Modeling Mammalian Extinction Risk: GIS & Anthropogenic Impacts

Abstract

Global wildlife populations are experiencing continued, and in many cases increasing, threats from various forms of human disturbance. Mammals represent one of the most vulnerable taxa to these disturbance impacts due to their unique spatial, resource, and physiological needs. Among the human-driven drivers of wildlife decline are (1) extensive habitat loss due to population growth, urban expansion, and land conversion for extractive purposes, (2) the presence of additional disturbances associated with agricultural pursuits (pesticides, domestic livestock as disease vectors, etc.), and (3) the introduction of invasive species that harm native ecosystems and the wildlife within these. Understanding how different anthropogenic factors impact mammal decline and extinction risk specifically is important to effectively conserving these species. In this study, I incorporate multiple anthropogenic disturbance variables into a multi-criteria evaluation analysis in a global spatial context to build predictive models for identifying areas of key terrestrial mammalian extinction risk. Using data describing regional human population, extent of agricultural land conversion, and invasive species counts, I created five models of potential extinction risk, calculated the proportion of overlap with a validation dataset of current threatened mammal ranges, and compared these results to identify the best performing model. Based on my analysis, the most accurate model included human population at the highest factor weight, agricultural conversion as the next more important variable, and invasive species as the least impactful component. This model accounted for 22.6% of the highest extinction risk validation data, out of a possible 62% overlap given the datasets’ differing areas. These findings suggest that multi-criteria evaluation as a framework for extinction risk modeling has significant merit, but requires further refinement to more accurately capture current risk and be able to project future risk scenarios.