Satellite NDVI combined with LightGBM for crop yield prediction in North-Central Nigeria with a Streamlit dashboard.
# Agricultural Crop Yield Forecasting & Farmland Monitoring
Satellite-imagery and ML-powered yield prediction platform for smallholder farmers across North-Central Nigeria, integrating NDVI time-series, weather, soil properties, and farm management data to forecast crop yields per hectare per growing season.
---
## Problem Statement
Smallholder farmers in Nigeria's food belt lack access to yield forecasts, making it impossible to plan inputs, storage, or marketing. This platform provides per-farm, per-season yield forecasts to improve food security and income planning.
---
## Features
| Feature | Description |
|---------|-------------|
| Multi-Crop Support | Maize, Sorghum, Millet, Cassava, Yam, Rice, Groundnut |
| NDVI / EVI Integration | Sentinel-2 and Landsat-8 spectral indices |
| LightGBM Regression | Yield prediction with feature importance ranking |
| Leave-One-Season-Out CV | Robust temporal generalisation testing |
| Streamlit Dashboard | Yield map, season trends, custom forecasting tool |
---
## Input Features
| Feature | Source |
|---------|--------|
| `ndvi_mean` | Sentinel-2 / Landsat-8 |
| `rainfall_mm` | NIMET rainfall stations |
| `soil_quality` | FAO HarvestChoice soil data |
| `temperature_mean` | ERA5 reanalysis |
| `farm_management` | Extension officer surveys |
| `crop_type` | Satellite classification |
---
## Tech Stack
| Layer | Technology |
|-------|-----------|
| Remote Sensing | NDVI/EVI from Sentinel-2, Landsat-8 |
| Machine Learning | LightGBM, scikit-learn |
| Geospatial | GeoPandas, Folium |
| Dashboard | Streamlit, Plotly |
| Data | pandas, NumPy |
---
## Quick Start
```bash
git clone
github.com
cd crop-yield-forecasting
pip install -r requirements.txt
streamlit run streamlit_app.py
```
---
## Author
**Momah Moses**, Geospatial AI Engineer & Data Scientist
GitHub · Portfolio