Climate change, encompassing both long-term warming trends and short-term extreme events like marine heatwaves (MHWs), profoundly affects marine ecosystems and human well-being. While the impacts of long-term warming have been extensively studied, the effects of episodic MHWs remain relatively less understood. MHWs, characterized by their intensity and variability, pose significant management challenges. Mobile species often respond to environmental changes by shifting their geographic ranges to maintain suitable conditions. This is well-documented for long-term warming and El Niño events, but less so for the rapid warming during MHWs. Temperatures during MHWs can mirror projected future conditions, providing insights into potential species redistributions in the coming decades. Previous studies relying on observational data have shown poleward or vertical shifts towards cooler waters, but these data are often patchy and limited in scope, lacking comparisons of multi-species responses across multiple MHWs. Statistical models offer a way to overcome data limitations by interpolating across space, time, and taxa, allowing for inferences on unobserved locations and events. They also allow for accounting for complex physical and biogeochemical changes beyond temperature increases. This study focuses on the Northeast Pacific Ocean, a region experiencing some of the most intense and prolonged MHWs on record, and a biodiversity hotspot with ecologically and commercially important top predator species. The high mobility of these predators and the occurrence of extreme temperature anomalies in areas of high predator density provide an opportunity to examine MHW impacts across multiple trophic levels. The study aims to quantify the impacts of four major North Pacific MHWs (2014, 2015, 2019, 2020) on the spatial distributions of 14 marine top predator species, examining how responses vary across species and MHWs, and analyzing how MHWs shift predators across political boundaries.

Existing literature highlights the widespread ecological and socioeconomic consequences of marine heatwaves. Studies have demonstrated the impacts of long-term climate change on species distributions, often showing poleward shifts in response to warming. However, there's a paucity of research comprehensively analyzing the multi-species responses to multiple, distinct marine heatwave events. Previous investigations have primarily relied on observational data, resulting in fragmented insights limited to specific species or single heatwave events. While some studies have shown range shifts in response to heatwaves, these are often limited by data availability and do not usually provide a comparison across different species and heatwave characteristics. The need for a multi-species, multi-heatwave framework using statistical modeling to address the limitations of observational data and provide a more comprehensive understanding of species responses is evident in the literature.

This study employed boosted regression tree (BRT) models to predict the redistribution of 14 top predator species' preferred habitats during four major North Pacific MHWs (2014, 2015, 2019, 2020). Extensive telemetry datasets spanning 2000-2010, sourced from the Tagging of Pacific Predators Project and private datasets, were used to train the models. The datasets included information for seabirds, mammals, turtles, tunas, and sharks. Species were categorized into three geographical groups (Northern, Coastal, Southern) based on telemetry data distribution. The BRT models incorporated multiple environmental variables, including sea surface temperature (SST), oxygen levels at 200m, mean primary productivity, surface chlorophyll-a, sea level anomaly, eddy kinetic energy, mixed layer depth and day of the year. All environmental variables were resampled to 0.25-degree resolution to match the coarsest resolution dataset. The models were validated using independent datasets from various public and government sources, such as fisheries observer programs, citizen science databases, and tagging data, ensuring robustness and generalizability. Pseudo-absences were generated at a 1:1 ratio with presences. Model performance was evaluated using multiple cross-validation techniques and independent datasets. MHW impacts were quantified using four metrics: displacement distance and direction of core habitat, percent change in range extent (compression or expansion), and percent change in habitat area. The models were predicted over the daily environmental data from 2000-2020 and spatially constrained within the minimum convex hull of the training data. Jurisdictional shifts in predator habitats across national Exclusive Economic Zones (EEZs) were also analyzed using version 11 EEZ boundaries. The study used a 50% prediction quantile as a threshold for defining core habitats, balancing the need for a suitable habitat definition with avoiding the complete loss of habitat during extreme events observed when using more conservative thresholds. Twenty replicate BRTs were generated to assess the impact of data variability on the outcome. To improve comparability, range compression/expansion and habitat area loss/gain were expressed as percent changes from mean conditions. The process of model development, validation, and analysis was carried out using R version 4.0.4.

The study revealed highly variable responses of top predator species to MHWs, both within and among events. Displacement distances, directions, range changes (compression or expansion), and habitat area changes varied greatly among species and across heatwave years. For instance, while some species showed northwestward displacement during the 2014 and 2015 events, others exhibited southeastward movement during the 2019 and 2020 events, likely due to emergent cool water refugia. Some species experienced range compression during certain events and expansion during others, indicating complex species-specific responses. The analysis highlighted significant cross-jurisdictional shifts in predator habitats, with considerable changes in the distribution of habitat across US, Mexican, Canadian EEZs, and the high seas. The US EEZ consistently gained predator habitat during each MHW, while the Mexican EEZ often experienced losses. The magnitudes of these shifts varied across heatwaves, underlining the need for dynamic, adaptable transnational management strategies. Coastal species showed the largest jurisdictional redistributions during the 2014 and 2015 events, while Southern species experienced the most significant shifts during 2019 and Northern species in 2020. The severity of impacts often correlated with regions experiencing the highest temperature anomalies. Temperature-only models showed significantly worse predictive performance compared to multivariate models, indicating the importance of considering multiple environmental factors beyond temperature in understanding species responses. The study also demonstrated the high predictability of species responses despite the high variability, with models performing well in extensive validation across space, time, and on novel data. A real-time dynamic ocean management tool, accessible online, was developed as a proof of concept, to predict predator distributions and responses to extreme conditions.

The findings underscore the need for novel management solutions capable of rapidly responding to MHW-driven species redistribution. The high variability of species responses necessitates a shift away from assuming consistent impacts across events. However, the high predictability of responses, demonstrated by the model's performance, enables real-time predictions of species distributions during future MHWs. This allows for proactive management strategies, guiding observation programs and informing real-time management decisions. The study highlights the importance of a multi-variable approach in understanding species responses to MHWs, demonstrating that factors beyond temperature significantly influence species distribution shifts. This emphasizes the necessity of monitoring not just temperature, but also other key ecological variables such as oxygen levels and primary productivity, to better anticipate and mitigate species impacts. The cross-jurisdictional shifts in species distributions necessitate international collaboration in managing shared resources and mitigating potential conflicts. The results suggest that the US will face particularly complex management challenges as MHWs increase the abundance of commercially valuable species within their EEZ, requiring careful management to ensure stock sustainability and avoid overexploitation. The redistribution of protected species also raises concerns about increased risk of bycatch and other threats.

This study provides a comprehensive assessment of the variable yet predictable impacts of marine heatwaves on top predator distributions in the Northeast Pacific. The findings reveal significant cross-jurisdictional shifts in predator habitats, highlighting the need for dynamic, adaptable transnational management strategies. The development and implementation of a real-time dynamic ocean management tool demonstrate the potential for proactive responses to these extreme climatic events. Future research should focus on improving the predictive capabilities of ecological forecasts, integrating behavioral and physiological data into species distribution models, and enhancing international collaboration in managing shared resources. The integration of real-time predictions with observation programs to improve model accuracy is a crucial next step.

While the study utilized extensive telemetry data and independent validation datasets, the models relied on correlative relationships and didn't explicitly incorporate species traits like physiology, movement syndromes, and life histories, which could affect their responses to MHWs. The models captured species' fundamental niches, but did not fully account for environmental preferences during reproductive behaviors or other ecological processes that influence distribution. The availability of satellite observations for certain variables limited the baseline period for some analyses. The use of a 50% threshold for defining core habitats, while balancing the need to avoid complete habitat loss during extreme events, might be viewed as less conservative than approaches used in some other climate change studies.