Pinnipeds are commonly monitored at land or ice-based sites using aerial photographic surveys. Although these counts form the basis for monitoring populations over time, they do not provide information regarding where animals occur in the water, which is often of management and conservation interest. The distribution of animals at-sea can be inferred by compiling spatially referenced sightings of animals from vessels. However, maps of occurrence probability created by naively pooling in-water sightings can produce misleading results. We developed a probabilistic model that links the number of pinnipeds counted ashore at terrestrial sties and sighted at-sea to produce estimates of availability (i.e., probability of being found on land), abundance, and spatial distribution of pinnipeds through time. Model performance was examined by estimating the abundance and distribution of harbor seals in Glacier Bay National Park and Preserve from 2006-2016, using a data set consisting of aerial photographic surveys and sightings of pinnipeds at-sea gathered during roving whale surveys.
Results/Conclusions
We succeeding in developing a hierarchical model for predicting pinniped abundance and space use. The model uses replicated counts of animals at terrestrial sites to estimate abundance and the probability that animals are available to be counted. A binomial point process links the predicted abundance of animals at-sea to sightings made by vessels. The structure of the model allows for the inclusion of predictors of abundance and availability, as well as spatially explicit predictors of probability of occurrence in the water.
When applied to harbor seals in Glacier Bay, the model provided reasonable estimates of harbor seal abundance, relative to those reported previously. Modeling also revealed that harbor seal occurrence in the water was positively related to proximity to terrestrial locations where harbor seals are known to aggregate. Predicted abundance and occurrence probabilities will be used to inform a decision model for managing island closures and vessel restrictions in the National Park. We discuss the assumptions, limitations, extensions of our modeling approach and place it in the context of related methods.