This dataset includes code and data for the manuscript of 'Zhao, R., Luo, X., Walker, A. P., Hoffman, F. M., Koh, L.P. (2025). Vegetation biogeography is a main source of uncertainty in modelling the land carbon cycle'. Update (Nov 2025): we have released the version 2 to include the raw data (Source_Data_Fig1.xlsx, Source_Data_Fig2.xlsx, Source_Data_Fig3.nc) and code (master_figure_clean.m, sup_code.zip) used for analysis and figure generation. Update (Jun 2025): we have released the version 1 to include the input, code and output for creating machine learning emulators for PFT GPP of each DGVM in TRENDY v9. The file 'input.zip' includes the predictor variables (i.e., annual precipitation, temperature, solar radition, pressure and wind speed during 2001-2019 from CRU JRA v2.1 datasets) and response (i.e., pixel PFT GPP rate) for each PFT of each DGVM. The file 'code4emulators.mlx' includes the code for training machine learning emulators (i.e., Random Forest models) for each PFT of each DGVM. The file 'model_emulators.zip' includes all trained emulators.