avishka19-web/FHRI-west-africa-early-warning-system_Data_Science

AI-powered Forest Health Risk Index (FHRI) Prediction and Responsive Early Warning System for West Africa using XGBoost, LSTM, GAN, Clustering, Remote Sensing, and Streamlit Dashboard. # AI-Based Forest Health Risk Index (FHRI) Prediction and Early Warning System ## West Africa — Nigeria · Ghana · Ivory Coast · Burkina Faso · Mali --- ## Project Overview A complete production-level end-to-end Data Science + AI system that: - Predicts the **Forest Health Risk Index (FHRI)** using real-world environmental datasets - Detects **wildfire-prone zones** and forest degradation - Sends **automated EMAIL and SMS alerts** when risk exceeds thresholds - Provides a **real-time Streamlit dashboard** with GAN visualizations and Agentic AI monitoring --- ## Architecture ``` Data Sources (NASA FIRMS, MODIS, ERA5, CHIRPS, GFW) │ ▼ Data Download Pipeline ──► Raw Data (CSV) │ ▼ Preprocessing & FHRI Computation │ ├──► Feature Engineering (lag, rolling, geospatial) │ │ │ ┌─────┼──────────┐ │ ▼ ▼ ▼ │ Classification Clustering DNN / LSTM / GAN │ (DT/RF/XGB) (KMeans/DBSCAN) (TensorFlow/Keras) │ │ ▼ ▼ Static Maps Agentic AI (LangChain/Rule-based) Folium Maps │ │ Alert System (Email + SMS) ▼ │ Streamlit Dashboard ◄─┘ ``` --- ## Dataset Details | Dataset | Source | Variables | |---------------|----------------------|----------------------------------| | NASA FIRMS | firms.modaps.eosdis.nasa.gov | Fire location, FRP, brightness | | MODIS NDVI | NASA/MODIS | Vegetation index (16-day) | | ERA5 Climate | Copernicus/ECMWF | Temperature, rainfall, wind, humidity | | Soil Moisture | ESA CCI | Surface soil moisture | | Drought Index | NOAA/FAO | SPI-based drought index | | Forest Loss | Global Forest Watch | Annual forest loss (hectares) | --- ## Installation ### Prerequisites - Python 3.10+ - pip or conda - Git ### Steps ```bash # 1. Clone or extract project cd fhri_project # 2. Create virtual environment python -m venv venv source venv/bin/activate # L …