MEPS 330:247-256 (2007)  -  doi:10.3354/meps330247

ABSTRACT: Marine ecosystems have fluid habitat features that can be modeled for application in management decisions. With less than 360 beluga whales Delphinapterus leucas remaining in Alaska’s Cook Inlet population and increasing development in the area, it is important to identify habitat requirements of this depleted population. We used a classification and regression tree (CART) model and a resource selection function (RSF) model to determine the importance of selected environmental variables in structuring the habitat use of beluga whales in Cook Inlet. Bathymetry, proximity to mudflats, and distance from rivers classified by flow accumulation values were evaluated with respect to the presence or absence of belugas. Although bathymetry, as applied to the model, was not a significant variable, mudflats were a significant predictor of beluga distribution during early summer months. The importance of flow accumulation varied slightly between the 2 models but, in general, belugas preferred higher flow accumulation areas. The CART model correctly classified 88% of the sightings, with the majority of beluga sightings found to be within 2.7 km of mudflats and 11.5 km of medium flow accumulation inlets. Using a 0.09 probability threshold for habitat, the RSF model correctly discriminated between beluga sightings and non-sightings 92% of the time. Similar regions of habitat were predicted by CART (habitat area, A = 2985 km²) and RSF (A = 2373 km²) models. The habitat models developed in this study will help managers meet current conservation goals and make future decisions to promote the recovery of Cook Inlet beluga whales.

KEY WORDS: Resource selection function · RSF · Classification and regression tree · CART · Geographic Information Systems · GIS · Receiver operating characteristic curve · ROC