Authors: John Maingi*, Miami University, Joseph M. Mukeka, Kenya Wildlife Service
Topics: Animal Geographies, Biogeography, Africa
Keywords: Maxent, Species distribution modeling, wildlife
Session Type: Paper
Start / End Time: 4:00 PM / 5:40 PM
Room: Grand Ballroom D, Astor, 2nd Floor
Presentation File: No File Uploaded
We modeled elephant habitat with maximum entropy (MaxEnt) and elephant presence data collected over five different times, between 2005 and 2017. Among the biophysical and anthropogenic predictor variables considered for analysis were distance to roads, distance to settlements, distance to rivers, distance to watering holes, 19 bioclimatic variables including 11 temperature and eight precipitation metrics, elevation, land use land cover (LULC) map, and vegetation phenology. We performed descriptive statistical analysis of predictor variables to determine their spatial variation and removed highly correlated variables in order to avoid spatial autocorrelation. Model performance was evaluated through receiver operating characteristic curve (ROC) plots of randomly selected training and testing data for elephant prediction. The area under the ROC curve (AUC) for both training and test data was over 0.90 indicating a stronger relationship between presence locations and predictor variables. We also described changes in distribution of elephants within the TCA.