Friday, August 6, 2010

PS 104-120: Operational ecosystem forecasting of migratory marine mammal habitat: Proof of concept prototype system for endangered North Atlantic right whales

Dan E. Pendleton1, Andrew Pershing2, Patrick Sullivan1, Moira Brown3, Changsheng Chen4, Christopher Clark5, Caroline P. Good6, Ruth Leeney7, Charles A. Mayo7, Bruce Monger1, Steven Phillips8, and Nicholas R. Record2. (1) Cornell University, (2) University of Maine & Gulf of Maine Research Institute, (3) New England Aquarium, (4) School of Marine Science and Technology, (5) Cornell Laboratory of Ornithology, (6) Duke University, (7) Provincetown Center for Coastal Studies, (8) AT&T Labs, Inc.

Background/Question/Methods

The ability to accurately predict the presence of highly migratory marine animals would help to identify unknown habitat, guide survey effort, and assess the impact of proposed conservation actions.  North Atlantic right whales (Eubalaena glacialis) migrate 2000+ km, from their calving grounds along the southeast coast of the USA to the Gulf of Maine, where they feed, socialize and mate.  Inter-annal variability in the distribution of right whales has made managing this species particularly difficult.  We built a system to forecast suitability of right whale habitat in Cape Cod Bay, an important nursing and feeding ground.

A high-resolution physical model of ocean circulation was coupled to a satellite-driven life-history model of two primary prey species of right whales. Modeled prey concentration, satellite derived temperature and chlorophyll, and depth, taken at right whale occurrences between 2003 and 2006, provided input to a maximum entropy density estimation model. Predictions of habitat suitability for 2009 in Cape Cod Bay were made by projecting the model onto environmental conditions for that year. To test the validity of the Cape Cod Bay-trained model in another region, the model was projected into neighboring Massachusetts Bay. These predictions were validated with acoustic detections of right whales.

Results/Conclusions

Predictive capacity of 2009 Cape Cod Bay forecasts was reasonable (AUC=0.726). Mean AUC for model training years, 2003 to 2006 was slightly better (AUC=0.80). Mean habitat suitability within three shipping lane choices in Cape Cod Bay were remarkably similar to one another, and all showed wide temporal variation. This suggests that, for the purpose of reducing a vessel's travel through high quality right whale habitat, when a vessel transits is more important than the route of vessel transit. A significant relationship was found between right whale habitat suitability and acoustic detections of right whales in the Boston shipping lane (r2=0.51, p<0.05), indicating that the model provided good information on the level of right whale activity outside of the model-training region. Our system is an important step toward managing endangered species in a changing world.