Synergistic interactions among growing stressors increase risk to an Arctic ecosystem

The study addresses how multiple, co-occurring environmental and anthropogenic stressors interact to affect marine ecosystems, focusing on the Chukchi Sea in the rapidly changing Arctic Ocean. Traditional models often assume additive effects, yet empirical evidence shows frequent non-additive (synergistic or antagonistic) interactions among stressors. The research goal is to quantify how suites of stressors—particularly chronic climate-related drivers (sea ice loss, warming, acidification) and acute human-related drivers (shipping noise and strikes, subsistence harvesting)—interact to influence food web dynamics and the probability of population collapse. The work is motivated by accelerated Arctic change, expanding human activity, and the need to anticipate tipping points and inform ecosystem-based management.

Prior ecosystem models commonly sum individual stressor effects, multiply stressors without isolating interaction terms, or apply ad hoc weightings. Meta-analyses show additive interactions account for only about 26% of cases, with synergistic and antagonistic interactions occurring in roughly 36% and 38% of studies, respectively, indicating non-linear interactions are common. Experimental quantification tends to focus on single or pairwise stressors, leaving limited understanding of multiple co-occurring stressors across full food webs. Existing models can predict impacts of individual stressors at fine scales, but failing to account for synergies can underestimate risks due to non-linear impacts. These gaps justify a modeling framework that explicitly represents interaction terms among multiple chronic and acute stressors across trophic levels.

The study applies the OSIRIS (Ocean System Interactions, Risks, Instabilities, and Synergies) network model to the Chukchi Sea food web over 2020–2040. Nodes represent biotic (biomass of species or functional groups from phytoplankton to polar bears) and abiotic components, linked by trophic and interaction pathways. The model assumes equilibrium initial conditions and uses literature-based parameters for growth, mortality, and interactions (Supplementary Tables). Forcing variables include seasonal light, sea surface temperature (SST), sea ice concentration and open water duration, ocean pH, nutrient flux through Bering Strait, shipping noise, probability of ship strikes, and subsistence harvest; these are extrapolated from recent observations for 20-year projections. Stressors are categorized as chronic (SST, pH, sea ice/open water, inflow) and acute (harvest, ship noise, ship strikes). Node-specific pairwise stressor interaction terms are defined for targeted nodes (no interactions for ice algae and phytoplankton). Interaction strengths span lower (antagonistic) to upper (synergistic) bounds per organism (Supplementary Table 4). All defined interactions are simultaneously and proportionally scaled across 41 discrete interaction-strength levels. For each level, 200 simulations are run using Latin hypercube sampling across parameter ranges to capture uncertainty. Outputs for the final simulation year are summarized as means and standard deviations of biomass per node. Probability of population collapse is defined as the fraction of simulations in which biomass falls below 10% of its initial value after 20 years. Three experiment sets are run: (1) all interactions included, (2) chronic-only interactions, and (3) acute-only interactions. A baseline scenario uses best-estimate parameters and interaction strengths. Baseline assumptions include: mean annual sea ice concentration declines by 14%; open water season lengthens from 135 to 165 days; max annual SST increases from 3 to 4 °C; pH declines from 8.1 to 8.0; nutrient flux scales with Bering Strait volume flux; shipping noise and strikes increase ~sixfold over 20 years.

• Baseline environmental change (2020–2040): sea ice concentration −14%; open water season 135→165 days; max SST 3→4 °C; pH 8.1→8.0; shipping noise/strikes ~6× increase.
• Primary producers: ice algae −20%; picophytoplankton +41%; planktonic diatoms +33%.
• Lower trophic consumers: small zooplankton +1.2%; large zooplankton +16%; clams +1.3%; amphipods +8.4%.
• Fish, birds, and mammals: Arctic cod −21%; Pacific salmon −8.9%; seabirds −10%; seals −9.1%; walrus −10.1%; polar bears −10.2%; bowhead whales −9.3%; gray whales −10.6%.
• Drivers: Increases at lower trophic levels tied to warmer SST and higher phytoplankton biomass; Arctic cod declines driven more by reduced sea ice/open water dynamics than SST; marine mammals and seabirds decline due to sea ice loss, prey changes, and acute stressors (noise, ship strikes, subsistence for bowheads).
• Interaction-strength experiments: As interaction strength becomes more synergistic, two patterns emerge: (a) biomass declines slightly for seabirds, seals, polar bears, Arctic cod, bowhead and gray whales, and walrus; (b) biomass increases substantially for small and large zooplankton, amphipods, clams, and Pacific salmon. Positive responses generally correspond to positive SST-growth relationships.
• Chronic vs acute: For groups with increasing biomass at higher interaction strength, chronic-stressor interactions dominate responses; organisms with declining biomass often respond more to acute stressors (e.g., whales to ship strikes/noise), while seals and Arctic cod respond strongly to chronic ice loss.
• Variability: Biomass variability increases with interaction strength when all interactions are included, especially for small zooplankton (~3-fold), large zooplankton (~2-fold), Pacific salmon (~5.5-fold), amphipods (~5-fold), and clams (~3-fold). Variability increase is negligible when only acute interactions are included, indicating chronic interactions drive variability.
• Collapse risk: With all interactions, several organisms have ≥0.1 probability of collapse at 20 years, increasing with interaction strength (clams, Arctic cod, seals, walrus). Amphipods and Pacific salmon rise from low risk at antagonistic levels to 0.3–0.4 at high synergy, representing ~40-fold and ~3.7-fold increases over baseline, respectively. Sensitivity of collapse risk to interaction strength disappears when only acute stressors are included; chronic stressor interactions drive the increased risk with synergy. At low interaction strengths with acute-only runs, some taxa can show higher risk due to antagonistic chronic interactions otherwise offsetting effects.
• Network properties: No relationship between number of connections or number/type of non-linear interactions per organism and collapse risk, indicating results are not artifacts of model connectedness.

The modeling demonstrates that synergistic interactions among multiple stressors substantially elevate risks in the Chukchi Sea food web, often more than doubling the probability of population collapse relative to models without non-linear interactions. Chronic stressors (warming, acidification, sea ice loss) generally outweigh acute stressors (ship noise/strikes, subsistence harvest) in determining biomass trends, variability, and collapse risks. Higher trophic, large-bodied organisms tend to decline with stronger synergies, whereas lower trophic levels often increase, likely due to positive thermal responses and enhanced productivity with ice loss. Increasing synergy amplifies internal model variability, making outcomes less predictable even under fixed forcing trajectories. The lack of correlation between network connectedness and collapse risk suggests biological sensitivities and stressor interactions, rather than model topology, drive outcomes. These findings underscore the need to incorporate non-linear, synergistic stressor interactions into ecosystem assessments and management to avoid underestimating risks and to better anticipate tipping points under accelerating Arctic change.

By applying the OSIRIS interaction-explicit framework to the Chukchi Sea, the study shows that neglecting non-linear interactions between co-occurring stressors can severely underestimate the risk of population collapse and mischaracterize ecosystem trajectories. Chronic stressor interactions dominate impacts and increasingly amplify both biomass changes and variability as synergy strengthens. The work highlights critical implications for ecosystem-based management in the Arctic, emphasizing precaution and the need to limit growth of multiple stressors. Future research should identify stressor pairs most prone to synergy, improve empirical constraints on organismal responses (especially to noise and ship strikes), and develop early-warning indicators to anticipate tipping points and prioritize mitigation levers.

Uncertainties include limited empirical data on the effects of some acute stressors (e.g., ship strikes and noise) on specific species such as bowhead whales, difficulty encapsulating diverse noise impacts into single parameters, and uncertainties in estimates of organismal responses and future stressor magnitudes and interactions. The model relies on extrapolated forcing data and assumes equilibrium initial conditions. While interaction structures are literature-informed, some interactions remain poorly constrained, and results depend on assumed bounds of interaction strengths. Thus, actual future impacts may be larger or differ where unmodeled or stronger synergies occur.