Date of Award

Winter 2006

Project Type


Program or Major

Natural Resources: Environmental Conservation

Degree Name

Master of Science

First Advisor

Andrew A Rosenberg


As a benthic foraging marine mammal, sea otters ( Enhydra lutris) present a unique opportunity for conducting a quantitative assessment of behavior based on habitat use as well as developing a habitat based density model using GIS because of the sea otter's well defined habitat requirements. Several studies have documented sea otter behavior but none have calculated the probability of occurrence of a particular behavior based on habitat attributes. Previous predictive models of sea otter density have been constructed, however these models have excluded offshore habitat. Seven aerial surveys, that included offshore habitats, were conducted between 1995 and 2005 in western Prince William Sound to estimate distribution and abundance of sea otters (Enhydra lutris).

The location and densities of sea otters that resulted from these surveys were used to explore relationships between sea otters and habitat attributes, both nearshore and offshore. These relationships described in western Prince William Sound were then used to construct habitat based models to predict sea otter carrying capacity and total abundance at different spatial scales. The data from the aerial surveys were also used to quantify the relationship between a sea otter's behavior and the habitat attributes associated with the location of the animals when the behavior occurred.

Stepwise logistic regression was used to describe relationships between behavior, diving or not diving (assumed resting) and habitat attributes. Three subsets of the data were examined; all animals, all single animals without pups and all single animals with pups. Bathymetry was consistently significant (alpha = 0.05) in determining the probability of a behavior being diving or not diving, regardless of size of group or reproductive status. Group size was the first variable to enter the stepwise regression analysis of all available sightings, regardless of reproductive status, with bathymetry as the second and final variable. Among single animals with pups bathymetry was the first variable and distance to shore was the second and final variable to enter the model. Bathymetry was the only significant variable in the analysis of single animals without pups.

The aerial survey data from western Prince William Sound, AK, was used to create a predictive density model based on five habitat attributes; bathymetry, distance to the closest shoreline, distance to the closest protected shoreline, distance to the closest tidewater glacier and distance to the closest anadromous stream. The mean predictive density estimate was 2.0316/km2 with a total corrected population estimate within the survey boundaries of 16,441, with a range of 14,468 to 18,803 (alpha = 0.05).

Special attention was given to northern Knight Island, an area heavily impacted by the Exxon Valdez oil spill in 1989. Predicted densities within that area were 1.5792/km2 with an estimated abundance of 384. The actual mean abundance estimate at northern Knight Island between 1995 and 2005 was 68 with a range of 34 to 102 (alpha = 0.05), illustrating a discrepancy between predicted estimates and of actual survey abundance estimates.

The analysis and results presented in this work give insight into the density and distribution variation of sea otters in Prince William Sound as well as contribute to the understanding of the sea otter's use of its nearshore habitat.