antarctica-aws/ea-forest-restoration

Code and workflow for East Africa forest restoration effectiveness analysis # Time-Varying Effects of Forest Restoration Shape Climate Mitigation Outcomes in East Africa ## Contact Corresponding author: Mengyu Liang (mliang77\@stanford.edu) ## Overview This repository contains the data and code developed for the manuscript on assessing forest restoration stretagies' climate mitigation potential in East Africa. ## Methods Summary The product generation and analysis of this project were conducted using a combination of programming languages to optimize computation efficiency and improve data visualization: - **GEDI data processing, machine learning model development, and aboveground biomass map generation**: Python 3.7 - **Landsat data processing and predictor variable derivation**: Google Earth Engine - **Econometric analysis (staggered difference-in-difference)**: R 3.6 - **Data visualization**: R 3.6 and QGIS 3.28.14 ## Repository Structure ``` ├── data/ │ └── boundaries/ # Study area boundary files (GeoJSON, shapefiles) ├── scripts/ │ ├── 01_data_acquisition/ # GEDI data download and subsetting │ ├── 02_preprocessing/ # Data preparation, Landsat extraction, and covariate-based matching │ ├── 03_modeling/ # XGBoost biomass modeling notebooks │ ├── 04_validation/ # Model validation scripts (R) │ └── 05_analysis/ # Statistical analysis and figure generation └── docs/ # Additional documentation when needed ``` ## Workflow Documentation ### 01_data_acquisition 1. Search for GEDI L4A footprints within the area of interest (AOI) 2. Subset GEDI orbits over the AOI 3. Convert GEDI HDF5 files to tabular CSV format 4. Search for GEDI L2A and L2B products over the AOI 5. Subset L2A and L2B products over the AOI ### 02_preprocessing 1. Landsat preprocessing using Google Earth Engine 2. Extraction of Landsat predictors at GEDI footprint locations 3. Statistical matching and preparation of modeling inputs ### 03_modeling 1. Development of tile-level XGBoos …