Accurate information on tree density and its long-term dynamics is essential for understanding forest demography, ecosystem restoration, and terrestrial carbon cycling. However, existing tree-density products for China are generally limited by coarse spatial resolution, restricted temporal coverage, or a focus only on forested areas. Here, we developed a 30 m tree density dataset for China (ChinaTD30) spanning 1985–2022 by integrating four cycles of the National Forest Inventory (NFI), each comprising 670,279 plot observations, with long-term satellite observations, forest attributes, land-cover information, and environmental variables. Three modelling strategies were evaluated to balance prediction accuracy and temporal consistency, and a temporally unified model was adopted for the reconstruction of the released product. The resulting dataset characterizes tree density in both forest and non-forest ecosystems and resolves substantial spatial heterogeneity across China. Comparison with existing tree-density products showed large discrepancies in densely forested regions, partly reflecting differences in training data, spatial resolution, and model structure. The reconstructed time series reveals a pronounced long-term increase in China’s tree population. Forest trees increased from approximately 106 billion in 1985 to about 182 billion in 2022, an increase of approximately 72 %, while trees outside forests increased from about 2.3 billion to 5.1 billion. Tree number increases were evident not only across northern restoration regions but also in densely forested southeastern China. Tree density exhibited distinct age-related patterns between natural and planted forests, with densities generally peaking in youngto-middle-aged stands before declining in older stands. This dataset provides a spatially detailed and temporally consistent basis for investigating forest dynamics, ecological restoration, tree mortality, carbon cycling, and the contribution of trees outside forests across China. The dataset is available at <a href="https://doi.org/10.5281/zenodo.23100944" target="_blank" rel="noopener">https://doi.org/10.5281/zenodo.23100944</a> (Xian et al., 2026a), with the nine associated part records detailed in the Data availability section.