Evaluation of spatial variation in chronic wasting disease risk with Bayesian Poisson log-Gaussian model

Chronic wasting disease (CWD) is a fatal neurodegenerative disease among cervids that has steadily spread across the United States and Canada. The year-to-year increase in the geographic spread of this disease among white-tailed deer and mule deer have raised concerns about conserving these species and sustainable big-game hunting. Knowledge of the spatial variation in CWD risk in Kansas, a state that attracts big game hunters nationwide is not fully understood. We explored the spatial variation in CWD risk and the potential effects of habitat-level covariates using surveillance data collected in Kansas from 2005 to 2023, with a Poisson log-Gaussian model in a Bayesian framework. Two models were considered; Model 1 included only spatial random effects and Model 2 included spatial random effects plus non-linear effects of habitat-level covariate. Following satisfactory convergence of model parameters, choropleth maps of posterior mean estimates for the risk of CWD presence, and measures of spatial heterogeneities were plotted. The impacts of the habitat-level covariates were deemed important predictors of CWD as Model 2 outperformed Model 1. The risk of CWD in the northwestern and southcentral portions of the state is likely driven by similar underlying spatial processes; however, no global smoothing effect was observed. The northwestern region is at higher risk for CWD presence but a gradual increase in risk toward the south and eastern sides of the state is apparent. We conclude that the data-driven Poisson log-Gaussian model is useful in assessing CWD and potentially other wildlife diseases from surveillance sources, and the different spatial patterns and habitat-level covariate association have relevance to CWD management in Kansas.