nyandajr/tanzania-climate-watch
Domain:
climateRecord type:
project
Creator:
nya
Host:
# Tanzania Climate Prediction Pipeline
Real-time weather monitoring and ML forecasting for 5 Tanzanian cities, updated automatically every 15 minutes via GitHub Actions.
**Live Dashboard →**
---
## Cities Covered
| City | Coordinates |
|------|-------------|
| Dar es Salaam | -6.79°, 39.21° |
| Dodoma | -6.17°, 35.74° |
| Arusha | -3.39°, 36.68° |
| Mwanza | -2.52°, 32.92° |
| Zanzibar | -6.17°, 39.20° |
## Pipeline Architecture
```
Open-Meteo API (free, no key)
│
fetch.py → data/raw/latest.json
│
preprocess.py → data/processed/climate_metrics.csv
│
predict.py → data/predictions/latest.json
│ ml/inference_state.json
anomaly.py → data/anomalies/anomaly_log.csv
│
dashboard.py → docs/data.json (GitHub Pages reads this)
│
GitHub Actions → git commit + push (1 commit/run)
```
## ML Model
- **Algorithm**: Ridge Regression (scikit-learn)
- **Features**: cyclic hour/day encoding, temperature, humidity, pressure, windspeed
- **Target**: next-hour temperature
- **Metric**: MAE tracked over time per city
- **Cold start**: API forecast used until 20 data points collected
## Features
- Live current conditions for all 5 cities
- 7-day temperature max/min forecast charts
- 30-day historical temperature trends
- Model accuracy (MAE) over time chart
- Anomaly detection (z-score + hard thresholds)
- Weekly markdown reports (auto-generated every Monday)
## Data Source
Open-Meteo — free, open-source weather API, no API key required.
## Local Development
```bash
pip install -r requirements.txt
python scripts/fetch.py
python scripts/preprocess.py
python scripts/predict.py
python scripts/anomaly.py
python scripts/dashboard.py
# open docs/index.html in browser (serve via local HTTP server)
python -m http.server 8080 --directory docs
```
## GitHub Pages Setup
1. Go to repo **Settings → Pages**
2. Source: **Deploy from a branch**
3. Branch: `main` / folder: `/docs`
4. Save — your dashboard will be live at `nyandajr.gith …