Spatiotemporal patterns of conflict-related sexual violence in Colombia

Abstract Sexual violence is one of the most severe forms of civilian victimization in the Colombian armed conflict, yet population-scale analysis of its patterns remains limited. Here, using two national victimization databases, we profiled victim demographics and applied log-Gaussian Cox process and Hawkes process modelling to examine geospatial clustering and temporal contagion across the entire conflict (1964–2024) and a recent period (2014–2024). Victims were predominantly female (~90%) and disproportionately from minority ethnic groups. Geospatial analysis indicates that sexual violence may be more widely present than raw event counts suggest and has shifted from concentrated, high-intensity incidents to more dispersed patterns. Unadjusted Hawkes models indicated near-critical temporal dependency in both periods, but Hawkes models adjusted for background conflict intensity suggest that event-to-event contagion was minimal in both periods. Identified areas of historically high prevalence remain priorities for victim support, and the temporal dynamics suggest that prevention efforts should proactively address long-term structural and contextual determinants.