Above Ground Biomass (AGB) Prediction using Random Forest

Project Overview

This project aims to predict Above Ground Biomass (AGB) using satellite imagery and machine learning techniques. It utilizes data from Sentinel-1, Sentinel-2, and ALOS-2 satellites, preprocessed in Google Earth Engine (GEE), and then uses Random Forest regression in Python to create a robust prediction model.

Features

• Data preprocessing using Google Earth Engine (GEE)
• Integration of multiple satellite data sources (Sentinel-1, Sentinel-2, ALOS-2)
• Random Forest regression model for AGB prediction implemented in Jupyter notebooks
• Comprehensive model evaluation metrics
• Visualization of results and model performance
• GeoTIFF output for predicted AGB values

Workflow

1. Data Preprocessing (Google Earth Engine)

   • Acquisition and preprocessing of Sentinel-1, Sentinel-2, and ALOS-2 data
   • Calculation of various spectral indices
   • Export of processed data as GeoTIFF files

2. AGB Prediction (Jupyter Notebook)

   • Data loading and preparation
   • Random Forest model training and prediction
   • Model evaluation and visualization
   • Export of predicted AGB as GeoTIFF

Requirements

• Google Earth Engine account
• Python 3.7+
• Jupyter Notebook
• pandas
• numpy
• scikit-learn
• matplotlib
• rasterio

Usage

Google Earth Engine Preprocessing

1. Log in to your GEE account
2. Use the provided GEE script to preprocess satellite data
3. Export the results as GeoTIFF files

Jupyter Notebook Analysis

1. Open the Jupyter notebook agb_prediction.ipynb
2. Update file paths to point to your preprocessed data
3. Run the notebook cells sequentially to:
   • Train the Random Forest model
   • Make predictions on the full dataset
   • Generate evaluation metrics and plots
   • Save the predicted AGB as a GeoTIFF file

Output

• Predicted AGB GeoTIFF file
• Evaluation metrics (R², MSE, MAE)
• Visualization plots:
  • Actual vs Predicted AGB
  • Residual plot
  • Feature importance
  • Error distribution
  • QQ plot of prediction errors