Multi-hazard risk mapping using machine learning

This study maps out Ghana’s multi-hazard risk of flood and drought by using machine learning (ML) models for susceptibility analysis, socioeconomic survey for vulnerability analysis and population density for exposure analysis. The ML models used were Logistic Regression (LR), Random Forest (RF) and Support Vector Machine (SVM) with inputs of location and features of natural hazards. Topographic, precipitation, temperature, hydrology, land cover and soil cover raster images were also used in these models. The value of the Area Under the Curve (AUC) of Receiver Operating Characteristic Curve (ROC) was above 0.80 for all models except the LR model for drought classification. The best performing model was RF, with an AUC of 0.84 and 0.82 for flood and drought classification.