Occupational Health Risk Mapping in the Niger Delta Oil and Gas Fields using Machine Learning and Environmental Exposure Indices

Abstract

The characterisation of occupational exposures in the Niger Delta oil and gas fields remains poor because of the nonlinear, dynamic, and spatially heterogeneous nature of occupational hazards associated with petroleum, as assessed using conventional monitoring methods. The research employed a combination of machine learning, an environmental exposure index, and a GIS-based model to forecast, categorise, and map occupational health hazards at drilling rigs, compressor stations, flare sites, processing units, and administrative areas in Rivers, Bayelsa, and Delta States. Six-month environmental monitoring included high-resolution measurements of PM2.5, PM10, VOCs, benzene, noise levels, WBGT, and meteorological parameters. Random Forest, XGBoost, ANN, and SVM will be used to develop composite indices (AQI, NEI, WBGT, HEI) using four models. IDW/Kriging and Getis-Ord Gi* were utilised to perform the spatial analyses. Findings indicated continuous overlimit of the WHO/NIOSH limits, with 87% and 63% of locations exceeding PM2.5 and heat-stress levels, respectively. Random Forest achieved the best results (Accuracy = 91.3%, AUC = 0.96) and identified WBGT, benzene, and PM2.5 as the most significant predictors. The GIS products showed hotspots of high-risk characteristics in the flare stacks and compressor stations. The presence of synergistic heat-chemical exposure interactions was supported by SHAP interpretation. The research concludes that the nature of occupational hazards in the Niger Delta is that of a multi-hazard, spatially clustered ecosystem that requires real-time and predictive surveillance. It suggests incorporating ML-GIS risk systems, specialised engineering control, heat-regulation strategies, and risk-specific PPE distribution to enhance the HSE management.