A Multi-Pathway Contamination Risk Model for Niger Delta Communities: Integrating Hydrocarbon Load, Heavy Metal Exposure, and Vegetation Stress Indices from Heterogeneous Observational Data

Dickson Ojochogwu Dickson 

The Niger Delta is one of the most petroleum-impacted environments in the world, yet spatially explicit contamination risk assessments that integrate multiple exposure pathways remain scarce. This study presents a Composite Risk Score (CRS) model that fuses heavy metal soil and water measurements from nine peer-reviewed studies, total petroleum hydrocarbon (TPH) data from the 2011 UNEP Environmental Assessment of Ogoniland, 15,630 georeferenced oil spill incidents from the National Oil Spill Detection and Response Agency (NOSDRA), and pre/post-spill Sentinel-2 NDVI indices for 500 spill events. Two XGBoost models were trained on 106 georeferenced sample points to predict exceedance of established contamination thresholds across two spatially extrapolatable pathways: hydrocarbon load exceeding EGASPIN limits, and normalised heavy metal burden relative to WHO reference values. The human health pathway was characterised separately by computing a hazard index directly from measured concentrations at each sample point, following the USEPA oral ingestion framework, rather than by model prediction, since it cannot be extrapolated to ungauged locations without measurements. The two model outputs were combined into a single CRS through PCA-derived pathway weighting and spatially interpolated across a 1 km2 grid covering Rivers, Bayelsa, Delta, Edo, Akwa Ibom, and Cross River States. The IDW-model blend ratio was optimised empirically via spatial leave-one-out cross-validation. The TPH exceedance classifier achieved a cross-validated AUC of 0.798 and the metal index regressor a cross-validated RMSE of 0.024. The interpolated risk surface identified 63 critical and 551 high-risk grid cells, concentrated in Ogoniland and the Warri-Escravos corridor. Sabotage-related spill density within 25 km explained the largest share of variance in the CRS, consistent with the predominantly anthropogenic contamination pattern documented in prior literature. The dataset and interactive risk map are publicly available at https://github.com/Ojochogwu866/nd-public and https://research-map.ojochogwu.dev/.