Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 131 - Methods for Spatial, Temporal, and Spatio-Temporal Data
Type: Contributed
Date/Time: Monday, August 9, 2021 : 1:30 PM to 3:20 PM
Sponsor: Section on Statistics and the Environment
Abstract #318161
Title: Continuous-Time Discrete-Space Movement Models Over Two- and Three-Dimensional Space
Author(s): Joshua Hewitt* and Robert S Schick and Alan E Gelfand
Companies: Duke University and Duke University and Duke University
Keywords: animal movement; bridged random walk; computational statistics; imputation

Continuous-time discrete-space (CTDS) models for animal movement model trajectories across discretely observed spatial domains, which arise via gridded remote sensing products in 2D, or via underwater sound propagation models in 3D. Covariate effects are approximately estimated from finite observations of animal location because exact likelihoods have O(N^3) computational complexity, where N is the size of the spatial domain. Typical approximations average estimates from multiple imputations of the complete, unobserved trajectory. However, imputations usually discretize output from continuous-space surrogate models which do not account for complex boundaries like coastlines and bathymetry. As a result, covariate estimates may be biased by imputations that unrealistically move along or cross physical barriers. We remedy such issues by using bridged, discrete-space random walks to sample the complete trajectory during estimation. We also extend CTDS models to spatial domains with irregular cell sizes, which is essential for 3D aquatic movement. We demonstrate the method in 2D and 3D via simulation and application to marine mammal data.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2021 program