Gilbertson, Kendra E., authorWebb, Colleen, advisorKosoy, Michael, committee memberSofaer, Helen, committee memberKading, Rebekah, committee member2025-06-022026-05-282025https://hdl.handle.net/10217/241048Most emerging infectious diseases that affect humans originate in animal species. In order to accurately predict disease risk, we must understand the spatial distribution of the various environmental, ecological, and host factors that allow pathogens to spillover into human populations. The most foundational of these requirements is the presence of one or more host species that allow the pathogen to persist in a given location. We accomplished this by using a Bayesian Additive Regression Trees (BART) model to predict the distributions of 15 medically-relevant small mammals species across the South Caucasus while also exploring different methodological strategies. Our flexible machine learning approach allowed us to create predictions that had high ecological accuracy and strong discrimination ability crucial for discerning the likelihood of species presence. We also found that models trained on random background points were well suited for our goal of maximizing discrimination ability without sacrificing biological realism. We recommend the consideration of this approach to those using species distributions to predict disease risk. Next we incorporated these predictions as predictive layers in a risk prediction model of tularemia in order to understand the spatial risk and ecological drivers of this complex pathogen. We found that tularemia was primarily a rural disease in humans, partially a result of the increased contact between humans and animals in agricultural settings. While we observed different distributions between clinical manifestations of tularemia, the causes of these diverging patterns warrant further inquiry. We extended our examination of zoonotic risk to include six additional pathogens. To paint a picture of overall disease risk in the South Caucasus, we combined the predicted distributions of all diseases to create a disease richness map. This is of public health import as it identified areas of elevated transmission risk across pathogens. We also found that zoonoses in the South Caucasus typically occur in rural areas, and those with occupational exposures were at the most risk. These areas are more likely to support higher diversity of host and vector species, allowing for circulation of pathogens in the environment and greater contact rates between humans and disease-carrying animals. Our research adds to a broader understanding of zoonotic risk and its ecological factors in the South Caucasus and we provided methodological recommendations relevant to the field of disease ecology.born digitaldoctoral dissertationsengCopyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright.spatial epidemiologytularemiadisease ecologyzoonosesspecies distribution modelingTextEmbargo expires: 05/28/2026.