Skip to content

A Google Earth Engine Land use (crops) classification workflow using Random Forest, one year of ground data, Sentinel-2, and Landsats; to produce multiyear annual 30-m crop maps

License

Notifications You must be signed in to change notification settings

ikramelhazdour/GEE_LULC_Multi-year_Classification

Repository files navigation

GEE LULC Multi-year Crop Classification

A comprehensive land use and crop classification workflow that utilizes Machine Learning for multi-year analysis using satellite imagery from both Sentinel-2 and Landsat.

Overview

This workflow implements a Random Forest classifier trained on a single year's ground data to perform land use and crop classification across two decades of satellite data. The model leverages seasonal median composites and multiple vegetation indices to achieve robust classification performance.

Input Data

Sentinel-2:

Surface reflectance data

Landsat Series:

Surface reflectance data (Landsat 5, 7, 8, 9)

Vegetation Indices

NDVI, NDWI, EVI2

Spectral Bands

Surface reflectance values from visible, NIR, and SWIR regions

Ground Data

Single year of training data for model training collected in Morocco Applied to classification across two decades (2000-2023)

STEPS

1. Data Preprocessing

Cloud masking and atmospheric correction Seasonal compositing using 45-day windows Calculation of vegetation indices (NDVI, NDWI, EVI)

2. Feature Compositing

Median value calculation per season Combination of spectral bands and indices Multi-sensor data harmonization

3. Model Training

Random Forest classifier implementation Training on single-year ground reference data Hyperparameter optimization

4. Classification

Application to multi-year satellite data Annual land use/crop classification Output generation (Rasters) for analysis

Technical Requirements

Google Earth Engine account

Usage

This workflow is designed for researchers and practitioners in remote sensing, agriculture, and environmental science who need to perform large-scale land use classification with limited ground truth data.

Note

The unique aspect of this workflow is its ability to leverage a single year's training data to perform classification across multiple years, making it particularly valuable for historical analysis where ground truth data may be limited.

Citation

The research paper related to this work will be available soon

About

A Google Earth Engine Land use (crops) classification workflow using Random Forest, one year of ground data, Sentinel-2, and Landsats; to produce multiyear annual 30-m crop maps

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published