ALPHA EARTH Canopy Height Modelling and Mapping (2020 and 2025)

This repository provides a reproducible Python/Google Colab workflow for AlphaEarth Foundation (AEF) embedding-based canopy-height modelling and mapping across Southwestern Nigeria for 2020 and 2025. It extracts the 64-dimensional AlphaEarth embedding bands, spatially matches them with quality-filtered GEDI RH98 observations, and prepares the resulting datasets for machine-learning prediction of canopy height. The workflow includes chunk-based AlphaEarth extraction with restartable processing, GEDI–embedding spatial matching, data cleaning and upper-tail filtering, XGBoost regression, five-fold cross-validation with out-of-fold predictions, and wall-to-wall canopy-height mapping. The same modelling framework is applied to 2020 and 2025, enabling consistent assessment of model performance and spatiotemporal canopy-height patterns.