Consistent effects of pesticides on community structure and ecosystem function in freshwater systems

Freshwater ecosystems are highly biodiverse and provide essential services but are threatened by pervasive pesticide contamination. Two key challenges impede prediction of ecosystem responses to pesticides: (1) whether effects are consistent across the many registered chemicals, particularly within pesticide classes (shared chemical structures) or types (shared target pests), and (2) how pesticides fit within biodiversity–ecosystem function relationships, given that anthropogenic drivers often act non-randomly across multiple trophic levels via direct and indirect pathways. The study’s objectives were to: (1) evaluate consistency of effects across pesticide types, classes, and individual pesticides on ecosystem processes and community structure; (2) assess whether effects arise from sublethal, non-target impacts versus changes in abundance of targeted taxa; and (3) test whether changes in composition, abundance, and richness of functional groups mediate ecosystem functional responses. Three hypotheses guided the work: (i) ecosystem processes respond consistently within pesticide types due to functional redundancy among taxa; (ii) communities respond consistently within classes due to taxa-specific sensitivities; and (iii) pesticide effects on ecosystem functions are mediated by changes in functional group abundance, composition, and richness.

Prior work has established links between biodiversity and ecosystem functioning and highlighted that anthropogenic drivers (e.g., climate change, nutrients, and synthetic chemicals) can alter multiple trophic levels via direct and indirect pathways. Pesticide risk assessments often rely on single-species toxicity tests or QSAR models, which may not capture taxa-specific sensitivities, species interactions, or indirect effects. Previous studies indicate that insecticides and herbicides can have predictable effects on aquatic taxa (e.g., differential sensitivities of cladocerans vs. copepods; strong bottom-up effects of herbicides). However, generality across many chemicals remains unclear. This study builds on these insights by testing consistency across pesticide classes/types in whole-community mesocosms and explicitly linking functional group changes to ecosystem processes.

Design: A randomized-block outdoor mesocosm experiment (Russell E. Larsen Agricultural Research Center, PA, USA) used 72 cattle-tank ponds (1100 L, 60% shade cloth). Eighteen treatments were each replicated four times: 12 pesticides (nested within 2 types and 4 classes), 4 simulated-pesticide manipulations, and 2 controls (water and solvent). Pesticide structure: two types (insecticides, herbicides), each with two classes—organophosphate insecticides (chlorpyrifos 64 µg/L nominal [60 measured], malathion 101 [105], terbufos 171 [174]); carbamate insecticides (aldicarb 91 [84], carbaryl 219 [203], carbofuran 209 [227]); chloroacetanilide herbicides (acetochlor 123 [139], alachlor 127 [113], metolachlor 105 [114]); triazine herbicides (atrazine 102 [117], simazine 202 [180], propazine 106 [129]). Concentrations were standardized to environmentally relevant estimated environmental concentrations (EPA GENEEC v2) and verified by chemical analysis 1 h post-application. Communities: Mesocosms received 800 L water, 300 g mixed hardwood leaves, and inocula of phyto-, periphyton, and zooplankton from local ponds. Prior to pesticide application, each mesocosm received common pond taxa at natural densities: snails (Helisoma/Planorbella trivolvis, Physa gyrina), larval anurans (Hyla versicolor, Lithobates palustris, L. clamitans), insect predators (Anax junius dragonfly larvae, Belostoma flumineum water bugs, Hydrochara sp. beetles, Notonecta undulata backswimmers), and larval salamanders (Ambystoma maculatum). Simulated-pesticide treatments: top-down and bottom-up manipulations intended to mimic lethal effects—(i) simulated insecticides: doubling zooplankton predators (A. maculatum and N. undulata) or direct zooplankton removal; (ii) simulated herbicides: doubling large herbivores (snails, tadpoles) or tripling shading (three layers of 60% shade cloth). Duration: 4 weeks (June–July). Measurements: - Primary producers: periphyton (chlorophyll-a on caged/uncaged tiles) and phytoplankton (chlorophyll-a from water samples). - Water chemistry and metabolism: pH, dissolved oxygen at dusk and dawn to estimate whole-system respiration (DO dusk − DO dawn), and turbidity (ordinal clarity score 1–5). - Decomposition: percent mass remaining of hardwood leaf packs. - Biota: zooplankton community (Daphnia, Diaphanasoma, Chydorus, Bosmina, Diaptomus, Cyclops; counts from integrated column samples), survival and mass of vertebrates/invertebrates, snail reproduction (eggs/hatchlings per H. trivolvis). End-of-experiment counts of remaining animals were conducted. Statistical analyses: Nested PERMANOVAs quantified variance explained by pesticide type, class, and individual pesticide, accounting for block effects. Separate models analyzed (1) ecosystem processes (pH, respiration, decomposition, turbidity, phytoplankton, accessible periphyton), (2) zooplankton community (square-root transformed densities; Bray–Curtis), (3) tri-trophic community (survival, growth, reproduction of predators/herbivores plus algae; accessible and inaccessible periphyton), and (4) simplified tri-trophic functional groups (algae, herbivores, predators). Visualization used dbRDA, PCoA, and two-way cluster diagrams. Structural equation modeling (piecewiseSEM) tested whether functional group composition (PCoA axis 1), abundance, and richness mediated pesticide effects on primary productivity (phytoplankton, accessible periphyton) and respiration. Herbicide SEMs emphasized bottom-up effects on algae and non-target effects on zooplankton; insecticide SEMs emphasized top-down effects on predators and zooplankton. Non-significant paths were pruned; model fits were evaluated with Fisher’s C.

- Ecosystem functions: Pesticide type explained 46% of variation in multivariate ecosystem responses (vs. 12% by individual pesticides). Herbicides decreased phytoplankton primary production and whole-system respiration; reduced phytoplankton likely increased light and periphyton production. Herbicides decreased pH (increased acidity), potentially via increased dissolved inorganic carbon from decomposing phytoplankton. Insecticides increased phytoplankton production, reduced periphyton production (light limitation via shading), and increased respiration; pH increased with phytoplankton growth. Decomposition of leaf litter was not strongly affected by pesticide type. - Zooplankton community: Pesticide type explained 44.2% of variance and class 18.8%. Insecticides largely eliminated cladocerans and increased copepods (competitive release), consistent with reduced grazing efficiency and increased phytoplankton. Herbicides reduced zooplankton abundance without strong compositional shifts; triazines exerted stronger bottom-up effects than chloroacetanilides, likely due to longer environmental persistence (soil half-lives ~110–146 vs. 14–26 days). - Tri-trophic community: Variance was distributed roughly equally among type, class, and individual pesticides. Insecticides reduced insect predator survival broadly (notably organophosphates and several carbamates), with aldicarb as an exception, and increased survival/growth of amphibian and snail prey (predator release). - Simulated treatments: Physical manipulations rarely reproduced the magnitude/specificity of actual pesticide effects, likely due to rapid population dynamics and complexity of disturbance effects. - Mediation by functional groups (SEMs): Herbicides decreased respiration and phytoplankton productivity via bottom-up reductions in phytoplankton abundance. Insecticides increased phytoplankton productivity and respiration via top-down changes in zooplankton composition and abundance (not richness). - Functional-role aggregation: Grouping responses by functional roles (algae, herbivores, predators) increased the proportion of variance explained by type to 29%, compared to 17.6% (class) and 17.3% (individual pesticide), supporting the utility of functional-group based predictions.

Findings support the hypotheses that ecosystem processes respond consistently within pesticide types and that community responses exhibit partial consistency within classes. Herbicides exert bottom-up effects by suppressing phytoplankton, lowering respiration, and shifting production to periphyton via increased light availability. Insecticides alter zooplankton composition—removing efficient cladoceran grazers and favoring copepods—producing top-down increases in phytoplankton and whole-system respiration, while reducing periphyton through shading. Structural equation models show that functional group abundance and composition (more than richness) mediate these ecosystem responses. Together, results suggest that the complexity of predicting pesticide impacts can be reduced by focusing on pesticide type/class and the responses of organismal functional groups, rather than individual species. This approach addresses the non-random, multi-trophic nature of anthropogenic disturbances and highlights why single-species or QSAR-based assessments may miss key indirect effects and community interactions.

The study demonstrates consistent, predictable effects of herbicides and insecticides on ecosystem functions across diverse freshwater mesocosm communities and links these effects to changes in functional group abundance and composition. By showing that pesticide type and class often explain more variation than individual chemicals and that functional-role aggregation increases predictive power, the work provides a path to streamline ecological risk assessment toward chemical classes/types and functional groups. Future research should: (1) extend analyses to additional pesticides and modes of action to test generality; (2) include diverse communities and ecosystems to evaluate how genetics, community composition/complexity, co-occurring stressors, and exposure history influence generality; and (3) develop standardized whole-community/ecosystem tests to detect exceptions and refine predictive frameworks.

- Scope limited to single-chemical exposures; mixtures were not tested, though mixtures commonly occur in the environment. - Simulated-pesticide manipulations did not consistently match real pesticide effects, likely due to rapid population dynamics and difficulty replicating complex disturbance pathways. - Short experimental duration (4 weeks) and pesticide persistence profiles constrain observation of multigenerational or long-term effects; however, authors argue multigenerational effects were unlikely given organism generation times and pesticide persistence. - Generality may vary with genetic makeup of populations, community composition/complexity, presence of other stressors, and prior exposure history. - Structural similarities among classes (QSAR-style groupings) did not explain observed community/ecosystem patterns, underscoring limits of structure-based predictions without species interactions. - Decomposition showed limited responsiveness and was excluded from SEMs, potentially overlooking subtle detrital pathway effects.