Combining citizen science and survey data in a log-Gaussian Cox process framework to estimate the monthly space-use of Southern Resident Killer Whales using R-INLA and inlabru

Species distribution models (SDMs) are useful tools to help ecologists quantify species-environment relationships, and they are being increasingly used to help determine the impacts of future climate and habitat changes on species. Estimating SDMs can be tricky from a statistical point of view since the effects of spatial and temporal autocorrelations, land cover and environmental covariates and detectability functions all need to be considered and inherently modeled. Furthemore, such models often assume that data have been collected from well-designed surveys and/or studies. In practice, data are often of the form of presence-only sightings collected from 'citizen scientists' and/or industry, and their 'search effort' can be difficult to quantify. Furthermore, search effort from such sources is often concentrated in areas in which the expected count of the species under study is high, and/or where population density is high. Ignoring the search effort can lead to severely biased estimates of the species distribution. 

We look at data collected on Southern Resident Killer Whales (SRKWs), an ecotype with designated 'species at risk' status found off the coast of Vancouver Island. Data from a variety of 'citizen science' sources and government surveys are considered. We present a method to combine the different data sources and estimate the monthly SRKW space-use in a statistically-rigorous manner using spatio-temporal log-Gaussian Cox processes within the R-INLA and inlabru packages. Improved (effort-corrected) estimates of the SRKW distribution will hopefully help ecologists and policy-makers safeguard the future of the SRKW.