Logo Lanfrica

rache3/floodwatch-ghana

Domain:

geospatialclimate

Record type:

softwareproject
Creator:
rac
Host:
Cloud-native flood risk intelligence system for Greater Accra, Ghana — automated 5-layer spatial pipeline serving live district-level risk scores # FloodWatch Ghana — Flood Risk Intelligence Application ### Greater Accra Region **Author:** Rachel Atia · GeoBuilders Africa **Version:** v0.1 (Baseline Risk) **Live map:** floodwatch.geobuildersafric… **Company:** geobuildersafrica.com **GitHub:** github.com --- ## Overview An automated flood risk mapping pipeline for Greater Accra, Ghana. The system ingests five geospatial datasets, normalises and combines them into a weighted composite risk score, and serves the output as a live interactive web map updated monthly. Greater Accra is the pilot region. The pipeline is designed to scale across all regions of Ghana and other African cities. --- ## Live map > **floodwatch.geobuildersafric… - Flood risk raster rendered with plasma colormap (0 = low risk, 1 = high risk) - All 29 districts of Greater Accra with boundary overlays - Click any district for mean, max, median, std deviation, and risk distribution - Layer toggles and opacity control - Validation badge — validated against May 18, 2025 Greater Accra floods --- ## Validation FloodWatch Ghana v0.1 was validated against the **May 18, 2025 Greater Accra flood event** — 132mm of rainfall, 4 deaths, 3,000+ displaced. Flooded districts were sourced from The Watchers, GDACS, and Copernicus EMS. **Result: Recall 0.86 · F1 0.40 · 6 of 7 flooded districts correctly flagged as high risk** | Metric | Value | |---|---| | Flooded districts mean risk | 0.782 | | Non-flooded districts mean risk | 0.765 | | Flooded districts flagged high risk (≥0.70) | 6 of 7 (86%) | | Recall | 0.86 | | Precision | 0.26 | | F1 Score | 0.40 | Six of 7 flooded districts — Tema West, Adenta, Tema, Ga East, La-Nkwantanang-Madina, Weija Gbawe — scored above 0.70. Accra Metropolis scored 0.636, narrowly below the high-risk threshold. The low precision (17 of 22 non-flooded districts also score high) reflects a known limitation of static structural models applied to a re …