Chytridiomycosis is one of the most alarming emerging infectious wildlife diseases to be discovered in the past few decades. This disease is linked to global amphibian population declines, and it has been directly or indirectly implicated in several extinctions. Research on the fungus causing the disease, Batrachochytrium dendrobatidis (Bd), has revealed several variables that predict its distribution, but most of these studies have been on continental or global scales. This study utilized a relatively new method of multivariate analysis, Bayesian Ordination and Regression Analysis (boral), and compared it to modeling techniques that have been previously used to understand Bd’s distribution (Maximum Entropy (MaxEnt), Generalized Linear Models (GLM), and General Additive Models (GAM)). This project focused on creating species distribution models for west-central Colorado, where populations of the state endangered boreal toad (Anaxyrus boreas boreas) have declined due to chytridiomycosis. Utilizing field data from Bd positive and negative amphibian populations, the final model’s results will predict which regions are at high risk of being infected by Bd, and provide insight into the best locations to reintroduce toad populations.
Results/Conclusions
Variable importance was determined for several known (temperature, precipitation, elevation, vegetation) and unknown (slope, aspect, land use) factors that have been associated with Bd in the literature, followed by an analysis of the four modeling techniques. Monthly average temperature, precipitation, and vegetation were calculated on a 15-year timescale, from 1995 to 2015. In addition, a normalized difference vegetation index (NDVI) was calculated for both 1995 and 2015; these variables, plus the difference in vegetation (NDVI diff) between those years, were also included in each model. Variable selection was performed using an AIC, and 11 variables were selected: elevation, July precipitation, NDVI diff, NDVI 1995, NDVI 2015, average temperatures from September, April, February, May, and June, and the total average temperature for the region were. Thus far, MaxEnt explains the most variation within the dataset, and it has the best predictive distribution model for Bd. Further analysis will be used to confirm this result. The results of this research will benefit management efforts aimed at conservation and restoration of amphibians, and may allow greater insight into disease management across a range of taxa and spatial scales.