# Sahan — AI Enabled Early Warning System for Somalia
Real-time satellite-driven drought, rainfall, vegetation, soil moisture, and
flash-flood monitoring for Somalia, with AI-generated plain-language
situation briefs for decision-makers. Built for the **IGAD Hackathon 2026**
("Smarter Early Warning, Stronger Communities") — selected as a **Top 10
finalist**, advancing to the physical evaluation workshop at ICPAC, Nairobi.
**Live demo:**
sahan-somalia-ews.vercel.app
*(backend runs on a free tier and may take 30–60s to wake up on first load)*
## What it does
- Interactive map of Somalia with 10 satellite-derived index layers (NDVI,
CHIRPS rainfall, SPI, SPEI, SMI, VHI, TCI, BSI, land-surface temperature,
temperature anomaly) plus a combined drought index (CDI).
- 1/3/7/14-day rainfall and temperature forecasts (NOAA GFS).
- Flash-flood risk overlay combining static terrain/soil/land-cover
susceptibility with live rainfall forecasts, across 5,000+ river sub-basins.
- Click any point on the map for a 12-month time-series chart of the active
index, with CSV/JSON export.
- Live water-source monitoring overlay (~1,900 surveyed boreholes, wells,
dams, berkads and springs) with functioning status and water-quality data.
- AI-generated narrative insights (situation, drivers, impact, risk level,
recommendations) for any region or district, grounded in the real computed
indices — see `docs/METHODOLOGY.md` for exactly how
each index is calculated and what's disclosed as a simplification, or
`docs/SAHAN_TECHNICAL_INSTITUTIONAL_REPORT.md`
for the full methodology writeup with file/function-level citations.
## Architecture
```mermaid
flowchart LR
subgraph Frontend["Browser — React + Leaflet (Vite)"]
Map["Map / Layer Controls"]
Panel["Stats Panel + AI Insights"]
end
subgraph Backend["FastAPI Backend (Python)"]
Routers["Routers: layers / stats / insights / admin / alerts"]
Services["Services: gee_utils, risk_engine, flood_processor"]
end
GEE["Google Earth Engine\n(CHI …