The University of St Andrews is currently hosting the 6th International Statistical Ecology Conference and our SMRU Consulting North America colleague and senior statistician Ruth Joy is in town to present.

This week-long conference is bringing together researchers to discuss recent developments in statistics applied to ecology and conservation throughout the globe.  The SMRU Consulting presentation is;

Describing Southern Resident Killer Whale Habitat in the Salish Sea using Multiple Sources of Data 

Ruth Joy, Joe Watson, Marie Auger-Méthé, Sheila Thornton, Dominic Tollit and Jason Wood.

The Southern Resident Killer Whale (SRKW) population is listed as Endangered under the Canadian Species at Risk Act and the American Endangered Species Act because of their small population size, low reproductive rate and the existence of a variety of anthropogenic threats.

The question of how SRKW use designated and newly proposed Critical Habitat in British Columbia is considered the fundamental piece of information that is needed to effectively assess risk from all anthropogenic stressors, not just acoustic ones.

Currently, there are no systematic cross-boundary US-Canadian surveys that could be used to describe the distribution of SRKWs in the Salish Sea. Instead, the available datasets vary in data type and in spatial effort. The core area of summer habitat is described largely by opportunistic sightings, while an area important for returning Chinook salmon (a key prey species), has both boat effort tracks and sightings; additionally, there are data from passive acoustic monitoring and satellite tracking efforts.

Ruth will be describing the methods used to obtain the spatial temporal point estimates of SRKW habitat use, together with the associated uncertainty estimates.  This work helps to better understand the fine-scale spatial and temporal distribution of SRKW during “Chinook season” (May – October).


Deep dive into the stats:

This project aims to synthesise disparate datasets within a spatial-temporal Gaussian random field model using R-INLA, with a hierarchy of dependence structures. Data from the different datasets are combined in a hierarchical structure with individual measurement error models, and transformations of the latent random fields to allow for joint estimation of SRKW habitat use. Connections between the stochastic partial differential equations and Markov random fields are used to obtain sparse matrices for the practical computation of posterior samples.