Quantifying the risk of pastoral conflict in 4 central African countries

Climate change has led to increased conflicts between farmers and herders in Africa. We use statistical and machine learning methods to study pastoral conflict in 4 African nations (Central African Republic, Cameroon, Chad, Democratic Republic of Congo) using a novel eight-year dataset (Jan 2015 to Sep 2022) that has fine-grained weather and terrain data about four African nations. Our principal goal in this paper is to investigate hypotheses linking these variables with pastoral conflict in geospatial regions. We generated risk maps that can be automatically updated for use by end-users such as potential decision-makers and policy analysts alike. These risk maps highlight high-risk areas including ones not covered in prior work (e.g. work reported in ACLED, FAO, and UN reports). Our models assess the risk that a given cell will experience pastoral conflict. We also study the variation of this predictive accuracy with the granularity of the cells.