The Qinghai–Tibetan Plateau is one of the least hospitable areas for agriculture. Nevertheless, humans have long conducted agricultural activities on the Qinghai–Tibetan Plateau, although the amount and extent of historical cropland remains unclear. We focused on the Qinghai–Tibetan Plateau and selected the Tubo Kingdom (AD 633–842) and the Yuan Dynasty (AD 1271–1368) as study periods to explore the human–Earth relationship in high–cold plateau areas. First, we estimated the population size of the Tubo Kingdom and extracted the household data of the Yuan Dynasty from historical literature. Then, to estimate the total cropland area of the Qinghai–Tibetan Plateau, we proposed a comprehensive method for cropland estimation due to the specificity of historical literature. Finally, we obtained the Qinghai–Tibetan Plateau cropland cover by allocating the cropland area into 5'×5' grids. Finally, the main findings were as follows. (1) Total cropland area was 3.87 × 10³ and 1.34 × 10³ km² during the Tubo Kingdom and Yuan Dynasty, respectively. (2) In the Tubo Kingdom, the fraction of cropland was highest in the Three-River Region (Yarlung Zangbo River, Lhasa River, and Nianchu River), with a gradual decline in the northeastward direction. During the Yuan Dynasty, cropland was concentrated in the Three-River Region and scattered in the valleys of the Nujiang and Lancang rivers. (3) Our reconstructions were highly consistent with regional development history, and addressed the gap in current global land cover datasets regarding the Qinghai–Tibetan Plateau. Thus, this study is vital to support prediction and simulation of regional and global environmental change.

Anthropogenic land use constitutes a key driver of land cover change, exerting profound impacts on terrestrial carbon/water cycles and ecosystems across global to regional scales. Consequently, long-term land cover datasets serve as fundamental data infrastructure for ecological effect assessment and paleoclimate modeling. Considering the dominant role of historical anthropogenic disturbances in vegetation change across the Hehuang Valley, we developed a 1 km × 1 km potential vegetation prior to land reclamation map through integrating remotely sensed land-use patterns with Random Forest-modeled vegetation predictions. Then, we derived changes in forest and grassland areas and their spatial patterns over the past two millennia by subtracting the spatially explicit cropland cover for six agricultural expansion stages. Finally, we compared our results with representative historical LUCC datasets. The main conclusions are as follows: (1) The potential vegetation of the Hehuang Valley was predominantly grassland, with forest occurring as a patchy and linear mosaic. Grassland and forest covered approximately 2.6 × 10⁴ km² and 0.7 × 10⁴ km², respectively. (2) Over the past two millennia, during the Han (202 BCE–220 CE), Tang (618–907 CE), Song (960–1279 CE), Ming (1368–1644 CE), and Qing (1644–1912 CE) dynasties, as well as the Republic of China period (1912–1949 CE), cropland reclamation reduced forest and grassland cover by 6% and 18%, respectively. Forest area decreased by 23.58–100.03 km², while grassland area decreased by 310.72–1513.05 km² across these periods. Grassland reduction was concentrated in valleys, whereas forest reduction was spatially scattered but locally intensive. These findings highlight the need to promote sustainable cropland development through technological innovation, improved management, and agricultural specialization rather than ecosystem conversion. (3) Compared with the HYDE 3.2 dataset and a national-scale reconstruction dataset, our framework better reflects the regional characteristics of the Hehuang Valley and provides a more detailed reconstruction of long-term forest and grassland changes.