Leveraging Water Data in a Machine Learning–Based Model for Forecasting Violent Conflict

The World Resources Institute presents a random forest model that captures 86% of future conflicts (defined as 10+ fatalities per year), though with a 50% false positive rate. While water-related variables correlate with conflict, they are not empirically significant for model decisions unless the conflict definition is modified. A web-based tool is provided to visualize these forecasts and vulnerabilities.

• The researchers developed a random forest model to forecast conflict, which is defined as organized violence causing at least 10 fatalities within a 12-month period. When tested, the model successfully captured 86 percent of future conflicts, although it produced a noisy signal where 50 percent of predictions were false positives.
• Water-related indicators were found to correlate with conflict outcomes but were not empirically significant for the model's decision-making process. However, the significance of these water variables increases if the definition of conflict is adjusted, such as by examining only emerging conflict or lowering the fatality threshold.
• A web-based tool has been created to house the model, enabling users to analyze indicators and forecasts spatially and temporally to identify underlying vulnerabilities and support water-related interventions for peace-building or conflict mitigation.

APA

World Resources Institute (n.d.). Leveraging Water Data in a Machine Learning–Based Model for Forecasting Violent Conflict. https://www.wri.org/research/leveraging-water-data-machine-learning-based-model-forecasting-violent-conflict

Chicago

World Resources Institute. Leveraging Water Data in a Machine Learning–Based Model for Forecasting Violent Conflict. n.d. https://www.wri.org/research/leveraging-water-data-machine-learning-based-model-forecasting-violent-conflict.

Wikipedia

{{cite report |author=World Resources Institute |title=Leveraging Water Data in a Machine Learning–Based Model for Forecasting Violent Conflict |url=https://www.wri.org/research/leveraging-water-data-machine-learning-based-model-forecasting-violent-conflict |access-date=17 August 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{worldresourcesinstitutendleveraging, author = {{World Resources Institute}}, title = {{Leveraging Water Data in a Machine Learning–Based Model for Forecasting Violent Conflict}}, institution = {World Resources Institute}, url = {https://www.wri.org/research/leveraging-water-data-machine-learning-based-model-forecasting-violent-conflict}, urldate = {2026-08-17}, note = {Indexed by Climate Insights Directory} }