High Spatiotemporal Resolution (8-Day, 500 m) Net Ecosystem CO₂ Exchange (NEE) Dataset for the Three-North Shelterbelt Engineering Region of China (2001–2023)

The dataset was generated using a random forest model trained with observations from 43 eddy covariance flux towers and integrated with multi-source data, including MODIS remote sensing products, climate data, and terrain data.

By incorporating ecological types and their process characteristics, the model effectively improves the representation of carbon exchange processes across different ecosystems and enhances its spatial generalization capability. Model evaluation demonstrated high simulation accuracy, with a coefficient of determination (R²) of 0.86 and a root mean square error (RMSE) of 0.84 g C m⁻² d⁻¹ for the validation dataset.