Marine pelagic fish are an important component of marine ecosystems as they serve as a critical forage base for many ecologically/commercially important species and respond rapidly to changing ocean conditions. Despite their importance, such species are often neglected because trawl surveys typically better represent bottom-dwelling fish. The potential of using traditional species distribution models to predict pelagic fish distributions is restricted by the limited availability of long-term monitoring data over broad geographical extents, because of the patchy nature and discrete times of field surveys. Maximum entropy (MaxEnt) models are promising tools as they can be applied to presence-only data (e.g., data collected from fishermen targeting a specific species or citizen-science programs). We used MaxEnt to relate fish occurrence records (Atlantic herring, Atlantic mackerel, and butterfish) from fishery-dependent data to environmental conditions, and compared model predictions with results from fishery-independent data. Environmental variables included sea surface temperature (SST) and chlorophyll-a from satellite remote sensing, bathymetry, and climate indices. Monthly habitat suitability maps for these species in the Northwest Atlantic Shelf area, relative influence of environmental factors on their distributions, and MaxEnt’s ability in predicting fish distributions were assessed between the two data sources for March to May and September to November.
Results/Conclusions
In this study, monthly suitable habitat areas for Atlantic herring, Atlantic mackerel, and butterfish estimated from fishery-dependent data were generally broader than those estimated from fishery-independent data. Both data sources suggested that SST variables had the greatest influence on pelagic fish distributions in spring and fall, and the influence of bathymetry increased during the fall. Using fishery-dependent data showed the advantages of being able to capture the great influence of chlorophyll-a variables on pelagic fish distributions in summer months, whereas trawl-survey data were typically not available for June to August. Fishery-independent observations were used for evaluating the predictive ability of MaxEnt in predicting fish occurrence and abundance from fishery-dependent data. The MaxEnt models based on fishery-dependent data showed good predictive performance in predicting pelagic fish distributions, especially for butterfish, and demonstrated the possibility of projecting pelagic fish distributions at both monthly and seasonal time steps. Our study showed a MaxEnt modeling framework combining high-resolution environmental data from satellite remote sensing and lower-resolution species occurrence data from citizen science or similar programs, which provides an alternative approach for better estimating the habitat and distributional shifts of marine pelagic fish species when field data are impractical across broad spatial or temporal extents.