Forecasting cetacean abundance patterns to enhance management decisions

Speciesenvironment models are increasingly recognized as valuable tools for as- sessing protected species distributions and developing measures to reduce or avoid adverse im- pacts. Cetaceanhabitat models can provide a finer spatial resolution than traditional abundance estimates, but model predictions are generally based on past observations rather than current or projected ocean conditions. We present and evaluate methods for near real-time and forecast mod- els of cetacean distribution based on remotely sensed and modeled oceanographic data. Recent advancements in processing satellite-derived data (e.g. microwave/infrared blended sea surface temperature (SST) products) have virtually eliminated data gaps due to cloud cover, allowing short-term forecasts based on single-day snapshots of oceanic conditions. Ocean circulation models (e.g. the Regional Ocean Modeling System (ROMS)) allow medium-range forecast predic- tions of oceanic variables, including SST, chlorophyll and salinity. We developed habitat models for striped dolphin, fin whale and Dall's porpoise using line-transect data collected from July to No- vember 1991�2005 in the California Current Ecosystem. We incorporated daily blended SST data and monthly ROMS SST forecasts as input variables to predict relative species density in 2008. Forecast ability was assessed by the models' ranked predictions across 8 geographic strata, and by visual inspection of predicted and observed distributions. For all 3 species, there was a significant correlation between model predictions using daily blended SSTs and actual survey observations (p < 0.05). Longer-term (3�4 mo) predictions also showed good concordance with observed sighting locations. Cetaceanhabitat models that allow weekly to monthly forecasting of cetacean abundance can greatly enhance short-term decision-making and advanced mitigation planning.