Hosts, pathogens, and species occurrence: Using occupancy models to make inference in disease systems

Background/Question/Methods

Boreal toads (Anaxyrus boreas boreas) are native to high-elevation areas in the Intermountain West of North America.  Boreal toad populations in the Southern Rocky Mountains have experienced drastic declines over the past 30 years. The suspected culprit is the fungal pathogen Batrachochytrium dendrobatidis (Bd), the causative agent of the disease chytridiomycosis. Our initial objective was to evaluate the potential influence of Bd on the occupancy, colonization, and extinction probabilities of boreal toads in Colorado using a historic dataset. We began to investigate this question using a dynamic single-species occupancy model, however the Bd detection information that was used as an explanatory variable in our modeling was obtained from swabbed boreal toads. Therefore, Bd detection information was available only at sites occupied by boreal toads in a particular season, inducing bias in parameter estimates.  In light of this bias, a dynamic two-species modelling approach was considered.  

Results/Conclusions

By modeling the dynamics of both boreal toads and Bd simultaneously, previous bias was eliminated (though some parameter redundancy was imposed).  We found that boreal toad extinction probability was higher when Bd was also present at a site, resulting in lower probabilities of boreal toad occupancy at these sites.  Our preliminary results indicate that the probability of Bd local extinction without simultaneous local extinction of boreal toads is rare, but possible.  Finally, we identified some parameters that were not estimable due to the unobservable state induced by our sampling design.  Our findings underscore the need to pursue alternative sampling methods for Bd in order to learn more about the dynamics of this pathogen, which may be especially important when amphibian reintroduction efforts are being considered. Our results allow us to obtain better estimates of occupancy and breeding probabilities in Colorado boreal toads and will be applicable to other host-pathogen systems and systems with unobservable states as well.