Synergistic interactions among growing stressors increase risk to an Arctic ecosystem

Oceans provide critical ecosystem services, but are subject to a growing number of external pressures, including overfishing, pollution, habitat destruction, and climate change. Current models typically treat stressors on species and ecosystems independently, though in reality, stressors often interact in ways that are not well understood. Here, we use a network interaction model (OSIRIS) to explicitly study stressor interactions in the Chukchi Sea (Arctic Ocean) due to its extensive climate-driven loss of sea ice and accelerated growth of other stressors, including shipping and oil exploration. The model includes numerous trophic levels ranging from phytoplankton to polar bears. We find that climate-related stressors have a larger impact on animal populations than do acute stressors like increased shipping and subsistence harvesting. In particular, organisms with a strong temperature-growth rate relationship show the greatest changes in biomass as interaction strength increased, but also exhibit the greatest variability. Neglecting interactions between stressors vastly underestimates the risk of population crashes. Our results indicate that models must account for stressor interactions to enable responsible management and decision-making.

The presence, magnitude, and direction of an interactive effect of multiple simultaneous 231 stressors are dependent on the biological variable being measured. Several meta-analyses 232 investigating pairwise stressors showed some consistency in terms of the type of interactive 233 effect and measured biological variables. Harvey et al. [18] showed that multiple stressors acted 234 synergistically on calcification, photosynthesis, reproduction, and survival while growth did not. 235 Kroeker et al. [19] report that survival, growth, and development had pairwise stressors acting 236 synergistically on them while calcification did not. Przeslawski et al. [21] report that calcifying 237 organisms are more sensitive to interactive effects than non-calcifying organisms measured for 238 the same biological variables. As the magnitude of the interactive effect can vary depending on 239 the choice of biological variable, it is important to consider this choice. A change in mortality 240 under multiple simultaneous stressors is an ultimate integrative measure, as all sub-lethal 241 changes will be incorporated in this biomass loss. However, other variables, such as 242 calcification, do not necessarily lead to a decrease in population but is expected to affect 243 calcifying organisms negatively.  Table 4), especially for those Arctic groups that had limited 254 information available. The reason why we included estimates from other geographical regions is 255 that the calculation of the interactive value (described below) results in a unit less interaction 256 coefficient, and the results we obtained did not vary much from the interaction coefficients 257 obtained from Arctic studies. 258 In order to calculate the interaction coefficients for pairwise stressors for different biological 259 variables, a study needed to include a control treatment, treatments of both stressors alone, and a 260 treatment of both stressors combined. The average effect of a particular biological variable was 261 determined, and the difference between the two single treatments and the control treatment was 262 used to calculate the additive estimate. The additive effect was determined by the direction of the 263 effect of two single stressor estimates. If both stressors yielded a higher estimate than the control, 264 the additive effect was calculated by adding the absolute differences between the control and 265 each single treatment to control, while if the two single stressor estimates yielded lower 266 estimates than the control, the additive effect was calculated by subtracting the absolute 267 differences between the control and the two single treatments from control. When the two single 268 stressor estimates were in the opposite direction, this was incorporated in the calculation of the 269 additive effect. The additive effect should be equal to the estimate of the combined treatment if 270 there was no interactive effect present between the two treatments.

271
The interaction coefficient was calculated as follows: where represents the interaction coefficient, µ AB represents the estimate of the combined 274 treatment, µ A+B represents the estimate of the additive effects, µ A represents the estimate of 275 treatment A, and µ B represents the estimate of treatment B.

276
Several studies reported on the effects of more than one biological variable, and in some 277 cases multiple treatment levels in one treatment were used. For example, Kunz et al. [22] 278 investigated the effects on survival, oxygen consumption, growth rate, and relative feed 279 conversion in Arctic cod under 0, 3, 6, and 10°C in combination with two pH levels (current pH 280 8.05 and future pH 7.6), which were all included to inform the literature range and highlight the 281 uncertainty around the interaction coefficient.

284
Sea ice in the Arctic is changing both in extent and thickness as a result of global warming 285 [23]. Sea ice has an important function in the Arctic; it functions as a habitat (e.g. ice algae), 286 refuge (e.g., juvenile Arctic cod), resting place (e.g. seals and walruses), hunting platform (e.g. 287 polar bears), reproductive site (e.g. seals and walruses; also in combination with snow, especially 288 for ringed seals which form the majority of polar bear diet), and it influences the soundscape in 289 the water by dampening the noise. Therefore, the loss of ice will have a significant impact on the   The Arctic Ocean currently is and has been relatively inaccessible to ships for large parts of 318 the year. The presence of sea ice, therefore, has played a protective role for many larger 319 organisms, like whales, which can be harmed by ships through physical impact as well as noise  Other whale species that live in industrial shipping areas can be informative about some of 332 the shipping-related impacts and potential future for Arctic whales. Noise pollution is however, 333 still a difficult and understudied aspect of shipping in terms of quantitative damage. While there 334 are many qualitative impacts described, such as displacement, increased stress, and even some 335 noise pollution has been linked to mortality, it remains difficult to quantity the full range of impacts that result in abundance or biomass loss. Below, the stressors are investigated in more 337 detail to illustrate how the impacts used in the OSIRIS model were derived.  incorporating the population of whales that is hit by a ship, the chance a hit is lethal, which is 393 dependent on the vessel speed, and the amount of shipping present in the Arctic. This is formally 394 represented as follows: where B i represents the relative biomass loss for species i as a result of ship strikes, H i represents 397 the proportion of the population of species i that is hit by ships, L represents the chance that the 398 hit is lethal, which in itself is dependent on ship speed and is looked up in the logistic regression 399 estimates, and S represents the relative amount of shipping in the Arctic, where 1 represents 400 current shipping levels.

401
The population of bowhead whales that is hit by a ship is difficult to determine, but based on