Abstract. High-spatial-resolution and long-term data on forest cover and plant functional types (PFTs) are crucial for elucidating the effects of forest cover change on the national terrestrial carbon balance. Since the 1980s, China has experienced a substantial expansion in its forested area, primarily driven by large-scale national afforestation programs. However, existing land cover products often fail to capture this long-term increasing trend, leading to an underestimation of forest cover change–related ecological processes. Here, we developed a high-resolution (1 km), annual forest cover dataset for China for 1981–2023. This dataset integrates spatial constraints from multisource remote sensing data with provincial-level statistics from China's national forest inventories (NFIs), providing a consistent and spatially explicit record of forest dynamics over four decades. Building on this primary dataset, we further produced an annual PFT dataset that disaggregates total forest cover into nine distinct plant functional types suitable for use in dynamic global vegetation models (DGVMs). Validation against independent data indicates that our reconstructed dataset achieves an overall accuracy (OA) of 84.86 % ± 1.18 % for five aggregated forest types (evergreen needleleaf forests, evergreen broadleaf forests, deciduous needleleaf forests, deciduous broadleaf forests, and mixed leaf forests), and it reproduces NFI-consistent forest dynamics (R2≈1). To evaluate its applicability, we implemented the dataset in the Lund–Potsdam–Jena General Ecosystem Simulator (LPJ–GUESS). Compared with the widely used PFT dataset from the European Space Agency's Land Cover Climate Change Initiative (ESA CCI) and the MODIS land cover type product (MCD12Q1), our product yields a markedly improved simulation of key biophysical and biogeochemical processes in China. Specifically, it reduced errors relative to over extensive regions, outperforming these baselines across 77.7 % and 85.2 % of the terrestrial area for gross primary productivity (GPP), 63.1 % and 69.7 % for net ecosystem exchange (NEE), 66.9 % and 77.3 % for the leaf area index (LAI), and 78.7 % and 85.3 % for actual evapotranspiration (ET). With its high spatial resolution, long-term temporal coverage, and detailed forest type classification, our dataset offers a robust foundation for assessing the ecological impacts of forest restoration and for constraining estimates of China's forest carbon sink since 1981. The dataset is freely available at https://doi.org/10.5281/zenodo.18448036 (Liu et al., 2026).
Liu et al. (Tue,) studied this question.