Repository logo
 

Statistical models for animal telemetry data with applications to harbor seals in the Gulf of Alaska

Date

2017

Authors

Brost, Brian M., author
Hooten, Mevin B., advisor
Small, Robert J., committee member
Wittemyer, George, committee member
Boone, Randall B., committee member

Journal Title

Journal ISSN

Volume Title

Abstract

Much is known about the general biology and natural history of harbor seals (Phoca vitulina), but questions remain about the aquatic and terrestrial space use of these marine mammals. This is in large part because methods for examining the spatial ecology of harbor seals are poorly developed. The objective of this dissertation is to pair existing telemetry data with contemporary spatio-temporal modeling to quantify the space use and resource selection of harbor seals in the coastal waters of southern Alaska. Recent extensions to models for analyzing animal telemetry data address complications such as autocorrelation and telemetry measurement error; however, additional challenges remain, especially in the context of analyzing Argos satellite telemetry data collected on marine mammals like harbor seals. For example, existing methods assume elliptical (or circular) patterns of measurement error, even though Argos satellite telemetry devices impose more complicated error structures on the data. Constraints, or barriers, to animal movement present another complication. Harbor seals and other marine mammals are constrained to move within the marine environment, and mechanistic models that do not adhere to movement barriers yield unreliable inference. Therefore, a primary goal of this research is to develop statistical tools that account for these nuances and provide rigorous, ecologically relevant inference. Even though the models presented in this dissertation were specifically developed with Argos satellite telemetry data and harbor seals in mind, the methods are general and can be applied to other species and types of telemetry data. This dissertation consists of five chapters. In Chapter 1, I briefly discuss the general biology of harbor seals, focusing on what is known about their spatial habits in Alaska. I then summarize trends in Alaskan harbor seal abundance, a topic that motivated my research as well as the work of many others. I describe the existing Alaska Department of Fish and Game telemetry data sets that are available for examining harbor seal spatial ecology, commonly-used statistical methods for analyzing animal telemetry data, and conclude with the objectives of my research and an outline for the remainder of the dissertation. In Chapter 2, I propose an approach for obtaining resource selection inference from animal location data that accounts for complicated error structures, movement constraints, and temporally autocorrelated observations. The model consists of two general components: a model for the true, but unobserved, animal locations that reflects prior knowledge about constraints to animal movement, and a model for the observed telemetry locations that is conditional on the true locations. I apply the model to simulated data, showing that it outperforms common ad hoc approaches used when confronted with telemetry measurement error and movement constraints. I then apply the framework to obtain inference concerning aquatic resource selection and space use for harbor seals near Kodiak Island, Alaska. Chapters 3 and 4 shift the focus from inference concerning aquatic space use and resource selection, to inference concerning the use of coastal resources (i.e., haul-out sites) by harbor seals. In Chapter 3, I present a fully model-based approach for estimating the location of central places (e.g., haul-out sites, dens, nests, etc.) from telemetry data that accounts for multiple sources of uncertainty and uses all of the available locational data. The model consists of an observation model to account for large telemetry measurement error and animal movement, and a highly flexible mixture model (a Dirichlet process) to identify the location of central places. Ancillary behavioral data (e.g., harbor seal dive data obtained from the satellite-linked depth recorders) are also incorporated into the modeling framework to obtain inference concerning temporal patterns in central place use. Based on the methods developed in Chapter 3, I present a comprehensive analysis of the spatio-temporal patterns of haul-out use for harbor seals near Kodiak Island in Chapter 4. Chapter 4 also extends previously developed methods to examine the affect of covariates on haul-out site selection and to obtain population-level inference concerning haul-out use. I conclude, in Chapter 5, with some general thoughts about analyzing animal telemetry data, as well as potential future research directions.

Description

Rights Access

Subject

Citation

Associated Publications