Kreutzweiser et al., 2008; Pestana et al., 2009). However, such studiestested higher concentrations of imidacloprid over longer duration than what wastested in our study. Thus, it is possible that the exposure to imidacloprid evalu-ated in this experiment was not sufficient to cause inhibitory effects on invertebrateshredders. For example, Kreutzweiser et al. (2008) observed inhibition of leaf pro-cessing by invertebrates in microcosms treated with 12 µg L-1 and higher. Anothermesocosm experiment evaluating three pulses of similar concentration (1.63 µgL-1) as our two pulses (1.97 µg L-1) also reported inhibition of leaf litter decompo-sition by invertebrates, but each of their pulses lasted 24 h which likely resulted inhigher exposure of the invertebrates to the insecticide compared to our treatment57(Pestana et al., 2009).3.4.3 Frequent antagonistic interactions between imidacloprid andother agricultural stressorsOur results generally support the hypothesis that imidacloprid can interact withnutrient enrichment and sedimentation at environmentally relevant concentrations.In agreement with our seventh prediction, imidacloprid interacted antagonisticallywith nutrient enrichment. This finding supports previous investigations document-ing antagonistic interactions between moderate concentrations of imidacloprid andhigh levels of nutrient enrichment (Alexander et al., 2013, 2016). For instance,Alexander et al. (2016) observed that the negative effect of lowest observable ef-fects concentrations of imidacloprid (LOEC: 1.39–1.60 µg L-1) on the sensitiveBaetis mayflies was masked by moderate and high levels of nutrient enrichment onstream microcosms. In our experiment, the insecticide independently did not havesignificant negative effects on benthic invertebrate richness, but dampened the pos-itive impacts of nutrient enrichment when the two stressors were combined.Given the tendency of imidacloprid to persist longer in sediment than in water(Van Dijk et al., 2013), our eighth hypothesis predicted synergy between imidaclo-prid and sediment, through increased contact exposure and ingestion of the insec-ticide by zooplankton and benthic invertebrates. However, our observations sug-gest that the combination of the two stressors interacted antagonistically to affectzooplankton communities in the pond mesocosms. We attribute this antagonisticinteraction to the fact that the two stressors tended to affect the same sensitive zoo-plankton genera in the ponds. Thus, their combination caused less taxa losses inthe system than predicted by additivity. This observation can be explained by thecommunity co-tolerance hypothesis proposed by Vinebrooke et al. (2004). This hy-pothesis predicts that when the species’ sensitivities to the stressors are positivelycorrelated, either stressor eliminates certain sensitive species but leaves speciesthat are likely to be tolerant to the second stressor, so there are no further losses inthe system. This response is known as stress-induced community tolerance (Vine-brooke et al., 2004), and has been reported by other multiple stressor studies on58freshwater ecosystems (Gardestro¨m et al., 2015; Louhi et al., 2017).This is the first experimental examination of 3-way interactions between imida-cloprid, nutrient enrichment, and sedimentation. In keeping with results observedfor their 2-way interactions, we observed complex 3-way interactions only on di-versity metrics, and they were all antagonistic at the levels tested. These findingsfurther support the hypothesis of stress-induced community tolerance of pelagicand benthic communities to the combined impacts of sedimentation, nutrient en-richment, and imidacloprid.3.4.4 ImplicationsMesocosm experiments are a useful tool to isolate stressor impacts while incorpo-rating enough ecological complexity to test indirect effects of the stressors, as wellas sufficient replication for rigorous hypothesis testing (Culp et al., 2000; Spivaket al., 2011; Stewart et al., 2013). However, they require the simplification of verycomplex systems, and are conducted in limited temporal and spatial scales, there-fore the extrapolation of our experimental results to field conditions should be donewith care. As an example, in our experiment we did not observe significant effectsof any of the stressors on leaf decomposition, which generally contrasts with anumber of studies reporting direct impacts of nutrient enrichment, and sedimenta-tion, and suggesting potential indirect effects of imidacloprid contamination. Thelack of response of leaf decomposition to either of the treatments could be associ-ated with the season when the experiment was conducted (spring). For instance,a previous pond mesocosm study conducted across several seasons, reported thatthe impacts of nutrient subsidies on decomposition varied with season, with littleeffects during the spring, but strong effects during the summer (Greig et al., 2012).However, despite the normal limitations, our study still offers novel empirical in-formation about the nature of some multiple stressor interactions, and documentspotential whole-ecosystem effects of a popular neonicotinoid insecticide.Most environmental impact assessment frameworks currently employed to man-age anthropogenic stressors on freshwater ecosystems are developed for single59stressors and are largely focused on stressor characteristics (Crain et al., 2008;Segner et al., 2014; Gessner and Tlili, 2016). These frameworks implicitly as-sume that multiple stressors interact additively, so their cumulative impacts on abiological system are predicted by the addition of single stressor effects. Our find-ings support a growing body of literature suggesting that additive frameworks maynot be adequate to predict impacts of multiple stressors on freshwater ecosystems(Shurin et al., 2012; Piggott et al., 2015b; Wagenhoff et al., 2016). Furthermore,we observed that the outcome of the multiple stressor interactions was stronglydependent on the correlation between the species’ tolerance to the individual stres-sors, a finding that fully supports emerging environmental assessment frameworksthat increasingly focus on properties of biological receptors, rather than propertiesof stressors (e.g. Rohr et al., 2006; Segner et al., 2014).Our study also supports the notion that field mesocosm experiments should beused to complement laboratory bioassays, in order to assess risk of toxicants onfreshwater ecosystems (Colombo and Mohr, 2013; Alexander et al., 2016). Notonly did we detect negative effects of imidacloprid on zooplankton at concentra-tions lower than the LC50 of several common species in the community, but we alsoobserved previously unreported effects of the insecticide on ecosystem functions.Furthermore, our results showed the complexity of predicting impacts on ecosys-tem function based on changes observed on ecosystem structure, further supportingprevious investigations suggesting that functional metrics should be implementedin routine ecological assessments, in order to obtain a reliable evaluation of ecosys-tem health (Gessner and Chauvet, 2002; Young et al., 2008; Piggott et al., 2015a).60Table 3.1: ANOVA summary of linear models evaluating impacts of the stressor treatments on water quality and habitatcharacteristics of the freshwater mesocosms. Significant effects (P < 0.05) are indicated in bold.N S I N * S N * I S * I N * S * IResponse variables df F P F P F P F P F P F P F PPO4-P 1, 24 235.125 < 0.001 0.185 0.671 0.400 0.533 0.005 0.944 0.613 0.441 0.877 0.358 1.676 0.208NH4-N 1, 24 1758.047 < 0.001 0.013 0.909 0.604 0.445 0.025 0.875 0.026 0.872 < 0.001 0.986 1.716 0.203NO3-N 1, 24 2.345 0.139 0.685 0.416 5.192 0.032 0.033 0.858 0.017 0.899 0.144 0.707 0.090 0.767DIP:DIN 1, 24 0.115 0.737 0.561 0.461 4.095 0.054 < 0.001 0.992 0.092 0.764 0.028 0.868 0.282 0.600Dissolved oxygen 1, 24 35.519 < 0.001 0.251 0.621 3.789 0.063 0.880 0.357 0.001 0.973 0.481 0.495 3.177 0.087Conductivity 1, 24 1.211 0.282 0.137 0.714 1.234 0.278 0.042 0.839 0.045 0.834 0.208 0.653 0.058 0.812pH 1, 24 36.899 < 0.001 0.012 0.915 9.999 0.004 0.022 0.884 0.062 0.805 0.012 0.915 1.513 0.231Turbidity 1, 24 1.202 0.284 256.721 < 0.001 0.603 0.445 0.921 0.347 0.235 0.632 1.551 0.225 0.008 0.930Sedimentation 1, 24 0.300 0.589 97.404 < 0.001 0.700 0.411 1.167 0.291 0.993 0.329 0.258 0.616 1.848 0.18761Table 3.2: Average imidacloprid concentration measured in all tanks treatedwith the insecticide, and three randomly selected insecticide controls(tanks not treated with the insecticide), on days 1 and 35 of the exper-iment.Measured imidacloprid(µg L-1)Mean (± SE)Treatment Day 1 Day 35Insecticide control n.d. n.d.I 1.985 (0.2) 0.071 (0.01)N*I 2.038 (0.05) 0.094 (0.01)S*I 1.943 (0.06) 0.04 (0.01)N*S*I 1.916 (0.06) 0.034 (0.01)Mean insecticide tanks 1.97 (0.01) 0.06 (0.001)62Table 3.3: ANOVA summary of linear models evaluating impacts of the stressor treatments on habitat, benthic in-vertebrates, zooplankton, and ecosystem function throughout the experiment. Significant effects (P < 0.05) areindicated in bold.N S I N * S N * I S * I N * S * IResponse variables df F P F P F P F P F P F P F PPelagic communityPhytoplankton Chl-a 1, 24 15.223 0.001 0.001 0.981 0.794 0.382 0.047 0.831 0.002 0.962 0.941 0.342 0.152 0.700Zooplankton density 1, 24 2.102 0.160 2.463 0.130 7.668 0.011 0.001 0.970 0.244 0.626 3.040 0.094 0.003 0.959Zooplankton biomass 1, 24 27.188 <0.001 4.961 0.036 0.216 0.647 0.076 0.785 0.627 0.436 0.062 0.806 0.309 0.583Zooplankton richness 1, 24 1.233 0.278 5.842 0.024 0.107 0.747 0.346 0.562 0.960 0.337 5.842 0.024 8.642 0.007Zooplankton Shannon-Wiener diversity 1, 24 2.897 0.102 1.968 0.173 0.031 0.862 0.006 0.938 0.381 0.543 1.646 0.212 13.249 0.001Zooplankton evenness 1, 24 2.604 0.120 0.374 0.547 0.029 0.866 0.073 0.789 0.044 0.835 0.097 0.759 9.469 0.005Benthic communityPeriphyton Chl-a 1, 24 2.032 0.167 0.840 0.369 0.133 0.719 0.257 0.617 0.062 0.806 1.494 0.233 1.341 0.258Abundance 1, 24 0.018 0.894 9.158 0.005 2.539 0.124 3.260 0.083 3.089 0.092 2.300 0.141 0.023 0.880Biomass 1, 24 0.324 0.574 1.303 0.265 0.244 0.626 1.141 0.296 0.475 0.497 0.616 0.440 0.592 0.449Mean size 1, 24 0.128 0.723 1.899 0.181 1.664 0.209 2.678 0.115 0.917 0.348 0.096 0.759 0.047 0.830Abundance small size 1, 24 0.007 0.935 8.398 0.008 1.259 0.273 1.635 0.213 0.228 0.638 0.175 0.679 0.259 0.616Abundance medium size 1, 24 0.153 0.699 7.551 0.011 3.971 0.058 3.562 0.071 3.221 0.085 0.687 0.415 0.032 0.861Abundance large size 1, 24 0.316 0.579 1.052 0.315 0.202 0.657 0.166 0.687 0.418 0.524 5.244 0.031 0.123 0.728Richness 1, 24 9.328 0.005 3.358 0.079 0.731 0.401 1.209 0.282 5.388 0.029 0.373 0.547 0.134 0.717Shannon-Wiener diversity 1, 24 1.496 0.233 1.386 0.251 0.142 0.709 0.166 0.687 0.0001 0.994 0.064 0.802 0.756 0.393Evenness 1, 24 0.372 0.548 0.087 0.770 0.019 0.892 1.636 0.213 2.178 0.153 0.530 0.474 0.282 0.600Ecosystem functionLeaf decomposition 1, 24 0.241 0.628 1.100 0.305 0.085 0.773 1.100 0.305 0.914 0.349 0.204 0.656 0.722 0.404Net ecosystem production 1, 24 68.869 <0.001 8.835 0.007 3.462 0.075 0.058 0.811 0.035 0.853 0.034 0.855 0.911 0.349Biochemical oxygen demand 1, 24 9.765 0.005 1.367 0.254 0.003 0.957 0.180 0.676 2.997 0.096 0.752 0.394 0.474 0.49863Table 3.4: Overview of hypotheses and study resultsHypothesis Stressor Biological response Hypothesizedeffect/interaction Result1 NutrientsLeaf decomposition, primaryproduction, invertebrate biomassPositiveAccepted for pelagic compartment,not accepted for benthic2 SedimentMost community metrics andecosystem processesNegative Accepted3 Imidacloprid Benthic invertebrate density NegativeNot accepted, but observed significantchanges in composition4 Imidacloprid Zooplankton density No effectNot accepted, observed significant reductionsat lower concentrations than expected5 Imidacloprid Net ecosystem productivity Positive Marginally significant positive effects6 Imidacloprid Leaf decomposition Negative Not accepted7 Imidacloprid + NutrientsPelagic and benthic invertebratemetricsAntagonistic Accepted8 Imidacloprid + SedimentPelagic and benthic invertebratemetricsSynergisticNot accepted, observed mostlyantagonistic interactions64Figure 3.1: Photo of the experimental set-up consisting in 32 outdoor fresh-water mesocosms located at the University of British Columbia’s exper-imental pond facility, Vancouver, Canada.65Figure 3.2: Temporal dynamics of phytoplankton biomass (a, b), zooplank-ton biomass (c, d), and net ecosystem production (e, f) in freshwatermesocosms exposed to nutrient, sediment, and insecticide pulses. Eachpoint represents the mean (± SE) for each treatment (n = 4). Lettersare used to represent stressor treatments (N, nutrients; S, sediment;I, insecticide, N+S, nutrient + sediment; N+I, nutrient + insecticide;S+I, sediment + insecticide; N+S+I; nutrient + sediment + insecticide).Dashed vertical lines indicate the dates for nutrient and insecticide ad-ditions, while dotted vertical lines indicate sediment additions. Pointsare slightly jittered to ease interpretability66Figure 3.3: Standardized effect size (Hedge’s d ± 95% CI) for significantstressor main and interactive effects on zooplankton communities. Let-ters in the y-axis represent main effects (N, nutrients; S, sediment; I,insecticide) and interaction terms (N*S, nutrient x sediment; N*I, nu-trient x insecticide; S*I, sediment x insecticide; N*S*I; nutrient x sedi-ment x insecticide). For main effects positive values indicate increasesin the response variable whereas negative values indicate the opposite.For interactions, confidence intervals overlapping zero indicate addi-tive interactions, positive values indicate synergies, and negative valuesare antagonisms or reversals (reversals are marked with an R). Symbolsare used to represent significance of the effect according to the linearmixed effects models presented in Table 3.3: ***P< 0.001, **P< 0.01,*P < 0.05, ·P < 0.1.67Figure 3.4: Bar plots illustrating significant interactive effects of the stressorson zooplankton (a-c) and benthic invertebrate (d-e) metrics. Letter no-tation for the treatments is consistent with Figure 3.2. Dotted horizontallines are used to represent the predicted additive effect of the stressors.All shown interactions were classified as antagonistic, except the S*Iinteraction on abundance of large-sized benthic invertebrates (d) whichwas a reversal. Bars represent the mean of four replicates (± SE).68a.b.Figure 3.5: Redundancy analysis plot showing the effects of the stressorpulses on zooplankton (a) and benthic invertebrate (b) community com-position by the end of the experiment. Only the most abundant taxa arelabeled. Solid arrows indicate significant main effects (P < 0.05) anddashed arrows indicate marginally significant effects (P < 0.1), follow-ing a permutation test of the full model with 999 randomizations. Letternotation for the treatments is consistent with Figure 3.3. The treatmentstogether explained 35% of the variation on zooplankton compositionand 33% of variation on benthic invertebrate composition.69Figure 3.6: Standardized effect size (Hedge’s d ± 95% CI) for significantstressor main and interactive effects on benthic invertebrate communi-ties. Letter notation for the treatments and interpretation of effects isconsistent with Figure 3.3. Symbols are used to represent significanceof the effect according to the linear fixed effects models presented inTable 3.3: **P < 0.01, *P < 0.05, ·P < 0.1.70Figure 3.7: Redundancy analysis plot showing the effects of the stressorpulses on the distribution of biomass among food web components inthe mesocosms by the end of the experiment (day 36). Solid arrowsindicate significant main effects (P < 0.01), following a permutationtest of the full model with 999 randomizations. Letter notation for thetreatments is consistent with Figure 3.3. The treatments together ex-plained 27% of the variation on biomass (a). Barplot comparing aver-age biomass (n= 16) of food web components among nutrient treatmentlevels (b).71Figure 3.8: Standardized effect size (Hedge’s d ± 95% CI) for significantstressor main and interactive effects on ecosystem function. Letter no-tation for the treatments and interpretation of effects is consistent withFigure 3.3. Symbols are used to represent significance of the effectaccording to the linear mixed effects models presented in Table 3.3:***P < 0.001, **P < 0.01, *P < 0.05, ·P < 0.1.72Chapter 4Multiple-stressor interactions intributaries alter downstreamecosystems in stream mesocosmnetworks4.1 IntroductionRiver systems are dendritic networks in which contributing streams join into thetributaries of the main river. This pattern follows a hierarchical configuration, in-creasing in size and decreasing in number in the downstream direction (Bendaet al., 2004; Campbell Grant et al., 2007). Thus, every large river basin is composedof nested subcatchments that have a longitudinal connection through the unidirec-tional flow of water (Fisher, 1997; Allan and Castillo, 2007; Campbell Grant et al.,2007). The connectivity imposed by the unidirectional flow and the spatial ar-rangement of river networks, greatly influences ecological dynamics, as subsidiesof energy, nutrients, and organic materials are constantly transported across inter-connected stream reaches (Gomi et al., 2002; Campbell Grant et al., 2007; Brown73et al., 2011). This interconnected set of habitats are the reason why river networksare considered meta-ecosystems (Loreau et al., 2003; Brown et al., 2011). In rivermeta-ecosystems, local community composition, habitat condition, and ecosystemservices cannot be understood by focusing solely on local-scale processes, becausethey are also influenced by processes operating in upstream reaches at the regionalscale (Campbell Grant et al., 2007; Brown et al., 2011; Tomscha et al., 2017).Land-use practices can dramatically alter the input of materials into headwaterstreams, subsequently affecting subsidies to downstream ecosystems (Gomi et al.,2002; Wipfli et al., 2007). For instance, agricultural practices often increase inputsof fine sediment and nutrients into stream ecosystems (Allan, 2004; Riseng et al.,2011). Numerous studies have demonstrated that nutrient enrichment generally in-creases standing stocks and processing rates of basal resources in aquatic ecosys-tems, causing bottom-up effects in aquatic food webs at moderate levels of enrich-ment, and negative effects at higher levels (Allan, 2004; Woodward et al., 2012b;Wagenhoff et al., 2013). On the other hand, fine sediment inputs are reported tohave strong negative effects on stream communities, reducing primary productiondue to increased turbidity, smothering, and abrasion, filling interstitial spaces forbenthic invertebrates, and slowing down organic matter processing (Wood and Ar-mitage, 1997; Matthaei et al., 2006). Furthermore, recent experimental investiga-tions have demonstrated that sedimentation and nutrient enrichment often interactsynergistically, causing cumulative effects that are more detrimental than predictedon the basis of the stressors’ individual effects (Matthaei et al., 2010; Piggott et al.,2012; Wagenhoff et al., 2012).The degradation of headwater watersheds with these and other ecological stres-sors may cause alterations to downstream hydrology, water quality, geomorphicprocesses, and even biota, as food resource subsidies coming from headwater ecosys-tems are disrupted (Freeman et al., 2007). The transmission of disturbances (e.g.,fine sediment, nutrients, and other contaminants) from headwaters to downstreamecosystems is governed by a complex array of routing processes taking place at thescale of headwater systems (storage, transformation, and disturbance mechanism),and network systems (synchronized or desynchronized outflows from headwaterbasins, basin size, basin shape, drainage density, and network geometry) (Gomi74et al., 2002; Wipfli et al., 2007). Empirical evidence regarding the interaction ofprocesses happening at multiple scales is necessary to develop predictive under-standing about the consequences of cumulative headwater degradation. However,studies addressing cumulative effects on river systems often focus on specific sec-tions of the river network, either evaluating responses at the headwater scale orat the network scale (Rasmussen et al., 2012; Schneider et al., 2013). To datefew studies explicitly integrate both scales to measure how processes in individualheadwater systems interact to affect inputs of material and function on downstreamecosystems (but see Patrick and Fernandez, 2013).Here we present the results of a mesocosm experiment designed to study howinteractions among multiple stressors within tributaries may affect downstreamecosystem function, diversity, and physical habitat. Using a mesocosm model ofa stream network, we manipulated sediment and nutrient levels in the tributariesof second-order channels, to determine individual and combined effects of distur-bances on tributaries and recipient ecosystems. We chose this stressor combinationas they offered a good model to study cumulative effects of headwater degrada-tion at the network scale. First, as mentioned earlier, there is already empiricalknowledge about their potential non-additive interactions on stream ecosystems(e.g., Matthaei et al., 2010; Piggott et al., 2012; Wagenhoff et al., 2012). Second,they may have different rates of delivery from headwaters to downstream ecosys-tems, due to different storage and transformation mechanisms within headwaters(Bernhardt et al., 2005; Wipfli et al., 2007).Our treatments were designed to test: i) individual and combined effects ofnutrient enrichment and sedimentation on ecological structure and function of trib-utary streams; ii) the potential effect of complex stressor interactions within thetributaries on recipient second-order channels; and iii) how the level of distur-bance within the tributaries affects ecological function and structure of recipientdownstream ecosystems. Our overarching hypothesis was that stressor additionsin tributary streams would have detectable effects on the structure and function ofdownstream recipient ecosystems. Specifically, for tributary channels we hypoth-esized that: 1) nutrient additions would have positive effects on most biologicalresponses; 2) sediment additions would have negative effects on most biological75responses; and 3) there would be complex non-additive interactions between nu-trient enrichment and sedimentation. For second-order channels we hypothesizedthat: 4) complex multiple-stressor interactions within the tributaries would influ-ence responses of recipient downstream ecosystems; and that 5) increasing levelsof disturbance in the tributaries would cause proportional increases of disturbanceon downstream ecosystems.4.2 Methods4.2.1 Experimental designWe built a network of 36 stream mesocosm channels in the Malcolm Knapp Re-search Forest of the University of British Columbia, near Maple Ridge, BritishColumbia, Canada (49◦ 16 N, 122◦ 34 W; Figure 4.1). In the design, 24 meso-cosms, which will be referred to as “first-order” channels, converged downstreamin pairs to form 12 “second-order” channels (channel dimensions: 6.8 m x 0.15 m,Figure 4.2). All channels were stocked with a 4 cm layer of washed, mixed grav-els (0.5 - 3.0 cm grain size range) and were continuously gravity-fed with waterfrom Mayfly Creek, an adjacent oligotrophic stream which is described in detailin Richardson (1991). Water from Mayfly Creek was distributed from two headerboxes with outflow valves that controlled water flow, set to about 1 L s-1 to eachchannel throughout the experiment. Average slope was 0.06 m m-1 for the first-order channels and 0.03 m m-1 for the second order channels. Four weeks priorthe beginning of the experiment, we allowed immigration of invertebrates fromMayfly Creek to colonize the channels via drift. As this method has proven ap-propriate to obtain consistent invertebrate densities in previous experiments at thesame location (Lecerf and Richardson, 2011; Richardson, 1991), we did not collectpremanipulation samples for this study.Our 22-day experimental manipulation was conducted in the fall season, fromNovember 7 to November 28 of 2013. We manipulated fine sediment and nutrientsin the 24 first-order channels on day 1 of the experiment. We evaluated two levels76of each stressor (added, ambient) in the following combinations (Figure 4.2): i)ambient levels of sediment and nutrients (control treatment, n = 12); ii) added nu-trients and ambient levels of sediment (N treatment, n = 4); iii) added sediment andambient levels of nutrients (S treatment, n = 4); and iv) added nutrients and addedsediment (N+S treatment, n = 12). High levels of deposited sediment (approxi-mately 60% stream bed coverage) were achieved by adding 3 L of sand (< 0.5 mm,“medium sand” on the Wentworth scale) as evenly as possible to each sediment-addition first-order channel. These sedimentation levels are equivalent to valuesreported in rivers affected by agricultural practices (Townsend et al., 2008; Wagen-hoff et al., 2012), and are similar to those used in several experiments evaluating theeffects of sedimentation on stream ecosystems (e.g., Matthaei et al., 2010; Louhiet al., 2017). We used 4-month, slow-release fertilizer pellets (Florikote, NPK:15-5-15), to achieve continuous nutrient enrichment throughout the experiment innutrient addition channels. We added 14 g of Florikote to each treated channel fora target phosphorus concentration of 3 µg L-1 above background nutrient levels(approximate background concentrations: 3.7 µg P-PO4 L-1, 123.4 µg dissolvedinorganic nitrogen L-1, and 74.5 N:P ratio; Garcı´a et al., 2017). This enrichmentlevel is equivalent to nutrient additions that have been previously reported to causesignificant responses in experimental channels fed by Mayfly Creek (Kiffney andRichardson, 2001).Treatments on first-order channels were assigned so each second-order channelwas exposed to one of the following four tributary treatments (Figure 4.2): a) twocontrol tributaries; b) one control tributary and one tributary with both stressors;c) nutrients added in one tributary and sediment in the other; and d) nutrients andsediment simultaneously added in both tributaries. Each tributary treatment hadthree replicates for a total of 12 second-order channels.4.2.2 Response variablesWe measured variables reflecting the effect of the treatments on habitat condition(sedimentation rates, water nutrient concentrations), benthic invertebrate commu-nities (density, functional feeding groups, drift rates), and ecosystem function (leaf77decomposition) in first- and second-order channels. Water samples were collectedon day 22 to determine nutrient concentrations (PO4-P, NH4-N, and NO3-N) inthe second-order channels, and were analyzed by Maxxam Analytics, Burnaby,British Columbia, using standard methods (APHA, 2005). To quantify sedimen-tation rates in the channels, we collected all the substrate present in three randomquadrants (0.1 m x 0.8 m) of each first- and second-order channel. Substrate col-lection was carried out using a small D-net built to fit in the channels. We placedthe d-net downstream from the sample quadrant, and disturbed the area within thequadrant until all the substrate was collected in the d-net (64 µm mesh size). Thecollected substrate was filtered through a 2 mm sieve stacked on top of a 0.5 mmsieve to separate the gravel. The filtered sand was stored in sealed plastic bucketsand transported to the laboratory. In the laboratory the contents of each bucketwere oven-dried at 40◦C for 10 days, weighed, ashed at 500◦C for three hours, andreweighed to quantify ash-free dry mass (AFDM).We collected invertebrate drift from all the experimental channels twice, oncenear the beginning of the experiment (day 2) and once at the end of the experiment(day 22). Drift samples were collected by placing 250 µm mesh nets at the endof each channel. Second-order channels were sampled for 24 h periods, whereasfirst-order channels were sampled for 2 h periods (10 am to 12 pm) to avoid dis-rupting the transport of materials to second-order channels. All drift samples werepreserved in 80% ethanol and transported to the laboratory. We sampled benthic in-vertebrate communities in all the experimental channels once, at the end of the ex-periment (day 22). One composite invertebrate sample was collected in each chan-nel using a small Surber sampler (0.017 m-2, 250 µm mesh size) in four randomlocations (total sampled area in each channel: 0.068 m2). Composite Surber sam-ples were stored in 80% ethanol and sorted later under the dissecting microscopefor invertebrates. All benthic invertebrates were enumerated, identified to the low-est practical taxonomic level (usually genus), and classified into functional feedinggroups according to Merritt and Cummins (1996). We computed 11 invertebratevariables for each first- and second-order channel: 1) total invertebrate density;2) total taxa richness; 3) Shannon-Wiener diversity; 4) density of Ephemeroptera,Plecoptera, and Trichoptera (EPT) taxa; 5) richness of EPT taxa; 6) density of78scrapers; 7) shredders; 8) predators; and 9) collectors; 10) drift rate (total numberof individuals drifting per hour); and 11) EPT drift flux (number of EPT driftingper sampling period).To measure leaf decomposition, we introduced one leaf pack of red alder leavesin coarse-mesh bags (3 g air-dry weight; 10 mm mesh) into each channel one daybefore treatment application (day 0). On day 22, leaf packs were removed from thechannels and stored at -18◦C until processed in the laboratory to estimate remain-ing AFDM, according to standard methods (Hauer and Lamberti, 2007). Posteriorprocessing involved defrosting, rinsing, and drying at 60◦C for 5 days. Remainingleaf material was then weighed, ashed at 500◦C, and reweighed to calculate AFDM.4.2.3 Data analysisIn order to test our first three hypotheses, we used linear fixed-effects models toquantify individual and combined effects of the stressors on response variablesrecorded in the first-order channels. For each response variable the model testedwas: y = b0 + b1N + b2S + b3N ∗ S, where N is the nutrient treatment and S thesediment treatment. Significance levels for all our tests was P < 0.05 and wasevaluated with ANOVAs. Following the recommendation of Nakagawa and Cuthill(2007) we present standardized effect size estimates for all findings with P < 0.1,so readers can judge the biological importance of the results. Hedges d estimatesof effect size (Gurevitch and Hedges, 2006) were calculated from the t values ofour linear models using the formulas provided by Nakagawa and Cuthill (2007).In order to improve the graphical representation of our results, we coded signifi-cant main effects to represent the direction of the response of manipulated versuscontrol mesocosms (i.e. positive effect sizes indicate increases in the response vari-able, while negative effect sizes indicate the opposite). Further, we coded 2-wayinteraction effect sizes to represent the classification of the interaction according tothe framework proposed by Jackson et al. (2016). Thus, positive interaction effectsizes represent synergistic interactions (i.e. the combined effect of the stressorsis greater than the sum of their individual effects), whereas negative effect sizes79represent either antagonistic interactions (i.e. the combined effect of the stressorsis less than the sum of their individual effects) or reversal interactions (i.e. thecombined effect of the stressors is in the opposite direction than the sum of theindividual effects). Given that the counts of EPT taxa drifting out of the first-orderchannels were low and contained many zeros, we could not apply a linear modelto this variable. Instead, we used a zero-inflated Poisson (ZIP) regression. TheZIP model has two components: a Poisson model with log link (count model) thatevaluates the effect of the explanatory variables on the counts; and a negative bino-mial with logit link model (zero model) that evaluates the effect of the explanatoryvariables on the probability of zero counts in the data.Data recorded in the second-order channels were divided into two subsets totest our fourth and fifth hypotheses. For our fourth hypothesis, we used data corre-sponding to tributary treatments b and c (Figure 4.2) and compared the two treat-ments using independent t-tests. Thus, we compared whether applying both nutri-ents and sediment in the same tributary (cumulative effect; tributary treatment b)had the same effect downstream as applying nutrients in one tributary and sedimentin the other tributary (single effect: tributary treatment c). For our fifth hypothesis,we used data corresponding to tributary treatments a, b, and d to evaluate the effectof tributary disturbance level on recipient second-order channels (Figure 4.2). Weexcluded tributary treatment c from this analysis in order to avoid potential con-founding effects of multiple-stressor interactions within the tributaries. This waywe ensured we were able to test a linear increase in tributary disturbance. We usedsingle linear regressions of the form: y = b0 +b1x1, where x1 was a 3-level factorrepresenting the degree of disturbance in the first-order tributaries (i.e. the numberof tributaries treated with the two stressors). The three levels included: 0) a controlwith no stressors in the tributaries; 1) only one tributary with both stressors; and2) two tributaries with both stressors (tributary treatments a, b, and d in Figure 4.2,respectively).804.3 Results4.3.1 Stressor effects on first-order channelsThe sediment treatment caused significantly higher sedimentation rates in the first-order channels (Table 4.1). By the end of the experiment, first-order channelstreated with sediment had on average 61 times more sediment accumulated in thesubstrate (1821.7± 312.8 g m-2) than untreated channels (29.8± 13.8 g m-2). Sed-iment additions also increased the absolute amount of organic sediment deposited(sediment AFDM) by 2.6 times (Table 4.1). However, on average, sediment de-posited in these channels had significantly lower organic content (organic contentsediment treatment: 0.6% ± 0.15%, mean ±SE) than sediment in the remainingfirst-order channels (organic content: 30% ± 6%, Table 4.1).We detected significant negative effects of the sediment treatment on total in-vertebrate density in the first-order channels (Table 4.1). Channels treated withsediment had on average 34% fewer individuals than channels without sediment(Figure 4.3a). According to our findings, this reduction was associated with nega-tive impacts on invertebrate predators (P = 0.005, 55% reduction) and, to a lesserextent, collectors (P = 0.07, 46% reduction). In terms of ecosystem functioning,we observed a significant nutrient x sediment interaction affecting leaf decompo-sition in the first-order channels (Figure 4.4). This interaction was classified as anantagonism; while both stressors tended to have positive effects on leaf decompo-sition independently (the individual effect of sediment was significant, while theindividual effect of nutrients was not), nutrient x sediment channels had similardecomposition rates as the control treatments (Figure 4.4). In other words, leafpacks in nutrient x sediment channels lost in average 25% less ash free dry massthan expected if the effect of the two stressors was additive.Nutrient and sediment additions did not significantly impact the total number ofinvertebrates drifting per hour on days 2 and 22 after treatment application. How-ever, we observed significant effects of both treatments on drift flux of individualsfrom the orders Ephemeroptera, Plecoptera, and Trichoptera on day 22 (Table 4.2).81According to the zero-inflated Poisson regression, nutrient and sediment additionsdid not influence the probability of excess zeros on the 2-hour drift samples (Fig-ure 4.5a), but had strong effects on the counts of EPT drifting from the channels(Figure 4.5b). Nutrient additions had a negative effect on EPT drift flux, as therewere 90% fewer individuals from these orders drifting from channels treated withnutrients. On the other hand, sediment had the opposite effect; sediment-channelshad twice as many EPT drifting as the other first-order channels (91% increase indrift flux).4.3.2 Stressor effects on second-order channelsThe t-tests used to compare the impact of applying both stressors in the same tribu-tary versus the impact of applying them in different tributaries (hypothesis 4) indi-cated that there were no significant differences between sediment deposition ratesand nitrate concentration between the two tributary treatments (Table 4.3). Unfor-tunately, phosphorous concentrations were under detection levels (5 µg L-1) in allwater samples, so we could not determine whether there were significant differ-ences for this element. However, we observed that EPT density and EPT richnesswere significantly different among the two treatments. There was on average 38%higher density and 45% higher richness of EPT taxa in second-order channels fedby tributaries where the stressors were applied separately (single treatment), whencompared to channels fed by tributaries where the stressors were applied simulta-neously (cumulative treatment).Linear models evaluating response variables in the second-order channels as afunction of disturbance level in the tributaries (hypothesis 5, Table 4.4), indicatedthat sediment dry mass (total sediment, 4.6a) and sediment ash-free dry mass (or-ganic sediment, 4.6b) significantly increased with the level of disturbance to thetributaries. According to Tukey’s HSD post-hoc test, total and organic sedimentdeposition in the maximum level of disturbance (level 2) were significantly higherthan sediment deposition in the control and the first level of disturbance (Figure4.6a-b). However, even at the highest level of tributary disturbance, average sedi-ment deposition in second order channels (351.6 ± 93.84 g m-2, mean ± SD) was82much lower than deposition in first order tributaries treated with sediment (1821.7± 1083.8 g m-2). On the other hand, we did not detect a significant relationshipbetween tributary level of disturbance and the concentration of nitrate in second-order channels (Table 4.4). In terms of the biological responses, the only variablethat significantly responded to the tributary disturbance was EPT density (Table4.4). According to our observations, EPT density tended to increase with increas-ing level of disturbance (Figure 4.6c). Tukey’s HSD tests indicated that EPT den-sity in the highest level of tributary disturbance (both tributaries treated with bothstressors) was significantly higher than EPT density in the control second-orderchannels (both tributaries with natural levels of nutrients and sediment).4.4 DiscussionOur results generally supported the hypothesis that upstream disturbance can in-fluence ecological function and structure of downstream recipient ecosystems ina stream network. However, most of the downstream effects we observed in ourstudy did not support our initial predictions. Due to the small scale of our experi-mental stream network, our hypotheses were based on the assumption that tributarytreatments would affect downstream ecosystems mostly through the transmissionof disturbances, or the movement of sediment and nutrients from tributaries tosecond-order channels. Counter to this assumption, most downstream impacts inour study were a result of within network dispersal of sensitive taxa, as a responseto stressor additions in the tributaries.4.4.1 Stressors altered invertebrate communities and ecosystemfunction in first-order channelsIn our first hypothesis, we predicted that nutrient additions would enhance pri-mary production and organic matter decomposition in our experimental systems,exerting positive bottom-up effects on invertebrate communities (Woodward et al.,2012b; Rosemond et al., 2015). However, we did not detect significant individ-ual effects of the nutrient additions on invertebrate density or leaf decomposition83rates. The only variable that was significantly affected by nutrient additions wasEPT drift flux, which was strongly depressed by the enrichment. This observationmay indicate there was indeed an increase in food availability in nutrient-enrichedchannels, which resulted in lower drift of EPT taxa relative to control channels.Active drift is a known mechanism for patch selection of some EPT taxa withhigh behavioural drift tendency (Naman et al., 2016). For example, previous ex-perimental manipulations of resource availability have found inverse relationshipsbetween food availability and active drift (Hammock and Wetzel, 2013). Further-more, O’Callaghan et al. (2015) also observed strong nutrient effects on drift pat-terns, even in the absence of strong effects on invertebrate densities within theexperimental units. They attributed this discrepancy to the duration of their ex-perimental manipulations (28 days), arguing their experiment may have been tooshort to detect long-term changes in community abundance, or shifts in competi-tive interactions due to changes in food availability (O’Callaghan et al., 2015), aconsideration that likely applies to our 22-day manipulation.Partially in support of our second hypothesis, sediment additions had strongnegative effects on benthic invertebrate densities, while also increasing EPT driftflux from our first-order channels. These findings largely support most publishedliterature reporting deleterious effects of sediment on benthic invertebrate commu-nities (e.g., Matthaei et al., 2006; O’Callaghan et al., 2015; Piggott et al., 2015b).Increased inputs of fine sediment have been reported to fill interstitial spaces instream ecosystems, reducing habitat availability for benthic invertebrates, and caus-ing direct negative effects on sensitive species due to coating of gills and respiratorysurfaces (Wood and Armitage, 1997; Allan, 2004; Wagenhoff et al., 2012). More-over, sediment additions often induce behavioural drift as a response to impairedhabitat quality (Connolly and Pearson, 2007; O’Callaghan et al., 2015; Wagenhoffet al., 2012; Piggott et al., 2015b). Specifically, the EPT orders contain severaltaxa known to respond to sedimentation by drifting short distances in order to findbetter habitat patches (O’Callaghan et al., 2015; Naman et al., 2016).On the other hand, our second hypothesis was not supported regarding the ef-fects of sedimentation on leaf decomposition. Instead of reducing leaf processingin the streams, as has been generally reported in the literature (Young et al., 2008;84Tank et al., 2010; Danger et al., 2012; Louhi et al., 2017), sediment additions inisolation actually increased leaf decomposition in first-order channels. This unex-pected positive response has been previously reported in a few experimental ma-nipulations of sediment in stream ecosystems (e.g., Matthaei et al., 2010; Piggottet al., 2012, 2015a), and might be due to increased anaerobic respiration of leafmaterial buried in sediment (Piggott et al., 2015a).Leaf decomposition was the only response that supported our third hypothe-sis predicting non-additive interactions among nutrients and sediment in first-orderchannels. Leaf decomposition showed clear antagonistic effects of the two stres-sors; while in isolation both stressors tended to increase decomposition, in combi-nation their effect was completely inhibited, resulting in values similar to those ofthe control treatments. Similar antagonistic nutrient x sediment interactions werereported for measures associated with leaf processing by Piggott et al. (2015a).They observed sediment additions dampened the positive effect of the nutrient en-richment on leaf respiration and cotton tensile strength loss in experimental streammesocosms. Piggott et al. (2015a) suggested that microbial communities respon-sible for increased decomposition rates under high nutrient concentrations, hadless access to nutrients and oxygen in the water column when sediment was ap-plied (Piggott et al., 2015a). However, in our case the presence of nutrients alsodampened the positive effects of sediment, which suggests that nutrient additionsinhibited anaerobic microbial respiration in buried leaf packs, a puzzling result thatdeserves further study.4.4.2 Tributaries influenced downstream ecosystems throughdispersal of sensitive taxaPartially in agreement with our fourth hypothesis, we detected significant differ-ences between cumulative- and single-stressor tributary treatments. However, toour surprise these differences were observed for variables without significant nu-trient x sediment interactions within the tributaries: EPT density and EPT richness.Both EPT metrics were higher in downstream ecosystems with single-stressor trib-utaries. Because the treatments did not affect EPT density or richness within tribu-85taries, we attribute this result to their opposing effects on EPT drift from tributariesto second-order channels. As mentioned earlier, sediment additions augmented thenumber of EPT drifting out of first-order channels, nutrient additions decreased it,and the combination of both treatments resulted in additive effects on EPT flux.However, higher EPT density in second-order channels with individual-stressortributaries seems to suggest that the combination of the two stressors within thesame tributary reduced EPT drift further than the addition of the stressors in sep-arate tributaries. The fact that such an interactive effect was not detected in driftresponses, may be an artifact of the timing and duration of our first-order driftsamples. In an attempt to not disrupt subsidies from tributaries to downstreamchannels, we collected only two-hour samples during daytime. Previous researchsuggests that this limited sampling window may not have captured complex inter-active effects of the treatments on drift behaviour. For example, in an experimentalmanipulation of flow carried out at the same experimental facility, Naman et al.(2017) found that drift responses to their treatments were stronger at night-time,indicating that time of the day may impact the likelihood of reliably capturing driftresponses to stressors in this particular system.Our fifth hypothesis, predicting that disturbance in second-order channels wouldincrease proportionally with disturbance in their tributaries, was partially supportedfor sedimentation rates in second-order channels. However, in contrast to our ex-pectations, most biological responses in second-order channels did not show in-creasing negative impacts with this trend. On the contrary, the density of EPTtaxa, a group generally considered sensitive to organic contamination and habi-tat degradation, increased proportionally with tributary disturbance. We againattribute this paradoxical result to the impacts of the stressors on the flux of or-ganisms from tributaries to downstream channels. Higher disturbance within thetributaries caused higher EPT drift, likely increasing immigration rates in down-stream ecosystems. Even though downstream sedimentation increased with trib-utary disturbance, there was less sediment deposited per unit area in downstreamecosystems than their first-order tributaries. Thus, individuals emigrating from thetributaries may have found in downstream channels more suitable habitat patchesthan their source patch. These findings generally fit the mass effects model of the86metacommunity framework. According to this model, even though local habitatconditions and species interactions are important in shaping community composi-tion, high rates of dispersal may override local effects and allow species to persistin unfavourable conditions (Brown et al., 2011). This mass effect has been rec-ognized as a particularly important force in stream networks, where downstreammovement of material and individuals from the tributaries may have a dispropor-tionate effect on downstream ecosystems (Campbell Grant et al., 2007).4.4.3 ImplicationsWe used a simplified model of a stream network to link upstream disturbance witheffects on downstream ecosystems. However, we do not contend that a short-termexperiment in a small channel network can encompass all processes occurring inreal river networks. Our experiment was not realistic in terms of spatial and tem-poral scales, network complexity, and material inputs from colluvial processes intodownstream ecosystems. In spite of these limitations, we argue that some inter-esting patterns and small-scale processes, such as short-distance dispersal of or-ganisms, dilution, and transfer of materials, could be reliably measured in our ex-perimental channel network. Thus, with some caveats in mind, our study offersinteresting insights about the potential impacts of multiple stressor interactions onmeta-ecosystem dynamics of river networks.To our knowledge, this is the first experiment explicitly linking multiple stres-sor effects in tributaries to effects in downstream recipient ecosystems in a rivernetwork. Thus, we had little empirical information to compare our results with.However, our observations generally support previous research highlighting the po-tential role of spatial species interactions within tributaries on downstream ecosys-tem function. For instance, Patrick and Fernandez (2013) observed that the spatialdistribution of shredder invertebrate species among tributaries, had the potential toregulate particulate organic matter exports from headwaters to downstream ecosys-tems, due to competitive interactions.Our results showed that stressor additions in tributaries can strongly influence87ecological function and structure of downstream ecosystems. In our experimentalsystem, most of these effects were due to impacts on dispersal patterns of sensitivetaxa, underscoring the importance of metacommunity frameworks to understandhow disturbances at the scale of the tributaries may influence population dynamicsin downstream ecosystems (Brown et al., 2011; Campbell Grant et al., 2007).Additionally, our observations in first-order channels further support previousstudies reporting that deposited fine sediment is a predominant determinant of in-vertebrate community dynamics and ecological functioning in stream ecosystems(Louhi et al., 2017; Piggott et al., 2015b). As such, our results suggest that manage-ment measures that reduce the input of fine sediment into headwater streams shouldbe prioritized to restore functional and structural integrity of stream ecosystems.88Table 4.1: ANOVA summary of linear models evaluating impacts of the stres-sor treatments on response variables of the first-order channels. Signifi-cant effects (P < 0.05) are indicated in bold.N S N*SResponse variables df F P F P F PHabitat variablesSediment dry mass 1,21 0.592 0.450 114.206 <0.0001 1.061 0.315Sediment AFDM 1,21 2.218 0.151 11.349 0.003 2.284 0.146Sediment % organic 1,21 0.984 0.333 19.115 <0.0001 0.823 0.375Benthic invertebratesTotal density 1,18 0.002 0.966 5.975 0.025 0.000 0.988Total richness 1,18 0.240 0.630 0.004 0.953 2.452 0.135EPT density 1,18 0.320 0.578 1.343 0.262 0.576 0.458EPT richness 1,18 0.110 0.744 0.504 0.487 2.440 0.136Scraper density 1,18 1.483 0.239 0.007 0.934 1.144 0.299Shredder density 1,18 0.298 0.592 1.270 0.275 0.381 0.545Predator density 1,18 0.062 0.806 10.492 0.005 0.016 0.902Collector density 1,18 0.035 0.854 3.585 0.074 0.065 0.801drift rate day 2 1,18 0.033 0.857 0.828 0.374 0.132 0.720drift rate day 22 1,18 1.769 0.205 0.087 0.773 0.023 0.881Ecosystem functionLeaf decomposition 1,18 2.351 0.141 2.185 0.155 8.520 0.00889Table 4.2: Parameter estimates of the zero-inflated Poisson regression ex-plaining EPT drift flux from first-order channels on day 22. The modelincludes two components: a count model (Poisson with log link), anda zero-inflation model (binomial with logit link). Significant effects(P < 0.05) are indicated in bold.Count model Zero-inflation modelFactor Estimate SE z P Estimate SE z PN 1.735 0.307 5.650 < 0.0001 3.302 154.596 0.021 0.983S -0.619 0.307 -2.060 0.044 -2.689 154.596 -0.017 0.986N*S -0.015 0.307 -0.034 0.973 2.969 154.596 0.019 0.98590Table 4.3: Summary of independent t-test’s to compare response variables insecond-order channels fed by tributaries where the nutrients and sedi-ment were applied in combination (cumulative; tributary treatment b inFigure 4.2) and second-order channels fed by tributaries where the stres-sors were applied separately (single; tributary treatment c in Figure 4.2).Significant effects (P < 0.05) are indicated in bold.Cumulative Single t-testResponse variable Mean SD Mean SD t PHabitat variablesSediment dry mass (g m-2) 133.53 113.53 502.79 249.26 -2.4 0.076Sediment AFDM (g m-2) 16.18 3.68 23.82 15.29 -0.6 0.659Nitrate (µg L-1) 76.5 0.5 80.5 0.5 -1.1 0.446Benthic invertebratesTotal density (ind m-2) 1083.3 770.1 823.6 420.3 0.5 0.642Total richness (taxa per mesocosm) 11.3 3.1 11.7 2.3 -0.2 0.888EPT density (ind m-2) 156.9 44.9 254.9 37.0 -2.9 0.045EPT richness (taxa per mesocosm) 3.3 0.6 6.0 1.0 -4.0 0.025Scraper density (ind m-2) 44.1 29.4 122.5 44.9 -2.5 0.074Shredder density (ind m-2) 39.2 30.6 58.8 58.8 -0.5 0.644Predator density (ind m-2) 78.4 55.7 39.2 22.5 1.1 0.350Collector density (ind m-2) 142.1 51.6 122.5 17.0 0.6 0.586drift rate day 2 (ind h-1) 1.3 1.2 1.3 0.3 0.0 0.976drift rate day 22 (ind h-1) 0.3 0.1 0.6 0.1 -3.2 0.123Ecosystem functionLeaf decomposition (g AFDM lost) 0.5 0.1 0.7 0.1 -1.8 0.14091Table 4.4: ANOVA summary of linear models evaluating the impact of distur-bance level within the tributaries on response variables of second-orderchannels. Significant effects (P < 0.05) are indicated in bold.Stress levelResponse variables df F PHabitat variablesSediment dry mass 2,6 19.024 0.003Sediment AFDM 2,6 13.852 0.006Sediment % organic 2,6 2.113 0.202Nitrate 2,6 0.106 0.901Benthic invertebratesTotal density 2,6 1.344 0.329Total richness 2,6 4.167 0.081EPT density 2,6 5.187 0.049EPT richness 2,6 1.716 0.232Scraper density 2,6 1.972 0.220Shredder density 2,6 2.197 0.192Predator density 2,6 1.839 0.238Collector density 2,6 0.736 0.518drift rate day 2 2,5 0.299 0.754drift rate day 22 2,6 1.760 0.264Ecosystem functionLeaf decomposition 2,6 0.300 0.75192a.b.Figure 4.1: Photo of the experimental channel-network set-up consisting in24 first-order channels converging in pairs to feed 12 second-order chan-nels (a). Detail of a pair of first-order channels and their second-orderrecipient (b).93Figure 4.2: Schematic representation of our experimental channel-networkset up. Each pair of first-order channels (represented as gray bars) isa treatment for its second-order receptor (black bars). Initials are usedto represent stressor additions in first-order treatments (N, nutrients;S, sediment; N+S, nutrient + sediment). Letters are used to representsecond-order tributary treatments (a, two control tributaries; b, one con-trol tributary and one tributary with both stressors; c, nutrients added inone tributary and sediment in the other; and d, nutrients and sedimentadded in both tributaries). There were three replicates of each tribu-tary treatment for a total of 12 second-order channels and 24 first-orderchannels.94Figure 4.3: Standardized effect size (Hedge’s d ± 95% CI) for significantmain and interactive effects of the stressors on first-order channels. Let-ters are used to represent main effects (N, nutrients; S, sediment) andinteractions (N*S, nutrients x sediment). For main effects positive val-ues indicate increases in the response variable, whereas negative val-ues indicate the opposite. For interactions confidence intervals over-lapping zero indicate additive interactions, positive values denote syn-ergies, and negative values indicate antagonisms. Symbols are used torepresent significance according to the linear models presented in Table4.1: ***P < 0.001, **P < 0.01, *P < 0.05, ·P < 0.1.95Figure 4.4: Bar plot illustrating significant antagonistic effects of the stres-sors on leaf decomposition (as AFDM loss from 3 g leaf packs). Barsrepresent the mean of the treatments (± SE): control (n = 9), sediment(S, n = 3), nutrients (N, n = 3), nutrients + sediment (N+S, n = 9). Adotted line represents the predicted additive effect of the two stressors.Letters indicate significant differences among the treatments accordingto Tukey’s Honestly Significant Difference tests.Figure 4.5: Plots representing the impact of the stressor treatments on thefrequency of zeros (a) and log-transformed counts (b) of individualsfrom the orders Ephemeroptera, Plecoptera, and Trichoptera drifting outof the first-order channels on day 22. Letter notation for the treatmentsis consistent with Figure 4.4.96Figure 4.6: Bar plots illustrating the effect of tributary level of disturbance ontotal sediment deposition (a), organic sediment deposition (b), and EPTdensity (c) in recipient second-order channels. Treatments in the x-axisrepresent: no disturbance in the tributaries (0, tributary treatment a inFigure 4.2), one tributary disturbed (1, tributary treatment b in Figure4.2), two tributaries disturbed (2, tributary treatment d in Figure 4.2).Bars represent the mean of each tributary treatment (±SE, n = 4). Let-ters indicate significant differences among the treatments according toTukey’s Honestly Significant Difference tests.97Chapter 5Conclusions: synthesis andimplications5.1 OverviewMy thesis presented a detailed, mechanistic investigation of freshwater ecosystemresponses to three widespread agricultural stressors, i.e. nutrient enrichment, sedi-mentation, and insecticide contamination. These studies constitute some of the firstexperimental evaluations of three-way interactions among these stressors, and ad-dressed two freshwater ecosystem types, encompassing processes occurring at thescale of communities, ecosystems, and meta-ecosystems (river networks). My re-sults contribute novel evidence suggesting that nutrient enrichment, sedimentation,and insecticide contamination often interact non-additively, as I observed multipleantagonistic and reversal interactions altering functioning and biological composi-tion of lentic and lotic ecosystems at different scales of organization.In my first experiment (Chapter 2), I observed that even low concentrations ofthe insecticide chlorpyrifos have the potential to interact with fine sediment, exert-ing deleterious impacts on small-sized invertebrates in a closed stream community.Furthermore, my results suggested that even though chlorpyrifos did not strongly98reduce invertebrate density, it still altered organic matter processing by shreddinginvertebrates, which resulted in significant reductions in leaf decomposition rates.With my second experiment (Chapter 3), I demonstrated that open, lentic ecosys-tems are also susceptible to complex interactions among nutrients, sediment, andinsecticide contamination, as I observed that imidacloprid frequently interactedantagonistically with moderate levels of nutrient enrichment and sedimentation.Moreover, my results indicated that imidacloprid has the potential to influenceecosystem metabolism through impacts on primary consumers.Finally, in my third experiment, I explored processes occurring at the scale ofthe river network meta-ecosystem, and observed that complex multiple-stressor in-teractions have the potential to strongly alter the flux of organisms from tributariesto downstream ecosystems. Furthermore, my results suggested that, at small spatialand relatively short time scales, these alterations of within-network migration pat-terns, may be more influential than the transmission of disturbances (e.g., nutrientand sediment loads) from headwaters to downstream recipient ecosystems.5.2 Single- and multiple-stressor responses across threefreshwater mesocosm manipulationsIn this section I discuss response patterns observed across the three experimentalmanipulations described in this dissertation. Even though the response variablesanalyzed here are not strictly independent, they show trends that were consistentacross two different ecosystem types and three different scales of biological orga-nization.5.2.1 Single stressor effects: the disproportionate impacts ofsedimentationAcross the three experiments conducted for this dissertation, I evaluated 51 univari-ate biological responses for both nutrient enrichment and sediment, 20 responsesfor chlorpyrifos, and 19 responses for imidacloprid (Figure 5.1a). In both lentic and99lotic ecosystems, and across all scales measured in these experiments, sedimenta-tion was consistently the most detrimental of the three types of agricultural stres-sors evaluated. Sediment additions significantly affected 17 biological responses,with negative effects on 15 variables and positive effects on only 2 (leaf decompo-sition, EPT drift flux, Figure 5.1a). These observations agree with literature sug-gesting that increased fine sediment inputs in agricultural landscapes are among themost pervasive ecological stressors for stream (Matthaei et al., 2010; Piggott et al.,2015b; Louhi et al., 2017) and lentic ecosystems (Luo et al., 1997; Skagen et al.,2008; Wood and Richardson, 2009). Furthermore, the observed impact mecha-nisms generally support previous experimental evidence. In keeping with a numberof previous investigations (e.g., Bilotta and Brazier, 2008; Wood and Richardson,2009; Wagenhoff et al., 2012; Piggott et al., 2015b; Louhi et al., 2017), I observedthat fine sediment additions consistently reduced benthic and pelagic invertebratedensity, mostly due to physical impacts on sensitive taxa (e.g., Ephemeroptera, Ple-coptera, and Trichoptera), and bottom-up effects from reduced primary productionand impaired leaf decomposition. However, my studies also offered new insightsabout the mechanisms by which the impacts of sedimentation scale up from com-munities to ecosystems and meta-ecosystems. With the pond mesocosm manip-ulation, I was able to establish clear cause-effect associations between inorganicsediment inputs and strong alterations of ecosystem metabolism. Moreover, withthe experimental channel network, I demonstrated that sedimentation in tributarychannels had unexpected effects on species composition of downstream ecosys-tems, as it promoted migration of sensitive invertebrate species within the rivernetwork meta-ecosystem.The second most influential stressor was nutrient enrichment, which had pos-itive effects on 5 (9.8%) of the variables tested, and negative effects on only onevariable (EPT drift flux, Figure 5.1a). This was not surprising because moderatenutrient additions have been extensively documented to produce positive bottom-up effects on benthic and pelagic food webs in oligotrophic freshwater ecosys-tems (Lienesch et al., 2005; Shurin et al., 2012). However, all these significanteffects were observed in pond mesocosms. In contrast to many investigations find-ing strong effects of nutrient enrichment in stream mesocosm experiments (Piggott100et al., 2015b; Garcı´a et al., 2017), the application of nutrients did not cause strongresponses in either of my stream experiments. This unexpected result indicatesthat nutrient additions in these experiments were not sufficient to cause a strong re-sponse that would be detectable by the time of sampling. However, in both streamexperiments I detected patterns that suggest there was an increase in biomass ofprimary producers that was quickly transferred to higher trophic levels (Wagen-hoff et al., 2012). For instance, in the microcosm experiment nutrient tanks hadon average higher periphyton biomass and higher invertebrate biomass, but theseresponses were not statistically significant due to their large variation among en-riched tanks. In the channel network experiment, I did not measure periphyton butthere was lower emigration of Ephemeroptera, Plecoptera, and Trichoptera taxafrom enriched channels, suggesting higher food availability due to the enrichment(Hammock and Wetzel, 2013).The results of this investigation also highlighted the influential role of insecti-cide contamination in lentic and lotic ecosystems, contributing new evidence of thepotential indirect impacts of insecticides on freshwater ecosystem function. Evenat relatively low concentrations, I observed significant negative effects of insecti-cide contamination on 2 (4%) of the biological responses tested (Figure 5.1a). Thisis a considerable proportion, taking into account these were aggregated communityand ecosystem responses, and do not measure taxon-specific effects of the insec-ticides. Out of these significant responses, one was due to direct toxic effects ofimidacloprid on zooplankton communities, and the remaining was documented forindirect effects of chlorpyrifos on ecosystem function. Some of these responses,such as the negative effects of chlorpyrifos on leaf decomposition, had been previ-ously documented in experimental and field studies (Brock et al., 1992a; Cuppenet al., 1995; Van den Brink et al., 1996). However, this is one of the first studiesdocumenting significant negative impacts of imidacloprid on zooplankton densityat concentrations that are not considered harmful according to standardized toxicitytests (Sa´nchez-Bayo and Goka, 2006). Moreover, I observed marginally significanteffects of imidacloprid on net ecosystem metabolism, a potential indirect effectthat has not been previously reported in the literature. These findings contributeto existing evidence suggesting that impacts of neonicotinoid insecticides may be101stronger in realistic scenarios than predicted by acute toxicity tests (Sa´nchez-Bayoet al., 2016).5.2.2 Cumulative effects: the importance of antagonistic andreversal interactionsOut of the 168 stressor combinations evaluated across the three studies, 157 (94%)were additive and 11 (6%) resulted in significant interactive effects (Figure 5.1b).From these non-additive interactions, 9 (82%) were antagonistic and 2 (8%) werereversals. The importance of antagonistic interactions on freshwater ecosystemshas been previously underscored by Jackson et al. (2016), who observed that antag-onisms comprised 41% of the 286 stressor combinations evaluated in their meta-analysis. However, my results differed from the trend documented by Jacksonet al. (2016), who observed that antagonistic interactions occurred most frequentlyon functional metrics, while diversity metrics showed mostly additive responses.In my studies, most antagonisms were observed for diversity metrics, while func-tional metrics exhibited mostly additive responses to cumulative stressors. Thisdiscrepancy further reinforces the notion that the outcome of multiple-stressor in-teractions is strongly determined by the context and the specific characteristics ofthe receptor biological systems (Segner et al., 2014; Clements et al., 2016).The importance of context-dependent responses to multiple stressors has beenrecently highlighted by Clements et al. (2016), who observed that our ability topredict variation in the response to stressors among communities is still limited.My studies contribute new evidence to this research gap, demonstrating that char-acteristics at the scale of microhabitats may have a disproportionate influence onthe response of aquatic invertebrates to cumulative stressors. My finding of antag-onistic effects on species diversity also supports the notion of context-dependency,as it suggests that community composition and tolerance patterns also determineresponses to stressors. Antagonistic impacts on diversity suggest that, in my focalcommunities, species tolerances to the stressors were positively correlated. Hence,each stressor eliminated certain species but was likely to leave species that weretolerant to the other stressors, a pattern that has been denominated “stress-induced102community tolerance” (Vinebrooke et al., 2004). Most importantly, the preva-lence of additive effects on functional measures of these systems indicates thatthe remaining species were not able to compensate functionally for the speciesloss (Jackson et al., 2016), offering new evidence of potential community andecosystem-level consequences of the development of community tolerance in fresh-water ecosystems.Although I did not observe synergistic interactions according to the classifi-cation framework I used, one of the reversal interactions I documented ultimatelyresulted in effects that were more detrimental than predicted by additivity. Thisis an interesting observation that further reinforces the importance of consideringreversal interactions in multiple-stressor classification frameworks (Piggott et al.,2015c; Jackson et al., 2016). With additive classification frameworks that do notaccount for reversals, (e.g., the classification framework used by Crain et al., 2008),the above interaction would have been classified as an antagonism, which wouldhave clearly underestimated its effects on freshwater communities. Reversal in-teractions are important because they represent the most extreme cases of “eco-logical surprises”, where the cumulative impact of two stressors actually goes inthe opposite direction as predicted by their single effects (Piggott et al., 2015c;Jackson et al., 2016). According to my results they should be carefully consid-ered in multiple-stressor studies, because they could represent instances where thepresence of one stressor inverts the effect of the dominant stressor with potentialdeleterious effects on freshwater ecosystems (Jackson et al., 2016).In agreement with my initial expectations, sediment x insecticide was the stres-sor combination that most frequently showed complex interactions (Figure 5.1b).The potential negative effects of pesticide-contaminated sediments on freshwaterecosystems have been recognized for some time (Warren et al., 2003; Burton andJohnston, 2010). Because some insecticides are moderately hydrophobic, they tendto adsorb onto sediment particles when they enter the water column (Brock et al.,1992a; Pablo et al., 2008; Gebremariam, 2011). Thus, deposited bed sedimentsmay become a sink for insecticides, thereby increasing the exposure of benthicorganisms through contact or ingestion of contaminated particles (Warren et al.,2003; Burton and Johnston, 2010). However, my studies are some of the first at-103tempts to explicitly evaluate potential synergistic interactions among sedimentationand insecticides at the scale of communities and ecosystems. As such, my resultsprovide novel evidence that increased fine sediment deposition has the potential toenhance the toxic effects of insecticides like chlorpyrifos on benthic invertebratecommunities.Interestingly, the presence of fine sediment did not consistently enhance thenegative effects of the insecticides in my experiments (Figure 5.2). Sediment addi-tions actually seemed to mask the negative effects of imidacloprid on pelagic inver-tebrate communities in the pond mesocosms. Because this was the first experimentevaluating multiple stressor interactions between imidacloprid and sedimentation,I did not have empirical information with which to contrast my results. However,according to my observations, the presence of both stressors in the system causedthe stress-induced community tolerance pattern described in previous paragraphs(Vinebrooke et al., 2004). A similar response was observed for the combination ofnutrients and the two insecticides (Figure 5.2), all of which were antagonistic andaffected diversity variables, supporting results of previous empirical evaluations(Alexander et al., 2013, 2016). Finally, in my studies, nutrients and fine sedimentonly interacted non-additively once, which contrasts with previous examinationsthat have found frequent complex interactions for this stressor combination (e.g.,Townsend et al., 2008; Piggott et al., 2012, 2015b). This is not entirely surprisingbecause these complex interactions have been reported for stream ecosystems andmy nutrient additions did not exert strong impacts in either of my stream mesocosmexperiments.5.3 ImplicationsMy studies contribute new evidence suggesting that complex interactions amongnutrient enrichment, sedimentation, and insecticide contamination are common indifferent types of freshwater ecosystems, and have distinct mechanisms operatingat different scales of organization. As such, my findings support a growing body ofliterature underscoring the need to account for potential detrimental effects arisingfrom these interactions, in order to control and mitigate agricultural impacts on104freshwater biodiversity and function (Matthaei et al., 2010; Piggott et al., 2015b;Liess et al., 2016).My results generally support the notion that in order to predict the outcome ofmultiple-stressor interactions it is necessary to shift from focusing solely on stres-sor characteristics, to focusing also on the properties of the receptor systems (Rohret al., 2006; Segner et al., 2014). According to my observations, physical andchemical habitat characteristics, the degree of connectivity of the ecosystem, andthe patterns of species co-tolerance to the individual stressors, are major determi-nants of the outcome of multiple-stressor interactions. An important consequenceof this context-dependency is that even though most significant interactions ob-served in these experiments were antagonistic, in a different context (e.g. differentmicrohabitat characteristics or different stressor levels), similar stressor combina-tions could result in synergistic effects or reversals, with potentially deleteriousconsequences for freshwater ecosystems. Moreover, these observations suggestthat theoretical frameworks like the community co-tolerance hypothesis (Vine-brooke et al., 2004) may be good starting points to predict cumulative stressoreffects on biological diversity. However, many more empirical studies are neededto better understand how multiple stressors impact diversity, and how these im-pacts on diversity will ultimately translate into impacts on freshwater ecosystemfunctioning (Sandin and Solimini, 2009). For instance, such studies could involvethe manipulation of species assemblages to evaluate multiple-stressor effects oncommunities that are predicted to exhibit stress-induced tolerance.One of the most important findings of this investigation is that environmentallyrelevant levels of insecticide contamination may interact non-additively with nutri-ent enrichment and sedimentation, and cause important effects on ecosystem func-tion. For example, I found that the presence of fine sediment in the substrate hasthe potential to enhance insecticide toxicity to small invertebrates. This observationhas major implications given the widespread co-occurrence of these two stressorsin freshwater ecosystems around the world (Vo¨ro¨smarty et al., 2010; Scha¨fer et al.,2016), and highlight the importance of including evaluations of sediment-boundtoxicity in risk assessment protocols for pesticides in freshwater ecosystems. Fur-thermore, across my studies sublethal insecticide exposures consistently altered105ecosystem processes through effects on trophic interactions. These findings sug-gest that risk assessment frameworks based on single-species toxicity tests mayunderestimate impacts on freshwater ecosystem function, as they cannot accountfor indirect effects through disruption of species interactions or complex interac-tions with other agricultural stressors. In this sense, my results support a growingbody of literature showing that mesocosm experiments are a valuable complemen-tary tool for pesticide risk assessment, which allows controlled and statisticallyrobust evaluations of toxicant effects in more realistic scenarios (Culp et al., 2000;Clements, 2004; Alexander et al., 2016).Finally, my experiments at different scales of ecological organization consis-tently showed that the effects at higher levels of organization cannot be easily pre-dicted from impacts detected at smaller, nested scales (e.g., impacts at the meta-ecosystem scale are not easy to predict from impacts on local ecosystems). Forinstance, I detected strong negative effects of sediment inputs on net ecosystemproductivity of the pond mesocosms, despite the fact that there were not significantperiphyton or phytoplankton biomass responses when analyzed separately. An-other example of surprising emergent properties in my study systems were the un-expected positive impacts of headwater disturbances on downstream ecosystems,due to mass effects within the river network meta-ecosystem. Together, these find-ings highlight the importance of choosing the correct scale of measurement to as-sess ecological condition, and underscore the usefulness of whole-ecosystem ex-periments to reliably evaluate stressor impacts on ecosystem structure and function(Loreau et al., 2003; Buck et al., 2004).The key management implication of my results is that sedimentation, nutrientenrichment, and insecticide pollution should be managed together in order to effec-tively protect freshwater ecosystems in agricultural landscapes. Freshwater man-agement decisions should take into account that, even at relatively low levels, thesestressors may act in combination to cause unexpected negative effects on aquaticbiota and ecosystem processes. In particular, my results suggest that sedimenta-tion was the most detrimental stressor and had the potential to enhance toxicity ofhydrophobic insecticides. Thus, management strategies that reduce the amount offine sediment entering aquatic ecosystems, such as soil conservation practices and106stabilization structures, should be a priority for restoration efforts. Additionally,the establishment of riparian vegetation buffers may be a very effective measureto mitigate the impacts of agriculture, as they can control nutrients, pesticides, andfine sediment entering water bodies via runoff (Udawatta et al., 2008; Lin et al.,2011; Chara´ et al., 2011).107a.b.Figure 5.1: Frequency distribution of individual stressor effects (a), andmultiple-stressor interactions (b) on freshwater ecosystem responsesevaluated across the three experiments conducted for this dissertation.Letters are used to represent individual stressor treatments (N, nutrients;S, sediment; I, insecticide) and interactions (N*S, nutrient x sediment;N*I, nutrient x insecticide; S*I, sediment x insecticide; N*S*I; nutrientx sediment x insecticide). The number of response variables analyzedfor each stressor (a) or stressor combination (b) is indicated in parenthe-ses.108Figure 5.2: Distribution of significant multiple-stressor interactions involv-ing insecticides across the experiments. Letters are used to denote thetreatments: nutrients (N), sediment (S), imidacloprid (IMI), and chlor-pyrifos (CPY).109BibliographyAbell, R., Allan, J. D., and Lehner, B. (2007). Unlocking the potential of protectedareas for freshwaters. Biological Conservation, 134(1):48–63. → pages 1Alexander, A. C. and Culp, J. M. (2008). Insulate or exacerbate? exploringnutrient masking of contaminant effects. Integrated Environmental Assessmentand Management, 4(2):263–264. → pages 7, 30Alexander, A. C., Culp, J. M., Baird, D. J., and Cessna, A. J. (2016).Nutrient-insecticide interactions decouple density-dependent predation pressurein aquatic insects. Freshwater Biology, 61(12):2090–2101. → pages 8, 25, 30,41, 42, 58, 60, 104, 106Alexander, A. C., Luis, A. T., Culp, J. M., Baird, D. J., and Cessna, A. J. (2013).Can nutrients mask community responses to insecticide mixtures?Ecotoxicology, 22(7):1085–100. → pages 14, 15, 29, 41, 58, 104Allan, J. D. (2004). Landscapes and riverscapes: the influence of land use onstream ecosystems. Annual Review of Ecology, Evolution, and Systematics,35(1):257–284. → pages 2, 3, 4, 13, 14, 26, 41, 42, 54, 55, 74, 84Allan, J. D. and Castillo, M. M. (2007). Stream Ecology, Structure and Functionof Running Waters. Springer, Dordrecht. → pages 1, 73Allan, J. D. and Johnson, L. (1997). Catchment-scale analysis of aquaticecosystems. Freshwater Biology, 37(1):107–111. → pages 2Andersen, T. and Hessen, D. (1991). Carbon, nitrogen, and phosphorus content offreshwater zooplankton. Limnology and Oceanography, 36(4):807–814. →pages 49Anderson, T. A., Salice, C. J., Erickson, R. A., McMurry, S. T., Cox, S. B., andSmith, L. M. (2013). Effects of landuse and precipitation on pesticides and110water quality in playa lakes of the southern high plains. Chemosphere,92(1):84–90. → pages 45APHA (2005). Standard Methods for the Examination of Water and Wastewaters.American Public Health Association, Washington D.C., USA, 21th edition. →pages 18, 46, 78Arar, E. J. and Collins, G. B. (1997). Method 445.0: In vitro determination ofChlorophyll a and Phaeophytin a in marine and freshwater algae byfluorescence. Technical Report September, U.S. Environmental ProtectionAgency, Cincinnati, Ohio. → pages 19, 46Aristi, I., Dı´ez, J. R., Larran˜aga, A., Navarro-Ortega, A., Barcelo´, D., and Elosegi,A. (2012). Assessing the effects of multiple stressors on the functioning ofMediterranean rivers using poplar wood breakdown. The Science of the totalenvironment, 440:272–9. → pages 7Atwood, T. B., Hammill, E., Kratina, P., Greig, H. S., Shurin, J. B., andRichardson, J. S. (2015). Warming alters food web-driven changes in the CO2flux of experimental pond ecosystems. Biology Letters, 11(12):20150785. →pages 3Bates, D., Maechler, M., Bolker, B., and Walker, S. (2015). Fitting LinearMixed-Effects Models Using lme4. Journal of Statistical Software, 67(1):1–48.→ pages 21, 49Beketov, M. a., Kefford, B. J., Scha¨fer, R. B., and Liess, M. (2013). Pesticidesreduce regional biodiversity of stream invertebrates. Proceedings of theNational Academy of Sciences of the United States of America,110(27):11039–43. → pages 4Benda, L., Poff, N. L., Miller, D., Dunne, T., Reeves, G., Pess, G., and Pollock,M. (2004). The network dynamics hypothesis: how channel networks structureriverine habitats. BioScience, 54(5):413–427. → pages 73Benke, A. C., Huryn, A. D., Smock, L. A., and Wallace, J. B. (1999).Length-Mass Relationships for Freshwater Macroinvertebrates in NorthAmerica with Particular Reference to the Southeastern United States. Journalof the North American Benthological Society, 18(3):308–343. → pages 18, 47Benoy, G., Sutherland, A. B., Culp, J. M., and Brua, R. B. (2012). Physical andecological thresholds for deposited sediments in streams in agriculturallandscapes. Journal of Environmental Quality, 41:31–40. → pages 14111Bernhardt, E. S., Likens, G. E., Hall, R. O., Buso, D. C., Fisher, S. G., Burton,T. M., Meyer, J. L., Mcdowell, W. H., Mayer, M. S., Bowden, W. B., Findlay, S.E. G., Macneale, K. H., Stelzer, R. S., and Lowe, W. H. (2005). Can’t see theforest for the stream? In-stream processing and terrestrial nitrogen exports.BioScience, 55(3):219–230. → pages 75Bevenger, G. and King, R. M. (1995). A pebble count procedure for assessingwatershed cumulative effects. Technical report, U.S. Department ofAgriculture, Forest Service, Rocky Mountain Forest and Range ExperimentalStation, Fort Collins, CO. → pages 7Bilotta, G. S. and Brazier, R. E. (2008). Understanding the influence of suspendedsolids on water quality and aquatic biota. Water Research, 42(12):2849–61. →pages 100Bo¨ttger, R., Feibicke, M., Schaller, J., and Dudel, G. (2013). Effects of low-dosedimidacloprid pulses on the functional role of the caged amphipod Gammarusroeseli in stream mesocosms. Ecotoxicology and Environmental Safety,93:93–100. → pages 41Boube´e, J. a. T., Dean, T. L., West, D. W., and Barrier, R. F. G. (1997). Avoidanceof suspended sediment by the juvenile migratory stage of six New Zealandnative fish species. New Zealand Journal of Marine and Freshwater Research,31:61–69. → pages 45Breitburg, D. L., Baxter, J. W., Hatfield, C. A., Howarth, R. W., Jones, C. G.,Lovett, G. M., and Wigand, C. (1998). Understanding effects of multiplestressors: ideas and challenges. In Pace, M. L. and Groffman, P. M., editors,Successes, Limitations, and Frontiers in Ecosystem Science, pages 416–431.Springer New York, New York, NY. → pages 5, 10, 11Brock, T. C. M., Crum, S. J. H., Van Wijngaarden, R., Budde, B. J., Tijink, J.,Zuppelli, A., Leeuwangh, P., and Vanwijngaarden, R. (1992a). Fate and effectsof the insecticide Dursban 4E in indoor Elodea-dominated and macrophyte-freefreshwater model ecosystems: I. Fate and primary effects of the activeingredient chlorpyrifos. Archives of Environmental Contamination andToxicology, 23(1):69–84. → pages 101, 103Brock, T. C. M., Van den Bogaert, M., Bos, A. R., Van Breuklen, S. W. F., Reiche,R., Terwoert, J., Suykerbuyk, R. E. M., and Roijackers, R. M. M. (1992b). Fateand effects of the insecticide Dursban 4E in indoor Elodea-dominated andmacrophyte-free freshwater model ecosystems II. Secondary effects on112community structure. Archives of Environmental Contamination andToxicology, 23(4):391–409. → pages 14, 27, 28, 29, 55Brosed, M., Lamothe, S., and Chauvet, E. (2016). Litter breakdown for ecosystemintegrity assessment also applies to streams affected by pesticides.Hydrobiologia, 773(1):87–102. → pages 27, 31Brown, B. L., Swan, C. M., Auerbach, D. A., Campbell Grant, E. H., Hitt, N. P.,Maloney, K. O., and Patrick, C. (2011). Metacommunity theory as amultispecies, multiscale framework for studying the influence of river networkstructure on riverine communities and ecosystems. Journal of the NorthAmerican Benthological Society, 30(1):310–327. → pages 73, 74, 87, 88Buck, O., Niyogi, D. K., and Townsend, C. R. (2004). Scale-dependence of landuse effects on water quality of streams in agricultural catchments.Environmental Pollution, 130(2):287–99. → pages 106Burdon, F. J., McIntosh, A. R., and Harding, J. S. (2013). Habitat loss drivesthreshold response of benthic invertebrate communities to deposited sedimentin agricultural streams. Ecological Applications, 23(5):1036–47. → pages 14Burton, G. A. and Johnston, E. L. (2010). Assessing contaminated sediments inthe context of multiple stressors. Environmental Toxicology and Chemistry,29(12):2625–2643. → pages 9, 103Campbell Grant, E. H., Lowe, W. H., and Fagan, W. F. (2007). Living in thebranches: population dynamics and ecological processes in dendritic networks.Ecology Letters, 10(2):165–175. → pages 73, 74, 87, 88CCME (2007). Canadian water quality guidelines: Imidacloprid. Technicalreport, Canadian Council of Ministers of the Environment, Winnipeg. → pages42Chara´, J. D., Giraldo, L. P., Chara´-Serna, A. M., and Pedraza, G. X. (2011).Beneficios de los corredores riberen˜os de Guadua angustifolia en la proteccio´nde ambientes acua´ticos en la Ecorregio´n Cafetera de Colombia. 2. Efectossobre la escorrentı´a y captura de nutrientes. Revista Recursos Naturales yAmbiente -CATIE, 61:60–66. → pages 107Chara´-Serna, A. M., Chara´, J. D., Giraldo, L. P., Zu´n˜iga, C., and Allan, J. D.(2015). Understanding the impacts of agriculture on Andean stream ecosystemsof Colombia: a causal analysis using aquatic macroinvertebrates as indicatorsof biological integrity. Freshwater Science, 34(2):727–740. → pages 13113Chen, G. K. (1992). Use of Basin Survey Data in Habitat Modelling andCumulative Watershed Effects Analyses Habitat and Fish Relationships. FishHabitat Relationship Technical Bulletin, 8:1–12. → pages 7Churchel, M. A., Hanula, J. L., Berisford, C. W., Vose, J. M., and Dalusky, M. J.(2011). Impact of imidacloprid for control of hemlock woolly aelgid on nearbyaquatic macroinvertebrate assemblages. Southern Journal of Applied Forestry,35(1):26–32. → pages 41Clements, W. H. (2004). Small-scale experiments support causal relationshipsbetween metal contamination and macroinvertebrate community responses.Ecological Applications, 14(3):954–967. → pages 30, 106Clements, W. H., Cadmus, P., and Brinkman, S. F. (2013). Responses of aquaticinsects to Cu and Zn in stream microcosms: Understanding differences betweensingle species tests and field responses. Environmental Science & Technology,47:7506–7513. → pages 8, 17Clements, W. H., Kashian, D. R., Kiffney, P. M., and Zuellig, R. E. (2016).Perspectives on the context-dependency of stream community responses tocontaminants. Freshwater Biology, 61(12):2162–2170. → pages 29, 102Colombo, V. and Mohr, S. (2013). Structural changes in a macrozoobenthosassemblage after imidacloprid pulses in aquatic field-based microcosms.Archives of Environmental Contamination and Toxicology, 65:683–692. →pages 56, 60Comte, L., Lek, S., de Deckere, E., de Zwart, D., and Gevrey, M. (2010).Assessment of stream biological responses under multiple-stress conditions.Environmental Science and Pollution Research, 17(8):1469–78. → pages 7Connolly, N. M. and Pearson, R. G. (2007). The effect of fine sedimentation ontropical stream macroinvertebrate assemblages: a comparison usingflow-through artificial stream channels and recirculating mesocosms.Hydrobiologia, 592(1):423–438. → pages 84Cooper, L. M. and Sheate, W. R. (2002). Cumulative effects assessment: A reviewof UK environmental impact statements. Environmental Impact AssessmentReview, 22:415–439. → pages 7Crain, C. M., Kroeker, K., and Halpern, B. S. (2008). Interactive and cumulativeeffects of multiple human stressors in marine systems. Ecology letters,11(12):1304–15. → pages 5, 6, 10, 60, 103114Culp, J. M., Podemski, C. L., Cash, K. J., and Lowell, R. B. (2000). A researchstrategy for using stream microcosms in ecotoxicology : integratingexperiments at different levels of biological organization with field data.Journal of Aquatic Ecosystem Stress and Recovery, 7:167–176. → pages 30,59, 106Cuppen, J. G. M., Crum, S. J. H., Van den Heuvel, H. H., Smidt, R. A., and Vanden Brink, P. J. (2002). Effects of a mixture of two insecticides in freshwatermicrocosms: I. Fate of chlorpyrifos and lindane and responses ofmacroinvertebrates. Ecotoxicology, 11:165–80. → pages 14Cuppen, J. G. M., Gylstra, R., van Beusekom, S., Budde, B. J., and Brock, T.C. M. (1995). Effects of nutrient loading and insecticide application on theecology of Elodea-dominated freshwater microcosms. III. Responses ofmacroinvertebrate detritivores, breakdown of plant litter, and final conclusions.Archiv Fur Hydrobiologie, 133(4):417–439. → pages 27, 101Daam, M. A., Crum, S. J. H., Van den Brink, P. J., and Nogueira, A. J. A. (2008).Fate and effects of the insecticide chlorpyrifos in outdoor plankton-dominatedmicrocosms in Thailand. Environmental Toxicology and Chemistry,27(12):2530–8. → pages 14Dama´sio, J., Ferna´ndez-Sanjuan, M., Sa´nchez-Avila, J., Lacorte, S., Prat, N.,Rieradevall, M., Soares, A. M. V. M., and Barata, C. (2011). Multi-biochemicalresponses of benthic macroinvertebrate species as a complementary tool todiagnose the cause of community impairment in polluted rivers. WaterResearch, 45(12):3599–613. → pages 7Danger, M., Cornut, J., Elger, A., and Chauvet, E. (2012). Effects of burial on leaflitter quality, microbial conditioning and palatability to three shredder taxa.Freshwater Biology, 57(5):1017–1030. → pages 85Davies-Colley, R. J., Hickey, C. W., Quinn, J. M., and Ryan, P. A. (1992). Effectsof clay discharges on streams 1. Optical properties and epilithon.Hydrobiologia, 248:215–234. → pages 55Diana, M., Allan, J. D., and Infante, D. M. (2006). The influence of physicalhabitat and land use on stream fish assemblages in southeastern Michigan.American Fisheries Society Symposium, 48:359–374. → pages 2Dodds, W. K. and Oakes, R. M. (2008). Headwater influences on downstreamwater quality. Environmental Management, 41:367–77. → pages 7115Dodds, W. K. and Whiles, M. R. (2010). Freshwater Ecology: Concepts andEnvironmental Applications of Limnology. Elsevier, New York, NY, 2 edition.→ pages 9Dudgeon, D., Arthington, A. H., Gessner, M. O., Kawabata, Z., Knowler, D. J.,Le´veˆque, C., Naiman, R. J., Prieur-Richard, A. H., Soto, D., Stiassny, M. L. J.,and Sullivan, C. A. (2006). Freshwater biodiversity: importance, threats, statusand conservation challenges. Biological Reviews of the CambridgePhilosophical Society, 81(2):163–82. → pages 1, 40Dunne, R. P. (2009). Synergy or antagonisminteractions between stressors oncoral reefs. Coral Reefs, 29(1):145–152. → pages 6Dyke, F. V. (2008). The Conservation of Aquatic Systems. In Dyke, F. V., editor,Conservation Biology: Foundations, Concepts, Applications, pages 313–346.Springer Netherlands, Dordrecht. → pages 1, 40Esselman, P. C. and Allan, J. D. (2010). Relative influences of catchment- andreach-scale abiotic factors on freshwater fish communities in rivers ofnortheastern Mesoamerica. Ecology of Freshwater Fish, 19(3):439–454. →pages 7Evans-White, M. A., Stelzer, R. S., and Lamberti, G. A. (2005). Taxonomic andregional patterns in benthic macroinvertebrate elemental composition instreams. Freshwater Biology, 50:1786–1799. → pages 49Fisher, S. G. (1997). Creativity, idea generation, and the functional morphologyof streams. Journal of the North American Benthological Society,16(2):305–318. → pages 73Folt, C. L., Chen, C. Y., Moore, M. V., and Burnaford, J. (1999). Synergism andantagonism among multiple stressors. Limnology and Oceanography,44(3-2):864–877. → pages 5, 6, 7Fox, J. and Weisberg, S. (2011). An {R} Companion to Applied Regression. SagePublications, Thousand Oaks, CA, 2 edition. → pages 21, 49Freeman, M. C., Pringle, C. M., and Jackson, C. R. (2007). Hydrologicconnectivity and the contribution of stream headwaters to ecological integrity atregional scales. Journal of the American Water Resources Association,43(1):5–14. → pages 9, 74116Garcı´a, L., Pardo, I., Cross, W. F., and Richardson, J. S. (2017). Moderate nutrientenrichment affects algal and detritus pathways differently in a temperaterainforest stream. Aquatic Sciences, 79:941–952. → pages 3, 42, 77, 101Gardestro¨m, J., Ermold, M., Goedkoop, W., and Mckie, B. G. (2015).Disturbance history influences stressor impacts: Effects of a fungicide andnutrients on microbial diversity and litter decomposition. Freshwater Biology,61:2171–2184. → pages 59Gebremariam, S. Y. (2011). Mineralization, sorption and desorption ofchlorpyrifos in aquatic sediments and soils. Phd thesis, Washington StateUniversity. → pages 14, 15, 28, 103Gessner, M. and Chauvet, E. (2002). A case for using litter breakdown to assessfunctional stream integrity. Ecological Applications, 12(2):498–510. → pages31, 60Gessner, M. O. and Tlili, A. (2016). Fostering integration of freshwater ecologywith ecotoxicology. Freshwater Biology, 61(12):1991–2001. → pages 4, 25, 60Giesy, J., Solomon, K., Coats, J., Kenneth, D., Giddings, J., and Kenaga, E.(1999). Ecological risk assessment in North American aquatic environments.Reviews of Environmental Contamination and Toxicology, 160:1–129. → pages14, 15, 28Giesy, J. P. and Solomon, K. R. (2014). Ecological risk assessment forchlorpyrifos in terrestrial and aquatic systems in the United States. Reviews ofEnvironmental Contamination and Toxicology, 231:269. → pages 14, 27Gilliom, R. J. (2007). Pesticides in U.S. streams and groundwater. EnvironmentalScience & Technology, 41(10):3408–3414. → pages 4Giroux, I. (2003). Contamination de l’eau souterraine par les pesticides et lesnitrates dans les re´gions en culture de pommes de terre. Technical report,Direction du suivi de l’e´tat de l’environnement, ministe`re de l’Environnement,Que´bec. → pages 45Gomi, T., Sidle, R. C., and Richardson, J. S. (2002). Understanding processes anddownstream linkages of headwater systems. BioScience, 52(10):905–915. →pages 9, 73, 74Greathouse, E. A., March, J. G., and Pringle, C. M. (2005). Recovery of a tropicalstream after a harvest-related chlorine poisoning event. Freshwater Biology,50:603–615. → pages 7117Greig, H. S., Kratina, P., Thompson, P. L., Palen, W. J., Richardson, J. S., andShurin, J. B. (2012). Warming, eutrophication, and predator loss amplifysubsidies between aquatic and terrestrial ecosystems. Global Change Biology,18(2):504–514. → pages 59Gurevitch, J. and Hedges, L. V. (2006). Meta-analysis: combining the results ofindependent experiments. In Scheiner, S. and Gurevitch, J., editors, Design andAnalysis of Ecological Experiments, pages 347–369. Oxford University Press,Cary, US. → pages 20, 48, 79Hammock, B. G. and Wetzel, W. C. (2013). The relative importance of driftcauses for stream insect herbivores across a canopy gradient. Oikos,122(11):1586–1593. → pages 84, 101Hart, R. C. (1988). Zooplankton feeding rates in relation to suspended sedimentcontent: potential influences on community structure in a turbid reservoir.Freshwater Biology, 19:123–139. → pages 55Hauer, F. R. and Lamberti, G. A. (2007). Methods in Stream Ecology. AcademicPress, Burlington, MA, USA. → pages 18, 47, 79Holmstrup, M., Bindesbøl, A.-M., Oostingh, G. J., Duschl, A., Scheil, V., Ko¨hler,H.-R., Loureiro, S., Soares, A. M. V. M., Ferreira, A. L. G., Kienle, C.,Gerhardt, A., Laskowski, R., Kramarz, P. E., Bayley, M., Svendsen, C., andSpurgeon, D. J. (2010). Interactions between effects of environmentalchemicals and natural stressors: a review. The Science of the TotalEnvironment, 408(18):3746–3762. → pages 5Horppila, J. and Liljendahl-Nurminen, A. (2005). Clay-turbid interactions maynot cascade - A reminder for lake managers. Restoration Ecology,13(2):242–246. → pages 55Jackson, M. C., Loewen, C. J. G., Vinebrooke, R. D., and Chimimba, C. T.(2016). Net effects of multiple stressors in freshwater ecosystems: Ameta-analysis. Global Change Biology, 22:180–189. → pages 5, 6, 8, 20, 28,41, 49, 79, 102, 103Jergentz, S., Mugni, H., Bonetto, C., and Schulz, R. (2005). Assessment ofinsecticide contamination in runoff and stream water of small agriculturalstreams in the main soybean area of Argentina. Chemosphere, 61(6):817–26.→ pages 18118Johnson, J. E., Patterson, D. A., Martins, E. G., Cooke, S. J., and Hinch, S. G.(2012). Quantitative methods for analysing cumulative effects on fish migrationsuccess: a review. Journal of Fish Biology, 81(2):600–31. → pages 7Johnson, L. B. and Host, G. E. (2010). Recent developments in landscapeapproaches for the study of aquatic ecosystems. Journal of the North AmericanBenthological Society, 29(February):41–66. → pages 2Johnston, T. A. and Cunjak, R. A. (1999). Dry mass-length relationships forbenthic insects: a review with new data from Catamaran Brook, NewBrunswick, Canada. Freshwater Biology, 41:653–674. → pages 18, 47Kiffney, P. M. and Clements, W. H. (1996a). Effects of metals on streammacroinvertebrate assemblages from different altitudes. EcologicalApplications, 6(2):472–481. → pages 28Kiffney, P. M. and Clements, W. H. (1996b). Size-dependent response ofmacroinvertebrates to metals in experimental streams. EnvironmentalToxicology and Chemistry, 15(8):1352–1356. → pages 17, 28, 29Kiffney, P. M. and Richardson, J. S. (2001). Interactions among nutrients,periphyton, and invertebrate and vertebrate (Ascaphus truei) grazers inexperimental channels. Copeia, 2:422–429. → pages 77Kiffney, P. M., Richardson, J. S., and Bull, J. P. (2003). Responses of periphytonand insects to experimental manipulation of riparian buffer width along foreststreams. Journal of Applied Ecology, 40(6):1060–1076. → pages 26Kiffney, P. M., Richardson, J. S., and Bull, J. P. (2004). Establishing light as acausal mechanism structuring stream communities in response to experimentalmanipulation of riparian buffer width. Journal of the North AmericanBenthological Society, 23(3):542–555. → pages 26Kimbrough, R. A. and Litke, D. W. (1996). Pesticides in Streams DrainingAgricultural and Urban Areas in Colorado. Environmental Science &Technology, 30(3):908–916. → pages 18Knowlton, M. F. and Jones, J. R. (1995). Temporal and spatial dynamics ofsuspended sediment, nutrients, and algal biomass in Mark Twain Lake,Missouri. Archiv Fur Hydrobiologie, 135(2):145–178. → pages 42, 55Ko¨hler, H. R. and Triebskorn, R. (2013). Wildlife ecotoxicology of pesticides:can we track effects to the population level and beyond? Science,341(6147):759–65. → pages 41119Kratina, P., Greig, H. S., Thompson, P. L., Carvalho-Pereira, T. S. A., and Shurin,J. B. (2012). Warming modifies trophic cascades and eutrophication inexperimental freshwater communities. Ecology, 93(6):1421–30. → pages 8Kreutzweiser, D. P., Good, K. P., Chartrand, D. T., Scarr, T. A., and Thompson,D. G. (2008). Toxicity of the systemic insecticide, imidacloprid, to foreststream insects and microbial communities. Bulletin of EnvironmentalContamination and Toxicology, 80(3):211–4. → pages 27, 41, 42, 57Kreutzweiser, D. P., Thompson, D. G., and Scarr, T. A. (2009). Imidacloprid inleaves from systemically treated trees may inhibit litter breakdown bynon-target invertebrates. Ecotoxicology and Environmental Safety,72(4):1053–7. → pages 27Langer-Jaesrich, M., Ko¨hler, H. R., and Gerhardt, A. (2010). Assessing toxicityof the insecticide thiacloprid on Chironomus riparius (Insecta: Diptera) usingmultiple end points. Archives of Environmental Contamination and Toxicology,58(4):963–72. → pages 41Lauridsen, R. B., Kronvang, B., and Friberg, N. (2006). Occurrence ofsediment-bound pyrethroids in danish streams and their impact on ecosystemfunction. Water, Air, and Soil Pollution, 6:59–68. → pages 27Lecerf, A. and Richardson, J. S. (2010). Litter decomposition can detect effects ofhigh and moderate levels of forest disturbance on stream condition. ForestEcology and Management, 259(12):2433–2443. → pages 26Lecerf, A. and Richardson, J. S. (2011). Assessing the functional importance oflarge-bodied invertebrates in experimental headwater streams. Oikos,120(6):950–960. → pages 76Legendre, P. and Anderson, M. J. (1999). Distance-based redundancy analysis:testing multispecies responses in multifactorial ecological experiments.Ecological Monographs, 69(1):1–24. → pages 20Legendre, P. and Gallagher, E. D. (2001). Ecologically meaningfultransformations for ordination of species data. Oecologia, 129(2):271–280. →pages 49Lienesch, P. W., McDonald, M. E., Hershey, A. E., O’Brien, W. J., and Bettez,N. D. (2005). Effects of a whole-lake, experimental fertilization on lake trout ina small oligotrophic arctic lake. Hydrobiologia, 548(1):51–66. → pages 100120Liess, M., Foit, K., Knillmann, S., Scha¨fer, R. B., and Liess, H.-D. (2016).Predicting the synergy of multiple stress effects. Scientific Reports, 6:32965.→ pages 8, 105Lin, C.-H., Lerch, R. N., Goyne, K. W., and Garrett, H. E. (2011). Reducingherbicides and veterinary antibiotics losses from agroecosystems usingvegetative buffers. Journal of Environmental Quality, 40(3):791–799. → pages107Loftis, J. C., Macdonald, L. H., Streett, S., Iyer, H. K., and Bunte, K. (2001).Detecting cumulative watershed effects: the statistical power of pairing.Journal of Hydrology, 251:49–64. → pages 7, 9Loreau, M., Mouquet, N., and Holt, R. D. (2003). Meta-ecosystems: A theoreticalframework for a spatial ecosystem ecology. Ecology Letters, 6(8):673–679. →pages 74, 106Louhi, P., Ovaska, M., Erkinaro, J., and Muotka, T. (2011). Does fine sedimentconstrain salmonid alevin development and survival? Canadian Journal ofFisheries and Aquatic Sciences, 68:1819–1826. → pages 4Louhi, P., Richardson, J. S., and Muotka, T. (2017). Sediment addition reducesthe importance of predation on ecosystem functions in experimental streamchannels. Canadian Journal of Fisheries and Aquatic Sciences, 74:32–40. →pages 17, 26, 59, 77, 85, 88, 100Luo, H.-R., Smith, L. M., Allen, B. L., and Haukos, D. A. (1997). Effects ofsedimentation on playa wetland volume. Ecological Applications,7(1):247–252. → pages 9, 100Macdonald, L. H. (2000). Evaluating and managing cumulative effects: processand constraints. Environmental Management, 26(3):299–315. → pages 5, 7Magbanua, F. S., Townsend, C. R., Hageman, K. J., and Matthaei, C. D. (2013).Individual and combined effects of fine sediment and the herbicide glyphosateon benthic macroinvertebrates and stream ecosystem function. FreshwaterBiology, 58(8):1729–1744. → pages 3, 14Main, A. R., Headley, J. V., Peru, K. M., Michel, N. L., Cessna, A. J., andMorrissey, C. A. (2014). Widespread use and frequent detection ofneonicotinoid insecticides in wetlands of Canada’s prairie pothole region. PLoSOne, 9(3):1–12. → pages 9, 41, 46121Malaj, E., von der Ohe, P. C., Grote, M., Ku¨hne, R., Mondy, C. P.,Usseglio-Polatera, P., Brack, W., and Scha¨fer, R. B. (2014). Organic chemicalsjeopardize the health of freshwater ecosystems on the continental scale.Proceedings of the National Academy of Sciences, 111(26):9549–9554. →pages 4Mallory, M. A. and Richardson, J. S. (2005). Complex interactions of light,nutrients and consumer density in a stream periphyton-grazer (tailed frogtadpoles) system. Journal of Animal Ecology, 74(6):1020–1028. → pages 25Marino, D. and Ronco, A. (2005). Cypermethrin and chlorpyrifos concentrationlevels in surface water bodies of the Pampa Ondulada, Argentina. Bulletin ofEnvironmental Contamination and Toxicology, 75(4):820–826. → pages 14Matthaei, C. D., Piggott, J. J., and Townsend, C. R. (2010). Multiple stressors inagricultural streams: interactions among sediment addition, nutrient enrichmentand water abstraction. Journal of Applied Ecology, 47(3):639–649. → pages 5,13, 14, 17, 41, 74, 75, 77, 85, 100, 105Matthaei, C. D., Weller, F., Kelly, D. W., and Townsend, C. R. (2006). Impacts offine sediment addition to tussock, pasture, dairy and deer farming streams inNew Zealand. Freshwater Biology, 51(11):2154–2172. → pages 4, 26, 42, 55,74, 84McArdle, B. H. and Anderson, M. J. (2001). Fitting multivariate models tocommunity data: a comment on Distance-Based Redundancy Analysis.Ecology, 82(1):290–297. → pages 20MEA (2005). Millenium Ecosystem Assessment: Synthesis Report. Island Press,Programme, United Nations Environment, Washington D.C., USA. → pages 2,11, 13Merritt, R. W. and Cummins, K. W. (1996). An Introduction to the Aquatic Insectsof North America. Kendall/Hunt Pub. Co., Dubuque, Iowa, 3rd ed. edition. →pages 78Mohr, S., Berghahn, R., Schmiediche, R., Hu¨bner, V., Loth, S., Feibicke, M.,Mailahn, W., and Wogram, J. (2012). Macroinvertebrate community responseto repeated short-term pulses of the insecticide imidacloprid. AquaticToxicology, 110-111:25–36. → pages 56Morrissey, C. A., Mineau, P., Devries, J. H., Sanchez-Bayo, F., Liess, M.,Cavallaro, M. C., and Liber, K. (2015). Neonicotinoid contamination of global122surface waters and associated risk to aquatic invertebrates: A review.Environment International, 74:291–303. → pages 4, 41, 56Nakagawa, S. and Cuthill, I. C. (2007). Effect size, confidence interval andstatistical significance: A practical guide for biologists. Biological Reviews,82:591–605. → pages 20, 48, 79Naman, S. M., Rosenfeld, J. S., and Richardson, J. S. (2016). Causes andconsequences of invertebrate drift in running waters: from individuals topopulations and trophic fluxes. Canadian Journal of Fisheries and AquaticSciences, 73:1292–1305. → pages 84Naman, S. M., Rosenfeld, J. S., Richardson, J. S., and Way, J. L. (2017). Speciestraits and channel architecture mediate flow disturbance impacts on invertebratedrift. Freshwater Biology, 62:340–355. → pages 86Nel, J. L., Roux, D. J., Abell, R., Ashton, P. J., and Cowling, R. M. (2009).Progress and challenges in freshwater conservation planning. AquaticConservation: Marine and Freshwater Ecosystems, 19:474–485. → pages 1Noble, B. F. (2010). Environmental effects and the tyranny of small decisions:towards meaningful cumulative effects assessment and management. NaturalResources and Environmental Studies Institute, 8:2–20. → pages 7Noble, B. F., Sheelanere, P., and Patrick, R. J. (2011). Advancing watershedcumulative effects assessment and management: lessons from the SouthSaskatchewan river watershed, Canada. Journal of Environmental AssessmentPolicy and Management, 13(4):567–590. → pages 8No˜ges, P., Argillier, C., Borja, A´., Garmendia, J. M., Hanganu, J., Kodesˇ, V.,Pletterbauer, F., Sagouis, A., and Birk, S. (2016). Quantified biotic and abioticresponses to multiple stress in freshwater, marine and ground waters. Scienceof the Total Environment, 540:43–52. → pages 8O’Callaghan, P., Jocque´, M., and Kelly-Quinn, M. (2015). Nutrient- andsediment-induced macroinvertebrate drift in Honduran cloud forest streams.Hydrobiologia, 758:75–86. → pages 84Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, D.,Minchin, P. R., O’Hara, R. B., Simpson, G. L., Solymos, P., Stevens, H.,Szoecs, E., and Wagner, H. (2016). Vegan: Community Ecology Package. Rpackage. → pages 21123Pablo, F., Krassoi, F. R., Jones, P. R., Colville, A. E., Hose, G. C., and Lim, R. P.(2008). Comparison of the fate and toxicity of chlorpyrifos–laboratory versus acoastal mesocosm system. Ecotoxicology and Environmental Safety,71:219–229. → pages 28, 103Patrick, C. J. and Fernandez, D. H. (2013). The β -richness of two detritivorecaddisflies affects fine organic matter export. Oecologia, 172(4):1105–15. →pages 75, 87Pennings, S. C. (1996). Testing for synergisms between chemical and mineraldefenses –A comment. Ecology, 77(6):1948–1950. → pages 6Pestana, J. L. T., Alexander, A. C., Culp, J. M., Baird, D. J., Cessna, A. J., andSoares, A. M. V. M. (2009). Structural and functional responses of benthicinvertebrates to imidacloprid in outdoor stream mesocosms. EnvironmentalPollution, 157:2328–2334. → pages 27, 31, 42, 57, 58Phillips, P. J. and Bode, R. W. (2004). Pesticides in surface water runoff insouth-eastern New York State, USA: seasonal and stormflow effects onconcentrations. Pest Management Science, 60(6):531–543. → pages 14Piggott, J. J., Lange, K., Townsend, C. R., and Matthaei, C. D. (2012). Multiplestressors in agricultural streams: a mesocosm study of interactions amongraised water temperature, sediment addition and nutrient enrichment. PLoSOne, 7(11):1–14. → pages 3, 74, 75, 85, 104Piggott, J. J., Niyogi, D. K., Townsend, C. R., and Matthaei, C. D. (2015a).Multiple stressors and stream ecosystem functioning: Climate warming andagricultural stressors interact to affect processing of organic matter. Journal ofApplied Ecology, 52(5):1126–1134. → pages 8, 17, 60, 85Piggott, J. J., Salis, R. K., Lear, G., Townsend, C. R., and Matthaei, C. D. (2015b).Climate warming and agricultural stressors interact to determine streammacroinvertebrate community dynamics. Global Change Biology,21(5):1887–1906. → pages 3, 4, 5, 26, 41, 54, 60, 84, 88, 100, 104, 105Piggott, J. J., Townsend, C. R., and Matthaei, C. D. (2015c). Reconceptualizingsynergism and antagonism among multiple stressors. Ecology and Evolution,5(7):1538–1547. → pages 5, 6, 28, 103Price, E., Prakash, B., Domino, M., Pepich, B. V., and Munch, D. (2009). Method527. Determination of selected pesticides and flame retardands in drining waterby solid phase extraction and capillary column gas chromatography/mass124spectrometry (GC/MS). Technical report, U.S. Environmental ProtectionAgency EPA, Office of ground water and drinking water, Cincinnati, Ohio. →pages 18Pusey, B. J., Arthington, A. H., and McLean, J. (1994). The effects of a pulsedapplication of Chlorpyrifos on macroinvertebrate comunities in an outdoorartificial stream system. Ecotoxicology and Environmental Safety, 27:221–250.→ pages 14Quinn, G. P. and Keough, M. J. (2002). Experimental Design and Data Analysisfor Biologists. Cambridge Univeristy Press, Cambridge, UK, 1 edition. →pages 20, 21, 49R Core Team (2016). R: a language and environment for statistical computing. →pages 21, 49Ramankutty, N., Evan, A. T., Monfreda, C., and Foley, J. A. (2010). Globalagricultural lands: croplands, 2000. → pages 13Rasmussen, J. J., Baattrup-Pedersen, A., Larsen, S. E., and Kronvang, B. (2011).Local physical habitat quality cloud the effect of predicted pesticide runofffrom agricultural land in Danish streams. Journal of EnvironmentalMonitoring, 13(4):943–950. → pages 30Rasmussen, J. J., Wiberg-Larsen, P., Baattrup-Pedersen, A., Monberg, R. J., andKronvang, B. (2012). Impacts of pesticides and natural stressors on leaf litterdecomposition in agricultural streams. The Science of the Total Environment,416:148–155. → pages 30, 31, 75Reid, L. M. (1993). Research and Cumulative Watershed Effects. Technicalreport, U.S. Department of Agriculture, Albany, CA. → pages 6, 9Reid, L. M. (1998). Cumulative Watershed Effects and Watershed Analysis. InNaiman, R. J. and Bilby, R. E., editors, River Ecology and Management:Lessons from the Pacific Coastal Ecoregion, pages 476–501. Springer-Verlag,New York, NY. → pages 9Rellstab, C. and Spaak, Æ. P. (2007). Starving with a full gut? Effect ofsuspended particles on the fitness of Daphnia hyalina. Hydrobiologia,594:131–139. → pages 55Richardson, J. S. (1991). Seasonal food limitation of detritivores in a montanestream: an experimental test. Ecology, 72(3):873–887. → pages 76125Richardson, J. S. and Wipfli, M. (2016). Getting quantitative about consequencesof cross-ecosystem resource subsidies on recipient consumers. CanadianJournal of Fisheries and Aquatic Sciences, 73:1609–1615. → pages 14Riseng, C. M., Wiley, M. J., Black, R. W., and Munn, M. D. (2011). Impacts ofagricultural land use on biological integrity: a causal analysis. EcologicalApplications, 21(8):3128–3146. → pages 3, 13, 74Riseng, C. M., Wiley, M. J., Seelbach, P. W., and Stevenson, R. J. (2010). Anecological assessment of Great Lakes tributaries in the Michigan Peninsulas.Journal of Great Lakes Research, 36(3):505–519. → pages 7Roessink, I., Koelmans, A. A., and Brock, T. C. M. (2008). Interactions betweennutrients and organic micro-pollutants in shallow freshwater model ecosystems.Science of the Total Environment, 406(3):436–442. → pages 8, 9Rohr, J. R., Kerby, J. L., and Sih, A. (2006). Community ecology as a frameworkfor predicting contaminant effects. Trends in Ecology & Evolution,21(11):606–613. → pages 60, 105Rosemond, A. D., Benstead, J. P., Bumpers, P. M., Gulis, V., Kominoski, J. S.,Manning, D. W. P., Suberkropp, K., and Wallace, J. B. (2015). Experimentalnutrient additions accelerate terrestrial carbon loss from stream ecosystems.Science, 347(6226):318–321. → pages 3, 14, 25, 83Sa´nchez-Bayo, F. (2011). Impacts of Agricultural Pesticides on TerrestrialEcosystems. In Sa´nchez-Bayo, F., Van den Brink, P. J., and Mann, R. M.,editors, Ecological Impacts of Toxic Chemicals, pages 63–87. Bentham SciencePublishers. → pages 4Sa´nchez-Bayo, F. and Goka, K. (2006). Influence of light in acute toxicitybioassays of imidacloprid and zinc pyrithione to zooplankton crustaceans.Aquatic Toxicology, 78(3):262–271. → pages 56, 101Sa´nchez-Bayo, F., Goka, K., and Hayasaka, D. (2016). Contamination of theAquatic Environment with Neonicotinoids and its Implication for Ecosystems.Frontiers in Environmental Science, 4:1–14. → pages 4, 41, 56, 57, 102Sandin, L. and Solimini, A. G. (2009). Freshwater ecosystem structure-functionrelationships: From theory to application. Freshwater Biology,54(10):2017–2024. → pages 105126Sanpera-Calbet, I., Chauvet, E., and Richardson, J. S. (2012). Fine sediment onleaves: shredder removal of sediment does not enhance fungal colonisation.Aquatic Sciences, 74(3):527–538. → pages xx, 16, 45, 135Saunders, D. L., Meeuwig, J. J., and Vincent, A. C. J. (2002). Freshwaterprotected areas: strategies for conservation. Conservation Biology,16(1):30–41. → pages 1Scha¨fer, R. B., Caquet, T., Siimes, K., Mueller, R., Lagadic, L., and Liess, M.(2007). Effects of pesticides on community structure and ecosystem functionsin agricultural streams of three biogeographical regions in Europe. The Scienceof the Total Environment, 382:272–85. → pages 31Scha¨fer, R. B., Ku¨hn, B., Malaj, E., Ko¨nig, A., and Gergs, R. (2016). Contributionof organic toxicants to multiple stress in river ecosystems. Freshwater Biology,61(12):2116–2128. → pages 4, 25, 30, 40, 105Scha¨fer, R. B., Von Der Ohe, P. C., Rasmussen, J. J., Kefford, B. J., Beketov, M.,Schulz, R., and Liess, M. (2012). Thresholds for the effects of pesticides oninvertebrate communities and leaf breakdwon in stream ecosystems.Environmental Science & Technology, 46:5134–5142. → pages 4, 27, 31Scheffer, M., Hosper, S. H., Meijer, M.-l., and Jeppesen, E. (1993). Alternativeequilibria in shallow lakes. Trends in ecology & evolution, 8(8):275–279. →pages 55Schneider, S. C., Kahlert, M., and Kelly, M. G. (2013). Interactions between pHand nutrients on benthic algae in streams and consequences for ecologicalstatus assessment and species richness patterns. The Science of the TotalEnvironment, 444:73–84. → pages 75Segner, H., Schmitt-Jansen, M., and Sabater, S. (2014). Assessing the Impact ofMultiple Stressors on Aquatic Biota: The Receptor’s Side Matters.Environmental Science & Technology, 48(14):7690–7696. → pages 60, 102,105Seitz, N. E., Westbrook, C. J., and Noble, B. F. (2011). Bringing science into riversystems cumulative effects assessment practice. Environmental ImpactAssessment Review, 31(3):172–179. → pages 5, 8Sheelanere, P., Noble, B. F., and Patrick, R. J. (2013). Institutional requirementsfor watershed cumulative effects assessment and management: Lessons from aCanadian trans-boundary watershed. Land Use Policy, 30(1):67–75. → pages 8127Shurin, J. B., Clasen, J. L., Greig, H. S., Kratina, P., and Thompson, P. L. (2012).Warming shifts top-down and bottom-up control of pond food web structureand function. Philosophical Transactions of the Royal Society,367(1605):3008–3017. → pages 41, 49, 54, 55, 60, 100Simon-Delso, N., Amaral-Rogers, V., Belzunces, L. P., Bonmatin, J. M.,Chagnon, M., Downs, C., Furlan, L., Gibbons, D. W., Giorio, C., Girolami, V.,Goulson, D., Kreutzweiser, D. P., Krupke, C. H., Liess, M., Long, E., Mcfield,M., Mineau, P., Mitchell, E. A., Morrissey, C. A., Noome, D. A., Pisa, L.,Settele, J., Stark, J. D., Tapparo, A., Van Dyck, H., Van Praagh, J., Van DerSluijs, J. P., Whitehorn, P. R., and Wiemers, M. (2015). Systemic insecticides(neonicotinoids and fipronil): trends, uses, mode of action and metabolites.Environmental Science and Pollution Research, 22(1):5–34. → pages 41Skagen, S. K., Melcher, C. P., and Haukos, D. A. (2008). Reducing sedimentationof depressional wetlands in agricultural landscapes. Wetlands, 28(3):594–604.→ pages 9, 100Smit, B. and Spanling, H. (1995). Methods for cumulative effects assessment.Environmental Impact Assessment Review, 15:81–106. → pages 7Smock, L. A. (1980). Relationships between body size and biomass of aquaticinsects. Freshwater Biology, 10:375–383. → pages 18, 47Smukler, S. M., Philpott, S. M., Jackson, L. E., Klein, A., DeClerck, F.,Winowiecki, L., and Palm, C. A. (2012). Ecosystem services in agriculturallandscapes. In Carter Ingram, J., DeClerck, F., and Rumbatitis del Rio, C.,editors, Integrating Ecology and Poverty Reduction: Ecological Dimensions,pages 17–51. Springer, New York, NY. → pages 2, 13Solomon, K. R., Williams, W. M., Mackay, D., Purdy, J., Giddings, J. M., andGiesy, J. P. (2014). Properties and uses of chlorpyrifos in the United States. InGiesy, J. P. and Solomon, K. R., editors, Reviews of environmentalcontamination and toxicology. Ecological risk assessment for chlorpyrifos interrestrial and aquatic systems in the United States., pages 13 – 34. SpringerOpen, New York, NY. → pages 14Spivak, A. C., Vanni, M. J., and Mette, E. M. (2011). Moving on up: can resultsfrom simple aquatic mesocosm experiments be applied across broad spatialscales? Freshwater Biology, 56:279–291. → pages 59Starner, K. and Goh, K. S. (2012). Detections of the neonicotinoid insecticideimidacloprid in surface waters of three agricultural regions of California, USA,1282010-2011. Bulletin of Environmental Contamination and Toxicology,88:316–321. → pages 45Statzner, B. and Beˆche, L. A. (2010). Can biological invertebrate traits resolveeffects of multiple stressors on running water ecosystems? Freshwater Biology,55:80–119. → pages 5, 6, 10Stehle, S. and Schulz, R. (2015). Agricultural insecticides threaten surface watersat the global scale. Proceedings of the National Academy of Sciences,112(18):5750–5755. → pages 4Stewart, R. I. A., Dossena, M., Bohan, D. A., Jeppesen, E., Kordas, R. L., Ledger,M. E., Meerhoff, M., Moss, B., Mulder, C., Shurin, J. B., Suttle, B., Thompson,R., Trimmer, M., and Woodward, G. (2013). Mesocosm Experiments as a Toolfor Ecological Climate-Change Research, volume 48. Elsevier Ltd., 1 edition.→ pages 59Strayer, D. L. and Dudgeon, D. (2010). Freshwater biodiversity conservation:recent progress and future challenges. Journal of the North AmericanBenthological Society, 29(1):344–358. → pages 1, 40Swank, W. T. and Bolstad, P. V. (1994). Cumulative effects of land use practiceson water quality. Hydrological, Chemical and Biological Processes ofTransformation and Transport of Contaminants in Aquatic Environments,219:409–421. → pages 7Tank, J. L., Rosi-Marshall, E. J., Griffiths, N. A., Entrekin, S. A., and Stephen,M. L. (2010). A review of allochthonous organic matter dynamics andmetabolism in streams. Journal of the North American Benthological Society,29(1):118–146. → pages 85Tomscha, S. A., Gergel, S. E., and Tomlinson, M. J. (2017). The spatialorganization of ecosystem services in river-floodplains. Ecosphere, 8(3). →pages 9, 74Townsend, C. R. and Thompson, R. M. (2007). Body size in streams:macroinvertebrate community size composition along natural andhuman-induced environmental gradients. In Hildrew, A. G., Raffaelli, D., andEdmonds-Brown, R., editors, Body Size: The Structure and Function of AquaticEcosystems, pages 77–97. Cambridge Univeristy Press, New York, NY. →pages 29129Townsend, C. R., Uhlmann, S. S., and Matthaei, C. D. (2008). Individual andcombined responses of stream ecosystems to multiple stressors. Journal ofApplied Ecology, 45:1810–1819. → pages 2, 5, 6, 7, 8, 17, 77, 104Traas, T. P., Janse, J. H., Van den Brink, P. J., Brock, T. C. M., and Aldenberg, T.(2004). A freshwater food web model for the combined effects of nutrients andinsecticide stress and subsequent recovery. Environmental Toxicology andChemistry, 23(2):521–9. → pages 14, 29Udawatta, R. P., Garrett, H. E., and Kallenbach, R. (2008). Agroforestry buffersfor nonpoint source pollution reductions from agricultural watersheds. Journalof Environmental Quality, 40(3):800–806. → pages 107Van den Brink, P. J., Crum, S. J. H., Gylstra, R., Bransen, F., Cuppen, J. G. M.,and Brock, T. C. M. (2009). Effects of a herbicide-insecticide mixture infreshwater microcosms: risk assessment and ecological effect chain.Environmental Pollution, 157(1):237–49. → pages 27Van den Brink, P. J., Van Wijngaarden, R. P. A., Lucassen, W. G. H., Brock, T.C. M., and Leeuwangh, P. (1996). Effects of the insecticide dursban R© 4E(active ingredient chlorpyrifos) in outdoor experimental ditches: II. Invertebratecommunity responses and recovery. Environmental Toxicology and Chemistry,15(7):1143–1153. → pages 27, 101Van Dijk, T. C., Van Staalduinen, M. A., and Van der Sluijs, J. P. (2013).Macroinvertebrate decline in surface water polluted with imidacloprid. PLoSOne, 8(5):1–10. → pages 58Vinebrooke, R. D., Cottingham, K. L., Norberg, J., Scheffer, M., Dodson, S. I.,Maberly, S. C., Sommer, U., Oikos, S., Mar, F., and Cottingham, L. (2004).Impacts of multiple stressors on biodiversity and ecosystem functioning: therole of species co-tolerance. Oikos, 104:451–457. → pages 10, 29, 58, 103,104, 105Vo¨ro¨smarty, C. J., McIntyre, P. B., Gessner, M. O., Dudgeon, D., Prusevich, A.,Green, P., Glidden, S., Bunn, S. E., Sullivan, C. A., Liermann, C. R., andDavies, P. M. (2010). Global threats to human water security and riverbiodiversity. Nature, 468(7321):334–334. → pages 1, 2, 3, 4, 11, 40, 105Wagenhoff, A., Clapcott, J. E., Lau, K. E., Lewis, G. D., and Young, R. G. (2016).Identifying congruence in stream assemblage thresholds in response to nutrientand sediment gradients for limit setting. Ecological Applications,27(2):469–484. → pages 54, 60130Wagenhoff, A., Lange, K., Townsend, C. R., and Matthaei, C. D. (2013). Patternsof benthic algae and cyanobacteria along twin-stressor gradients of nutrientsand fine sediment: a stream mesocosm experiment. Freshwater Biology,58:1849–1863. → pages 4, 5, 74Wagenhoff, A., Townsend, C. R., and Matthaei, C. D. (2012). Macroinvertebrateresponses along broad stressor gradients of deposited fine sediment anddissolved nutrients: a stream mesocosm experiment. Journal of AppliedEcology, 49(4):892–902. → pages 14, 17, 41, 42, 74, 75, 77, 84, 100, 101Wagner, E. J., Bosakowski, T., and Intelmann, S. (1997). Combined effects oftemperature and high pH on mortality and the stress response of rainbow troutafter stocking. Transactions of the American Fisheries Society,126(6):985–998. → pages 8Ward, S., Arthington, A. H., and Pusey, B. J. (1995). The effects of a chronicapplication of Chlorpyrifos on the macroinvertebrate fauna in an outdoorartificial stream system: species responses. Ecotoxicology and EnvironmentalSafety, 30:2–23. → pages 27Warren, N., Allan, I. J., Carter, J. E., House, W. A., and Parker, A. (2003).Pesticides and other micro-organic contaminants in freshwater sedimentaryenvironmentsa review. Applied Geochemistry, 18(2):159–194. → pages 9, 28,103Waters, T. E. (1977). Secondary production in inland waters. Advances InEcological Research, 10:91–164. → pages 49Waters, T. F. (1995). Sedimentation in Streams: Sources, Biological Effects, andControl. American Fisheries Society, Bethesda, Maryland. → pages 3Wentworth, C. K. (1922). A scale of grade and class terms for clastic sediments.The Journal of Geology, 30(5):377–392. → pages 3Wepener, V., van Dyk, C., Bervoets, L., O’Brien, G., Covaci, A., and Cloete, Y.(2011). An assessment of the influence of multiple stressors on the Vaal River,South Africa. Physics and Chemistry of the Earth, 36(14-15):949–962. →pages 7Williams, W. M., Giddings, J. M., Purdy, J., Solomon, K. R., and Giesy, J. P.(2014). Exposures of aquatic organisms to the organophosphorus insecticide,chlorpyrifos resulting from use in the United States. In Giesy, J. P. andSolomon, K. R., editors, Reviews of Environmental Contamination and131Toxicology. Ecological Risk Assessment for Chlorpyrifos in Terrestrial andAquatic Systems in the United States, pages 77 – 117. Springer Open, NewYork, NY. → pages 14, 18, 27Wipfli, M. S., Richardson, J. S., and Naiman, R. J. (2007). Ecological linkagesbetween headwaters and downstream ecosystems: transport of organic matter,invertebrates, and wood down headwaters channels. Journal of the AmericanWater Resources Association, 43(1):72–85. → pages 74, 75Wood, P. J. and Armitage, P. D. (1997). Biological effects of fine sediment in thelotic environment. Environmental Management, 21(2):203–217. → pages 3, 4,14, 26, 42, 55, 74, 84Wood, S. L. R. and Richardson, J. S. (2009). Impact of sediment and nutrientinputs on growth and survival of tadpoles of the Western Toad. FreshwaterBiology, 54(5):1120–1134. → pages 100Woodward, G., Brown, L. E., Edwards, F. K., Hudson, L. N., Milner, A. M.,Reuman, D. C., and Ledger, M. E. (2012a). Climate change impacts inmultispecies systems: drought alters food web size structure in a fieldexperiment. Philosophical Transactions of the Royal Society,367(1605):2990–2997. → pages 14Woodward, G., Gessner, M. O., Giller, P. S., Gulis, V., Hladyz, S., Lecerf, A.,Malmqvist, B., McKie, B. G., Tiegs, S. D., Cariss, H., Dobson, M., Elosegi, A.,Ferreira, V., Grac¸a, M. A. S., Fleituch, T., Lacoursiere, J. O., Nitorescu, M.,Pozo, J., Risnoveanu, G., Schindler, M. H., Vadineanu, A., Vought, L. B., andChauvet, E. (2012b). Continental-scale effects of nutrient pollution on streamecosystem functioning. Science, 336:1438–1440. → pages 3, 25, 74, 83Wooster, D. E., Miller, S. W., and Debano, S. J. (2012). An examination of theimpact of multiple disturbances on a river system: taxonomic metrics versusecological traits. River Research and Applications, 28:1630–1643. → pages 7Young, R. G., Matthaei, C. D., and Townsend, C. R. (2008). Organic matterbreakdown and ecosystem metabolism: functional indicators for assessing riverecosystem health. Journal of the North American Benthological Society,27(3):605–625. → pages 54, 57, 60, 84132Appendix ASupporting Materials133Table A.1: ANOVA summary of linear mixed effects models to test impactsof stressor treatments (fixed effects) and time (week: fixed effect) ondissolved oxygen, temperature, pH, and conductivity, with individualmicrocosm treated as a random effect. Symbols are used to representsignificance of the effects according to the ANOVAs: *** P < 0.001,**P < 0.01, *P < 0.05, ·P < 0.1.Factor d.f. F-value P-valueTemperatureNutrients 1,24 2.56 0.123Sediment 1,24 0.57 0.458Insecticide 1,24 2.04 0.166Week 1,31 1248.89 <0.0001 ***Nutrients * sediment 1,24 0.26 0.619Nutrients * insecticide 1,24 1.66 0.210Sediment * insecticide 1,24 0.28 0.601Nutrient * sediment * insecticide 1,24 0.25 0.620pHNutrients 1,24 0.06 0.814Sediment 1,24 4.17 0.052Insecticide 1,24 2.45 0.131Week 1,31 25.35 <0.0001 ***Nutrients * sediment 1,24 0.28 0.603Nutrients * insecticide 1,24 0.00 0.983Sediment * insecticide 1,24 0.49 0.491Nutrient * sediment * insecticide 1,24 0.02 0.881ConductivityNutrients 1,24 5.26 0.031Sediment 1,24 245.46 <0.0001 ***Insecticide 1,24 0.01 0.921Week 1,31 7.90 0.009 **Nutrients * sediment 1,24 3.52 0.073Nutrients * insecticide 1,24 0.27 0.607Sediment * insecticide 1,24 2.05 0.165Nutrient * sediment * insecticide 1,24 0.53 0.473Dissolved oxygenNutrients 1,24 3.05 0.093Sediment 1,24 0.92 0.348Insecticide 1,24 2.67 0.116 .Week 1,31 308.32 <0.0001 ***Nutrients * sediment 1,24 0.26 0.612Nutrients * insecticide 1,24 10.98 0.003 **Sediment * insecticide 1,24 0.18 0.677Nutrient * sediment * insecticide 1,24 1.40 0.248134Figure A.1: Schematic of the stream microcosm set-up. Darker rectangle inthe middle of the tank represents a horizontal glass plate introduced topartially divide the tank into an upper and a lower section. A bubbleralong with a plastic deflector placed in one end of the tank promotedcircular flow in the direction indicated by the blue arrows. Redrawnfrom Sanpera-Calbet et al. (2012).135Figure A.2: Untransformed univariate responses of gravel invertebratesacross the treatments on day 15. Boxes are drawn around upper andlower quartiles with whiskers indicating maximum and minimum val-ues, dark thick lines indicate the median, and points denote outliers.Y-axis does not always start in zero.136Figure A.3: Untransformed univariate responses of leaf invertebrates acrossthe treatments on day 15. Boxes are drawn around upper and lowerquartiles with whiskers indicating maximum and minimum values, darkthick lines indicate the median, and points denote outliers. Y-axis doesnot always start in zero.137Figure A.4: Untransformed ecosystem processes across the treatments onday 15. Boxes are drawn around upper and lower quartiles withwhiskers indicating maximum and minimum values, dark thick linesindicate the median, and points denote outliers. Y-axis does not alwaysstart in zero.138"@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2018-02"@en ; edm:isShownAt "10.14288/1.0362384"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Forestry"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "Attribution-NonCommercial-NoDerivatives 4.0 International"@* ; ns0:rightsURI "http://creativecommons.org/licenses/by-nc-nd/4.0/"@* ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Cumulative effects of multiple agricultural stressors on freshwater ecosystems"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/64138"@en .