Malaria outbreak of Ethiopia Early Warning System
# Ethiopia Malaria Early Warning System (MEWS)
Climate-driven malaria risk monitoring for Ethiopia, built entirely on **Ethiopia-specific, real,
freely-accessible data** — no pan-African or global datasets standing in for national ones.
## What's in this repo
| File | Purpose |
|---|---|
| `Ethiopia_MEWS_Colab.ipynb` | Full pipeline — fetch data, engineer features, train the risk model, export dashboard JSON. Run in Google Colab (needs open internet; this sandbox cannot fetch these APIs itself). |
| `Ethiopia_MEWS_Dashboard.html` | Single-file dashboard. No build step. Reads `data/dashboard/*.json`; falls back to labeled seed data if missing. |
| `scripts/fetch_ethiopia_data.py` | Headless equivalent of the notebook, for CI. |
| `requirements.txt` | Python deps for the script above. |
| `update-ews-data.yml` | GitHub Actions workflow — **move this to `.github/workflows/`** in your repo. |
## Data sources (Ethiopia-only)
- **Climate** — NASA POWER API, queried per Ethiopian region (temperature, rainfall, humidity). Free, no key.
- **National malaria burden** — WHO GHO OData API, `SpatialDim = ETH`. Free, no key.
- **ENSO phase** — NOAA Niño 3.4 index (East African rainfall teleconnection).
- **Regional 2024 case/death share** — WHO Disease Outbreak News, Ethiopia, 31 Oct 2024 (cited in the notebook).
**Known limitation, stated plainly:** Ethiopia's real sub-national monthly case counts live in EPHI's
PHEM/HMIS system, which has no public API. The regional-monthly series here apportions the real
national WHO trend by published regional shares + Ethiopia's kiremt/belg seasonality — it is a
documented approximation, not raw surveillance data. Swap in EPHI/PHEM or DHS figures once you have
institutional access (Cell 6 of the notebook / the equivalent block in the script) for a validated result.
## Deploy the dashboard on GitHub Pages
1. Push this folder structure to a GitHub repo.
2. Move `update-ews-data.yml` into `.github/workflows/update-ews-data.yml`.
3. Sett …