Nyando Basin Flood Risk Prediction System: A Production-Grade, Explainable Machine Learning Framework for Ward-Level Flood Susceptibility Mapping in Kisumu County, Kenya

Key Points

• The aim is to develop a machine learning system for predicting flood susceptibility in the Nyando River Basin.
• Utilized XGBoost classifier with AUC-ROC of 0.94.
• Trained on open satellite data including CHIRPS, NASA DEM, and Sentinel-1 SAR.
• Developed a Python package with a FastAPI prediction endpoint and a React web dashboard.
• Achieved an AUC-ROC of 0.94 indicating high prediction accuracy.
• Constructed a complete framework including source code, training data, and model weights.
• Ensured 100% compliance with Kenya DPA 2019 and GDPR standards.