Governor-netizen/Keta-Intelligence-System

Cloud-based compound coastal flood prediction and driver attribution framework for data-scarce barrier coasts (Keta, Ghana). Fuses Sentinel-1 SAR, FES2022 tides, WaveWatch III, GPM IMERG, CHIRPS, SMAP L4, and ERA5 with leakage-audited XGBoost models and calibrated alert thresholds. # Keta Intelligence System: Cloud-Based Compound Coastal Flood Prediction & Driver Attribution > **A zero-ground-infrastructure geospatial machine learning framework for compound coastal flood susceptibility mapping, hydro-meteorological driver attribution, wave-tide runup reconstruction, and calibrated early warning on data-sparse barrier coasts.** --- ## 📌 Repository Description (For GitHub About Section) ```text Cloud-based compound coastal flood prediction and driver attribution framework for data-scarce barrier coasts (Keta, Ghana). Fuses Sentinel-1 SAR, FES2022 tides, WaveWatch III, GPM IMERG, CHIRPS, SMAP L4, and ERA5 with leakage-audited XGBoost models and calibrated alert thresholds. ``` --- ## 🌊 Overview Low-lying barrier coasts along the Gulf of Guinea, such as the **Keta barrier in southeastern Ghana**, experience recurring compound flooding driven by spring tides, swell-driven coastal surges, intense tropical rainfall, and Volta River dam releases (Akosombo Dam). In the absence of local tide gauges and telemetry-equipped weather stations, operational early warning requires cloud-accessible Earth Observation (EO) and global reanalysis data. The **Keta Intelligence System** integrates **Google Earth Engine (GEE)** multi-sensor satellite imagery, **PyFES FES2022 tidal hydrodynamics**, **NOAA WaveWatch III swell forcing**, and **leakage-audited XGBoost machine learning classifiers**. The system achieves out-of-time flood susceptibility predictions with 80.9% accuracy and provides probabilistic driver attribution (rain-dominant vs. coastal-dominant vs. compound) supported by hourly Total Water Level (TWL) forensic reconstructions. --- ## ✨ Key Features - **Multi-Sensor Satellite Fusion (GEE Pipeline v4.2)**: Fuses Sentinel-1 SAR change detection (VH polarization, Otsu thresholding), GPM IMERG V07 rain rates, CHIRPS antecedent rainfall, ECMWF ERA5 meteorology, NASA SMAP L4 surface soil moisture, and NASADEM/SRTM topography at 30 m resolution. …