Dataset and Code for "Estimating Global Site-Level Net Ecosystem Exchange with a Geographically Weighted XGBoost Framework"

The two Excel spreadsheets contain datasets of flux observations and remote sensing-derived predictor variables, corresponding respectively to model training with internal validation, and independent external validation. The provided code includes implementations of Random Forest and XGBoost for global modeling, as well as geographically weighted Random Forest and geographically weighted XGBoost for geographically weighted modeling. The PDF file (titled Overview of flux tower sites used in this study.pdf) summarizes all flux tower sites employed for model development and validation. Detailed descriptions of all provided files, including their contents, purposes, and usage information, are available in the accompanying Readme.pdf file.