Grassland Mowing Detection (GMD)

Grassland Mowing Detection (GEE Prediction Pipeline)

This repository runs a Google Earth Engine (GEE) batch prediction pipeline to detect grassland mowing events from Sentinel-2 time series. The notebook notebooks/prediction/gee_prediction.ipynb is the entry point: it defines a geometry, creates tiles, runs SatSelect to build features, predicts with a pre-trained Random Forest model stored in Earth Engine assets, and exports results to Google Drive.

What you need

• Python 3.9+ with Jupyter
• A Google Earth Engine account

Run the prediction notebook

Open the notebook and run all cells:

jupyter notebook notebooks/prediction/gee_Mowing_prediction.ipynb

Inside the notebook, you can change:

• region and the asset paths for aoi and mask
• years in the loop
• the export folder name (geeGMI) if you want a different Drive folder

The notebook will:

1. Define geometry and mask
2. Create tiles and filter them by geometry
3. For each tile and year:
   • create features
   • predict with the pre-trained model stored in EE assets
   • export results to Google Drive

Exports are started as Earth Engine tasks and will appear in your Drive after completion.

Project layout

.
├── notebooks/
│   ├── prediction/
│   │   └── gee_Mowing_prediction.ipynb    # Main GEE prediction notebook
│   └── development/
│       ├── 01_cross_validation.ipynb  # Model development
│       ├── 02_spatial_transfer.ipynb  # Model development
│       └── 04_model_development.ipynb # Model development
├── src/
│   └── functions.py               # GEE helper functions + model loader
├── requirements.txt
└── setup.py

License

MIT License. See LICENSE.