Abstract: Many studies have demonstrated that land use and cover change (LUCC) has played a key role in global environmental change. The contemporary land cover is a result of human land use in the history. In order to simulate the LUCC's influence in climate and ecosystem, it is important to have a historical LUCC dataset, especially high- resolution land cover dataset. However, in China, such national coverage dataset is still missing, and this has limited the national environmental change simulations. So there is an urgent need to develop an effective way to reconstruct historical cropland distribution with high-resolution grids. Considering the complexity of the natural environment in China, in this study we developed a separate reconstruction method. First, we divided China into four regions based on a qualitative analysis: the traditional cultivated region, the northeastern region, the northwestern region, and the Qinghai-Tibet Plateau. This division is mostly consistent with other recent studies except for the northwestern region, which differs slightly from common delineation. Second, in every region we examined the relationship between cropland distribution and various natural and human factors and built a reconstruction model. In the traditional cultivated region and the northeastern region, we found that elevation, slope, and population density were the main contributing factors to cropland distribution. In other regions, how-ever, population density was the sole significant contributing factor. This model was then used to reconstruct the cropland distribution of China in 1913, 1933, 1950, 1970, 1990 and 2000 at a spatial resolution of 10 km×10 km. By comparing the reconstruction result with remote sensing data interpretation for 1990, we found that the reconstructed cropland distribution data are reliable not only at the county scale, but also at the grid scale. The comparison between the reconstructed change and the remote sensing data-derived change from 1990 to 2000 also supports this view, that is, the separate reconstruction method developed in this study is effective for capturing cropland change over time. The reconstructed dataset indicates the follows. (1) In the northeastern region, the cropland area slightly decreased at the beginning of the People's Republic of China in 1949; up to 1970, the cropland area had recovered and the modern distribution pattern formed; thereafter, the Sanjiang Plain was brought into agricultural development gradually. (2) In Xinjiang in western China, the first cropland development climax appeared in the republican period influenced by the agricultural policy; the second climax appeared between the 1950s and the 1970s, but most of the cropland was distributed in the area of the Tianshan Mountains. (3) Change in cropland distribution of the Qinghai-Tibet Plateau was not notable, but the area had increased much; the spatial distribution of cropland in the traditional cultivated region also did not change significantly, but the reclamation ratio has increased. In conclusion, cropland area in China had increased in the early 20th century and then decreased, and the inflection point was likely in the late 20th century. This trend occurred not only in cropland area, but also in reclamation ratio. However, the change varies in different regions and is more pronounced in the northeastern and northwestern regions.

Key words: China, cropland distribution, grid, separate reconstruction