Climate warming, a key aspect of anthropogenic global change, is significantly impacting marine ecosystems. The North Atlantic and Arctic Oceans are experiencing community changes, with warmer-water species moving northward and colder-water species retreating. The extent and consequences of these shifts remain unclear, particularly concerning the implications for fish biomass. Existing species distribution models (SDMs) often lack species abundance or biomass data, model species independently, or focus on limited, commercially important species. This research addresses these limitations by employing a J-SDM, specifically the Hierarchical Modelling of Species Communities (HMSC) framework, to analyze 107 Northeast Atlantic marine fish species, including biomass data for 61 species, across the continental shelf from the North Sea to the Barents Sea. The HMSC framework allows for a joint analysis of species responses to environmental factors and co-occurrence patterns, improving the estimation of model parameters, especially for rare species. This study aims to project richness, relative dominance, and species distributions and biomass under different future climate scenarios (SSP1-2.6, SSP2-4.5, SSP5-8.5) to 2050 and 2100. Understanding these future changes is crucial for ecosystem management, fisheries, and conservation efforts, especially in the rapidly warming Arctic, where species' capacity to shift northward is limited by the lack of a contiguous continental shelf.

Numerous studies document species range shifts driven by climate warming, with poleward expansions and equatorward contractions. These shifts are influenced by local climate velocities, which may not directly reflect global temperature patterns. The Arctic, warming at a rate nearly four times the global average, is experiencing increased species richness due to the influx of warmer-water species. These distributional shifts have significant implications for ecosystems and human activities, impacting the fishing industry and potentially causing transboundary conflicts. Climate-induced population displacement from marine protected areas and changes in species interactions due to range shifts also pose conservation challenges. While previous SDMs have provided insights, they often lack the comprehensiveness of a joint model accounting for species interactions and biomass.

This study utilized bottom trawling data from the FishGlob database, encompassing the North Sea, Norwegian Sea, and Barents Sea (2004-2022). The dataset was filtered to include data from the continental shelf (<500m depth), species sampled in at least 100 hauls and 10 years, and species with available environmental data. The final dataset included 16,345 hauls and 107 fish species. Environmental variables included bottom water temperature, sea ice concentration (where applicable), current components, bottom dissolved oxygen, phytoplankton concentration, and depth. Data were obtained from Copernicus Marine Environment Monitoring Service and BioOracle. Co-linearity was addressed using Variance Inflation Factors (VIF), resulting in the selection of seven variables. A two-part hurdle model within the HMSC framework was employed, using a binomial model for species occurrence (probit distribution) and a Gaussian model for log-transformed catch per unit effort (CPUE) to represent biomass. Spatial autocorrelation was accounted for using Gaussian Predictive Processes (GPP). A phylogenetic tree was constructed using NCBI Common Tree software to incorporate phylogenetic relationships in the model. Future environmental layers were derived from the IPSL-CMIP6 climate model under three Shared Socioeconomic Pathways (SSPs): SSP1-2.6 (+1.6°C by 2100), SSP2-4.5 (+2.6°C by 2100), and SSP5-8.5 (+4.5°C by 2100). Model convergence was assessed using Gelman-Rubin diagnostics, and model performance was evaluated using AUC and R² values, and five-fold cross-validation. Species traits (maximum length, age at maturity, fecundity, habitat, trophic level, preferred temperature, maximum depth, and zoogeography) were obtained from FishBase and used in regression analyses to explain species responses to climate change. Geographic range metrics (range, biomass, core range, core biomass), fragmentation indicators (number of polygons, mean area, mean distance), species richness, and relative dominance were calculated and analyzed to assess distributional changes under future scenarios. Biomass analysis was restricted to 61 species with mean R² >0.05 in cross-validation.

Depth and bottom temperature were the most significant environmental predictors in both occurrence and biomass models. Strong phylogenetic niche conservatism was observed, indicating that species' niches are highly influenced by phylogenetically structured traits. The model projections indicate overall increases in species richness and decreases in relative dominance across all scenarios. In the northern Barents Sea, species richness is projected to double around Svalbard and the north coast of Norway, while the deepest parts show slight decreases. Relative dominance decreased in the northern and eastern Barents Sea. Polar cod (Boreogadus saida), a dominant species in the present-day Barents Sea, is projected to practically disappear under the high-emission scenario (SSP5-8.5) by 2100. Norway pout, blue whiting, and whiting are projected to increase their relative dominance significantly. Species generally increased their distribution range and biomass under high-emission scenarios, particularly within their core ranges. However, Arctic and Arctic-boreal species strongly declined across all scenarios, while boreal and warmer-water species expanded. The decline in currently abundant species (polar cod and capelin) wasn't compensated by expanding boreal species, resulting in an overall decline in fish biomass with climate warming. Species zoogeography was the best predictor of range and biomass changes. Arctic and boreal-arctic species were projected to decline, while boreal, temperate, subtropical, and deep-water species increased. Community-wide range shifts were projected northwards and eastwards, increasing in magnitude with the severity of the climate change scenario. While the number of geographic range units increased and the mean unit area decreased with increasing emissions, the interpretation of geographic range fragmentation is complex, driven by contrasting trends in Arctic (declining) and warmer-water (expanding) species. The model projections show a decrease in the number of dominant species, indicating a homogenization of the community by the end of the century.

The study's projections align with the observed global redistribution of marine species, showing increasing richness at higher latitudes. The projected biomass increases in several boreal and warmer-water species are not sufficient to offset the decline of Arctic and boreal-arctic species, leading to an overall decline in biomass. The shifts in dominant species indicate a substantial community change. While richness is projected to slightly decline at lower latitudes, caution is required in interpreting these projections due to the potential influx of species from outside the study area. The increased richness and abundance under future climate warming scenarios improve the predictability of community properties. However, the model's predictive capability is lower for Arctic species, highlighting a need for better data collection to improve the accuracy of future projections. The projected increase in warmer-water species may offer new fishing opportunities, but also the potential for increased conflicts due to stock shifts across economic zones. The study also emphasizes the importance of including fishing impacts in future models. The high sensitivity of Arctic and boreal-arctic fish species contrasts with the relative robustness of Arctic benthic taxa, suggesting potentially novel species interactions. Further research is needed to understand the consequences of keystone species loss (e.g., polar cod) on higher trophic levels.

This study provides comprehensive projections of future marine fish distributions and biomasses in the Northeast Atlantic under different climate change scenarios. The results highlight the significant impacts of warming on Arctic and boreal-arctic species, with potential for local and global extinctions. While increases in species richness are projected, there is an overall decline in biomass, reflecting a dramatic restructuring of the community. These findings underscore the urgent need for improved data collection, particularly for Arctic species, and the importance of incorporating climate change projections into fisheries management and conservation strategies. Future work should focus on refining projections for poorly sampled regions and incorporating the effects of fishing.

The model's predictive power is lower for Arctic species due to data limitations, which could affect the accuracy of projections for these species. The projections exclude the effects of fishing, which may interact synergistically with climate change. The use of a single global earth system model for future projections introduces some uncertainty. The interpretation of richness and dominance changes in southern regions requires caution due to the potential for species expanding from outside the study area. The study focuses on demersal fish and may not fully represent the dynamics of pelagic species.