Spatial Prediction of Opisthorchis viverrini Infection Risk Using Multi-Source Remote Sensing, Hexagonal Grid Optimization, and Ensemble Machine Learning: A Case Study From Northeastern Thailand

Opisthorchis viverrini (OV) infection is a major public health problem in northeastern Thailand, where it causes cholangiocarcinoma. This study developed a spatial prediction model for OV infection risk in Sakon Nakhon Province by integrating Sentinel-2, Sentinel-3, and CHIRPS data with hexagonal grid optimization and ensemble machine learning. Eleven environmental indices were aggregated using a newly developed Hd index, which identified an optimal hexagonal grid size of d = 2,100 m. Three models-Random Forest, XGBoost, and Multi-scale Geographically Weighted Regression (MGWR)-were compared using R ², root mean square error, ROC-AUC at exceedance thresholds (1%, 2%, 3%), 5-fold cross-validation, and SHAP analysis. XGBoost outperformed the other models, achieving the highest test-set R ² (0.894) and AUC (0.955) at the >3% threshold, with stable cross-validated performance (mean R ² = 0.83). SHAP analysis identified GVMI as the most influential predictor (mean |SHAP| = 0.125), followed by MNDWI (0.090) and SPI-6 (0.070). Moderate moisture conditions and above-average rainfall increased risk, whereas deep open water reduced it. The resulting risk map supports targeted surveillance and resource allocation for OV control in endemic areas.