This dataset contains the annual woody vegetation's aboveground biomass (AGB) and belowground biomass (BGB) maps in China from 2003 to 2020. The spatial resolution is 1km (1/120 degree), while the uncertainty maps of AGB and BGB are also provided. To develop this dataset, we first integrated the up-to-date satellite-based forest AGB and height maps by referring to large quantities of woodland plots throughout China. Next, we developed an improved vegetation water content (VWC) dataset covering 2003~2020 by fusing the vegetation optical depth retrieved from various microwave remote sensors. VWC and the vegetation cover derived from optical remote sensing were subsequently linked to the calibrated reference AGB map for a continuous AGB mapping in China during 2003~2020, which was downscaled to ~ 1 km resolution. Afterwards, based on the collated AGB and BGB in forest and shrubland plots, we developed the random forest model or regression model that can transform the plot-level AGB into the woodland plots' BGB. Finally, we solved the problem resulting from the scale difference between plot measurements and satellite observations to ensure the application of the plot data-based models.