Evaluating the accuracy of Chinese pasture data in global historical land use datasets

Global historical land use datasets are widely used in global environmental change studies; however, uncertainties in the included pasture data have not been evaluated. In this study, using trend, relative difference ratio (RDR), and spatial comparisons, we evaluated the accuracy of China pasture data in the HYDE (versions 3.1 and 3.2), SAGE, and PJ datasets using historical document-based reconstructions, including the land use datasets for China covering 1935–1997 (hereafter Ge-dataset) and Northeast China covering the 1700s–2000 (hereafter Ye-dataset), and the satellite-based China’s Land-Use/cover Datasets (CLUDs) for 1980–2015. Four important results were obtained. (1) China pasture area in the HYDE, SAGE, and PJ datasets and grassland area in the Ge-dataset, Ye-dataset, and CLUDs show both disparate trends and large differences in absolute values. (2) Spatially, 50.9%, 52.8%, and 63.0% pasture/grassland grids in 2000 had RDRs greater than 60% between HYDE3.1, HYDE3.2, and SAGE datasets, and CLUDs, respectively. The percentage of grids with RDRs less than 20% were 24.9%, 26.7%, and 16.0%, respectively. (3) Based on HYDE3.2, the spatial distribution of pasture in Northeast China over the last 300 year has expanded, which is in contrast to the results from the Ye-dataset, which shows the spatial distribution of grassland shrinking because of human reclamation activities. (4) The large contrasts between the global datasets and Chinese native datasets are due to differences in pasture/grassland definitions, land use practices, and spatial reconstruction methods.