Skilful decadal-scale prediction of fish habitat and distribution shifts

Our understanding of climate change impacts typically focuses on long timescales (50-100 years), which are less useful for immediate decision-making by regional bodies, local governments, businesses, and individuals. Near-term climate predictions offer a potential solution, but predictive skill on annual-to-decadal timescales has been limited, particularly for the ocean. However, decadal ocean predictions could be invaluable for climate adaptation and sustainable development, especially in ocean-dependent nations and communities of the Global South. Climate-driven species redistribution is a major challenge, with shifts in species distribution occurring faster in the ocean than on land due to higher marine species vulnerability to warming. This leads to changes in fishing opportunities, requiring adaptation by local communities and fishers. It also creates potential for international conflicts over fishing rights as stocks move across international jurisdictions. Foreseeing these shifts is crucial for conflict avoidance and adapting marine fisheries to climate change. This study directly applies decadal climate predictions to forecast shifts in the habitat and distribution of marine species in the North Atlantic, focusing on three exemplar fish species: Atlantic mackerel, Atlantic bluefin tuna, and blue whiting. The study distinguishes between species habitat (where conditions are suitable) and distribution (where the species are actually found).

The paper references existing literature on climate change impacts on marine ecosystems, the development of near-term climate prediction capabilities, the challenges of decadal-scale predictability in the ocean, the impacts of species redistribution on human communities and international relations, and previous studies on the specific fish species considered (mackerel, bluefin tuna, and blue whiting). It cites studies showing the northward and westward expansion of mackerel, the appearance of bluefin tuna in previously unobserved areas, and the variability in the spawning distribution of blue whiting, linking these shifts to oceanographic changes. The authors reference studies on the existing habitat models for these fish species, linking them to specific environmental parameters such as sea surface temperature and salinity. The literature review also covers general methods for forecast verification and the use of multi-model ensembles to improve prediction skill. The importance of initialization in climate prediction models and its impact on forecast skill, particularly for decadal predictions are highlighted.

The study combined biological habitat models characterizing the environmental preferences of three North Atlantic fish species (Atlantic mackerel, Atlantic bluefin tuna, and blue whiting) with predictions from five decadal climate prediction systems. For each species, a habitat model was used which relates the species' presence or probability of occurrence to environmental variables. These models were based on existing literature and data, with sea surface temperature (SST) identified as a key driver for mackerel and bluefin tuna, and subsurface salinity for blue whiting. The five climate prediction systems used (CESM DPLE, EACEarth3, HadGEM3, MPI-ESM1.2-HR, and NorCPM) all followed the CMIP6 Decadal Climate Prediction Project (DCPP) protocol, providing a large ensemble of 85 retrospective forecasts. Model outputs (SST and salinity) were processed, regridded, and bias-corrected using observational data (HadISST v1.1 and EN4). Habitat models were then applied to both the bias-corrected forecasts and to observational data, generating estimates of suitable habitat area. This enabled comparison of forecast skill using retrospective forecasting. Forecast skill was assessed using the Pearson correlation coefficient, the Mean Squared Error skill score (MSESS), and the Continuous Ranked Probability skill score (CRPSS). A persistence forecast (tomorrow equals today) served as a baseline for comparison. The authors also compared the performance of initialized forecasts with uninitialized climate projections. For blue whiting, scientific monitoring survey data were used to assess the correspondence between habitat forecasts and observed distribution.

The study found statistically significant predictive skill in forecasting the physical drivers (SST and salinity) of fish habitat, with five-year forecast skill generally high. The predictive skill of these physical variables translated into significant skill in forecasting habitat area on multi-annual and decadal time scales. The habitat forecasts, using the ensemble of climate prediction systems, generally outperformed persistence forecasts, especially for lead times of three years or more. Multi-year averaging significantly improved forecast skill, with correlation coefficients reaching 0.95, 0.94, and 0.74 for decadal averages of mackerel, bluefin tuna, and blue whiting habitat, respectively. The improved skill with multi-year averaging was attributed to the ability of initialized climate models to capture low-frequency variability in the ocean system, effectively filtering out high-frequency noise. The study also showed that initialized climate predictions outperformed uninitialized climate projections, especially regarding the probabilistic representation of habitat. Although the study focused on habitat suitability, the results also showed some correspondence with observed distribution shifts. For blue whiting, the model accurately predicted the decline in spawning habitat, but observed distribution changes were more rapid due to fishing pressure.

The findings address the research question by demonstrating the feasibility and skill of using decadal climate predictions to forecast fish habitat and distribution shifts. The results highlight the potential of these forecasts as valuable tools for stakeholders in managing marine resources and mitigating conflicts arising from changing fish stocks. The superior performance of initialized decadal climate predictions over uninitialized projections underscores the importance of using models that incorporate current oceanographic conditions. While the focus was on habitat suitability, the authors acknowledge the limitations of habitat models in perfectly predicting actual species distribution, as other factors such as fishing pressure, predation, and biological interactions also play a role. This study provides evidence supporting marine ecological forecasting as a climate change adaptation tool, particularly relevant to ocean-dependent nations and communities in the Global South.

This study demonstrates the significant skill of decadal-scale climate predictions in forecasting changes in fish habitat and distribution in the North Atlantic. The findings show that these forecasts can outperform simpler baseline predictions and provide valuable information for stakeholders. Multi-year averaging enhances the skill of the forecasts, highlighting the importance of considering low-frequency variability in the ocean. The study underscores the potential of marine ecological forecasting as a climate adaptation tool, particularly beneficial for vulnerable communities and nations highly dependent on marine resources. Future research could focus on incorporating additional factors into habitat models, improving the representation of biological interactions, and expanding the geographic scope of the forecasts to other marine ecosystems.

The study's main limitation is the focus on habitat suitability rather than directly predicting species distribution. Other factors beyond environmental conditions (e.g., fishing pressure, predation, biological interactions) can influence species distribution. The habitat models used were based on existing literature and may not fully capture all the complexities of the species' ecological niches. The study also focused on a specific region (North Atlantic) and particular species, limiting the generalizability of the findings to other regions and species. The availability of suitable observational data for verifying forecasts varied among species.