The Geography of Grid Crime: Mapping and Predicting Electricity Infrastructure Vandalism Hotspots in Uganda

Electricity infrastructure vandalism is an increasingly serious threat to power-system reliability, utility financial sustainability, electricity-access expansion, and infrastructure resilience in Sub-Saharan Africa. In Uganda, theft and vandalism of distribution lines, transmission towers, transformer components, transformer oil, and underground cables continue to cause service interruptions, replacement costs, safety risks, and operational inefficiencies. Yet infrastructure protection remains largely reactive, with limited use of predictive tools capable of identifying vulnerable locations before incidents occur. This study developed a national spatial intelligence framework for predicting electricity infrastructure vandalism risk in Uganda by integrating Geographic Information Systems (GIS), Kernel Density Estimation, infrastructure exposure analysis, and ensemble machine learning.

The analysis used 1,536 georeferenced vandalism incidents recorded between 2011 and 2026 and modelled risk across a national 5 km × 5 km analytical grid. Predictor variables included historical vandalism density, proximity to previous incidents, road accessibility, settlement access, market access, police/security proximity, substation access, elevation, slope, and electricity infrastructure exposure. Three machine-learning algorithms (Random Forest, Gradient Boosting Machine, and XGBoost) were trained using a presence-background modelling framework and combined into an ensemble risk prediction system.

Results show that vandalism is highly clustered rather than randomly distributed. Distribution lines accounted for 50.7% of all incidents, followed by transmission towers (22.1%), transformer oil-related incidents (14.3%), and underground cable vandalism (13.0%). The models achieved exceptional predictive performance, with GBM producing the highest accuracy (AUC = 0.981; Accuracy = 96.6%; Kappa = 0.913), followed by Random Forest (AUC = 0.980), the ensemble model (AUC = 0.980), and XGBoost (AUC = 0.978). Sensitivity exceeded 95.8% across all models, confirming strong capacity to identify high-risk locations. Historical incident density and proximity to previous incidents emerged as the strongest predictors, validating hotspot and repeat-victimization theories. National risk mapping identified concentrated vulnerability around Kampala, Mukono, Jinja, Masaka, Mbarara, Butembe, and major economic corridors, while several districts in Karamoja and northern Uganda exhibited consistently lower predicted risk.

The study provides one of the first national-scale predictive assessments of electricity infrastructure vandalism in Africa. The analysis indicates that vandalism risk is shaped by repeat victimization, accessibility, infrastructure exposure, economic activity, and security gaps, and that these drivers can be accurately modelled using geospatial machine learning. The resulting framework provides a practical decision-support tool for utilities, regulators, and security agencies seeking to prioritize patrols, harden vulnerable assets, monitor emerging corridors, protect MV/LV networks, and shift from reactive repairs to proactive infrastructure security planning.

This study demonstrates that electricity infrastructure vandalism in Uganda is a highly predictable spatial phenomenon rather than a random occurrence. By integrating 1,536 georeferenced vandalism incidents with GIS-based accessibility indicators, infrastructure exposure variables, hotspot analysis, and advanced machine-learning techniques, the study achieved exceptional predictive performance (AUC ≈ 0.98) and identified clear national risk patterns concentrated around major economic corridors and urban growth centres. The findings confirm that repeat victimization, infrastructure accessibility, market connectivity, and security gaps are the dominant drivers of grid crime, with distribution networks bearing the greatest exposure. Importantly, the resulting national risk intelligence framework provides utilities, regulators, and security agencies with a practical tool for transitioning from reactive asset replacement to proactive infrastructure protection. Targeted surveillance, strategic asset hardening, community engagement, and intelligence-led deployment of security resources in identified hotspot areas have the potential to substantially reduce vandalism losses, improve network reliability, safeguard public investments, and accelerate Uganda’s electricity access and energy transition objectives.