umaribrahimkambiya/HI-EWS-NIGERIA

# HI-EWS-NIGERIA — REAL DATA V2.2 This package is a clean, reproducible Nigeria LGA flood early-warning pipeline. ## Real sources - 774-LGA geometry: geoBoundaries ADM2. - Administrative P-codes/names: BNDA reference supplied with the project. - Population: WorldPop Nigeria v3.0 LGA admin release. - Rainfall: CHIRPS v2.0 daily 0.05° (2018–2025), spatially averaged to each LGA. - Flood impacts: NEMA/OCHA Nigeria 2022 Flood_affected_areas ArcGIS layer. The query uses JSON instead of the failing GeoJSON request and expands first/last flood dates to daily observations. - Flood exposure: World Bank Space2Stats ADM2 flood exposure (population exposed >15 cm, 1-in-100 return period). ## Important scientific limitation The NEMA/OCHA flood layer is a 2022 LGA-level impact source. It is not a complete 2018–2025 LGA event archive. The package therefore does NOT pretend that it is. The model's historical label coverage is determined by the observed source. DFO is retained as an external historical validation/reference source rather than fabricated into LGA labels. ## Install ```powershell python -m venv .venv .\.venv\Scripts\Activate.ps1 python -m pip install --upgrade pip pip install -r requirements.txt ``` Place the supplied reference files here: ```text data/raw/reference/BNDA_NGA_2000-01-01_lastupdate.csv data/raw/reference/NGA_population_v3_0_admin/lga_pop_total_scaled.csv ``` ## Run everything ```powershell python scripts/run_all.py ``` CHIRPS downloads eight yearly NetCDF files and can take time and disk space. ## Run stages individually ```powershell python src/data/fetch_boundaries.py python src/data/build_lga_master.py python src/data/fetch_population.py python src/data/fetch_rainfall.py python src/data/fetch_flood_impact.py python src/data/fetch_vulnerability.py python scripts/real_data_preflight.py python src/preprocessing/build_panel.py python src/models/train.py ``` ## Dashboard ```powershell streamlit run dashboard/app.py ``` ## Outputs - `data/raw/refe …