Maxent modeling for predicting the potential distribution of human-elephant conflict risk in Sri Lanka

Abstract

Human-elephant conflict (HEC) is a major problem that causes loss of life to both humans and elephants. While HEC risk models have been developed in past studies, there has not been any HEC risk models developed for the country with the highest annual HEC-related elephant deaths which is Sri Lanka. Thus, this study aims to develop a nationwide model to predict the risk of HEC and identify the most significant predictor variables for HEC in Sri Lanka. HEC risk variables and thirteen predictor variables were prepared using GIS tools. The MaxEnt application was used to input the risk variables (as presence points) and predictor variables (as environmental layers) and model the probability of HEC risk at 500m resolution. The modeling showed that distance to elephant distribution areas was the most important predictor variable for HEC, followed by vegetation area, elevation, rangeland area, population density, and agricultural area. The results are supported by past studies that show the preference of elephants to remain within their usual range, but venturing out for food and water when resources are lacking. This is the first study to develop a nationwide HEC risk map for Sri Lanka using machine learning.