Dickson32-cell/Geohazard-Information-System-GeoHIS-

GeoHIS (Geohazard Information System) is a web-based platform designed to help municipal planners, emergency managers, and residents understand flood and landslide risks in New Juaben South Municipality, Ghana. # GeoHIS: Geohazard Information System An open-source web-based Geohazard Information System for flood and landslide susceptibility mapping and risk assessment. ## Overview GeoHIS integrates Multi-Criteria Decision Analysis (MCDA) techniques for municipal-level geohazard assessment: - **Flood Susceptibility**: Analytical Hierarchy Process (AHP) with 5 conditioning factors - **Landslide Susceptibility**: - **Statistical**: Frequency Ratio (FR), Information Value (IV), Certainty Factor (CF) - **Machine Learning**: Random Forest, XGBoost, Support Vector Machine (SVM), Logistic Regression - **Risk Assessment**: Hazard × Exposure × Vulnerability framework - **Scientific Rigor**: Spatial Cross-Validation, Uncertainty Quantification, Sobol Sensitivity Analysis **Validation Results**: AUC-ROC > 0.90 for ensemble models with 95% Bootstrap Confidence Intervals. ## Key Features ### Advanced Analysis Engine - **Spatial Cross-Validation**: Implements checkerboard splitting to prevent spatial autocorrelation data leakage. - **Uncertainty Quantification**: Provides epistemic uncertainty maps (prediction standard deviation) for decision support. - **Global Sensitivity Analysis**: Sobol indices to quantify input factor contributions and interactions. - **Hyperparameter Tuning**: Automated grid/random search for optimal model performance. ## Study Area New Juaben South Municipality, Eastern Region, Ghana (110 km², population 183,000) ## Architecture ``` ┌─────────────────────────────────────────────────────────┐ │ PRESENTATION LAYER │ │ React.js + Leaflet.js + Bootstrap │ ├─────────────────────────────────────────────────────────┤ │ APPLICATION LAYER │ │ FastAPI + Python Analysis │ ├─────────────────────────────────────────────────────────┤ │ DATA LAYER │ │ PostgreSQL/PostGIS + GeoJS …