Animal movement and prediction: modelling animal behaviour in a changing climate

For highly mobile marine predators, movement ecology fundamentally underpins the spatial distribution of populations and their response to change. Many researchers are interested in using individual-based electronic tracking data to explore and project species’ movement under changing environments. The currently popular techniques (e.g. habitat selectivity, resource utilisation) correlate animals’ use of space with environmental parameters but are generally blind to the processes that underlie animal movement patterns and interactions with their habitats. The open discussion at the 3rd Climate Impacts on Top Predators (CLIOTOP) Symposium highlighted a notable lack of mechanistic or process-based modelling approaches. While there is value in characterising species’ current habitat preferences, approaches that explicitly model movement dynamics, associated behavioural processes, and their ties to environmental features should be better able to provide robust projections of species’ future distributions in a changing environment. The CLIOTOP Task Team 2016-06 ‘Modelling Animal Behaviour in a Changing Climate’ is convened to address this key gap - and opportunity - for integrating a mechanistic understanding of how animals actually use marine areas (e.g. for feeding, migrating, resting) into spatial models of species habitat utilisation and distribution. Early work of the group has delivered a fast method (using R interfaced to C++) to model movements of colony-based predators (Michelot et al. accepted, Ecology. DOI: 10.1002/ecy.1880) using observational animal tracking datasets for estimation and simulation procedures. This framework incorporates external covariates and further work will focus on explicitly modelling environmental relationships. Concurrent efforts are exploring continuous-space parameterizations (using Template Model Builder), specifically random and correlated random walk models with an advection (drift) term for implementing an environmental bias, or a time-varying autocorrelation term to approximate search behaviour (where correlation is low). This presentation will overview task team work and plans for improving our current capabilities to project animal movements under a changing climate.