North Atlantic right whales (Eubalaena glacialis) are critically endangered, 350-400 individuals remain. Collisions with vessels and entanglement in fishing gear are among the primary threats to the continued existence of the species. Knowledge of when and where right whales are likely to be is needed for effective mitigation of human-induced right whale mortality. In the present work we employ a species distribution model (SDM) to estimate areas likely to host right whale aggregations on weekly timescales.

SDMs combine species occurrence records with environmental data to model the environmental niche occupied by the species. By projecting the model onto georeferenced environmental data layers, one can generate habitat suitability maps. In conventional usage, these maps represent the average range of the species over decadal or longer timescales. However, management actions take place on weekly timescales. To generate weekly estimates, we paired right whale occurrence records with near-realtime satellite derived and modeled data layers; sea surface temperature, chlorophyll concentration, modeled Calanus finmarchicus (prey) abundance, modeled salinity and subsurface temperature. Time of year and depth were also included as predictors.
Results/Conclusions

Resultant habitat maps show the movement of right whales from the wintertime Cape Cod Bay critical habitat to the summertime Great South Channel critical habitat. Historic whaling grounds were identified as areas likely to host right whales. Modeled prey abundance was the most important data layer – it dramatically increased the richness of habitat maps and improved the model’s predictive capacity. Model performance was assessed by the area under the receiver operator characteristic curve. Performance varied substantially depending on the out-of-sample year used to test the model. The methods developed in the present work are being used to generate realtime estimates of right whale habitat suitability, which are being distributed to resource managers as they are produced. This work as an important first step toward using species distribution models to locate cryptic, migratory species on an operational timescale.