Three global datasets, the History Database of the Global Environment (HYDE), Kaplan and Krumhardt (KK) and Pongratz of reconstructed anthropogenic land cover change (ALCC) were introduced and compared in this paper. The HYDE dataset was reconstructed by Goldewijk and his colleagues at the National Institute of Public Health and the Environment in Netherland, covering the past 12 000 years. The KK dataset was reconstructed by Kaplan and his colleagues, the Soil-Vegetation-Atmosphere Research Group at the Institute of Environmental Engineering in Switzerland, covering the past 8000 years. The Pongratz dataset was reconstructed by Pongratz and her colleagues at the Max Planck Institute for Meteorology in Germany, covering AD 800-1992. The results show that the reconstructed datasets are quite different from each other due to the different methods used. The three datasets all allocated the historical ALCC according to human population density. The main reason causing the differences among the three datasets lies on the different relationships between population density and land use used in each reconstructed dataset. The KK dataset is better than the other two datasets for two important reasons. First, it used the nonlinear relationship between population density and land use, while the other two used the linear relationship. Second, Kaplan and his colleagues adopted the technological development and intensification parameters and considered the wood harvesting and the long-term fallow area resulted from shifting cultivation, which were neglected in the reconstructions of the other two datasets. Therefore, the KK dataset is more suitable as one of the anthropogenic forcing fields for climate simulation over the past two millennia that is recently concerned by two projects, the National Basic Research Program and the Strategic and Special Frontier Project of Science and Technology of the Chinese Academy of Sciences.