Logo Lanfrica

kjdartaugust/flood-risk-ghana

Domaine:

climatemobility

Type de record:

software
Créateur:
kjd
Hôte:
# FloodWatch Ghana 🌊 A production flood-risk platform for Ghana. Two combined features: 1. **Flood-risk scoring & mapping** — check a flood-risk score for any location/area in Ghana before buying land, building, living, or starting a business. Built from historical flood records, rainfall time-series, elevation/DEM, drainage/slope, land cover, and known Accra hotspots (Kaneshie, Circle, Adabraka, Alajo). 2. **Trotro route flood alerts** — combines the risk layer with live/forecast rainfall to predict which trotro routes flood before and during rain, and pushes route-level warnings to commuters. ## Architecture ``` ┌──────────────┐ Open data ──ETL──▶│ PostGIS │◀── H3 hex scoring tiles (GPM/IMERG, │ (Supabase) │ SRTM DEM, └──────┬───────┘ Sentinel LC, │ OSM roads) ┌──────▼───────┐ ┌──────────────┐ │ FastAPI │◀────▶│ Redis │ cache + rate limit ML model ────────▶│ (inference) │ └──────────────┘ (logistic + └──────┬───────┘ gradient boost) │ OpenAPI / JWT (Supabase Auth) ┌──────▼───────┐ │ Next.js + │ MapLibre GL interactive risk map │ React │ + trotro route alerts └──────────────┘ ``` | Layer | Tech | |-----------|----------------------------------------------------------| | Frontend | Next.js 14 (App Router), React, MapLibre GL, TypeScript | | Backend | FastAPI, SQLAlchemy (async), Alembic, Pydantic v2 | | Data | Supabase Postgres + PostGIS, H3 hex index, Redis | | ML | scikit-learn (logistic baseline), LightGBM (upgrade) | | ETL | httpx workers, APScheduler cron (rainfall refresh) | | Infra | Docker Compose (local); Neon + Render + Vercel (deploy) | ## Quick start (local) ```bash cp .env.example .env # fill in Supabase + Mapbox/MapLibre keys docker compose up --build # postgis + redis + backend + worker + frontend # backend → localhost

Languages