Evaluation of global historical land use scenarios based on regional datasets on the Qinghai–Tibet Area

Global historical land use scenarios are widely used to model human-induced climate change from the regional to global scales. It is necessary to conduct regional scale assessments of these global scenarios, identifying their uncertainties and pointing out directions for improvement. Based on the regional reconstruction Li-dataset, remotely sensed dataset, and grazing intensity dataset, the uncertainties of land use area and geographical distribution in HYDE3.1, HYDE3.2, and SAGE (a global land dataset from the Center for Sustainability and the Global Environment) scenarios for the Qinghai–Tibet Area (QTA) are evaluated. The comparisons show that the cropland areas on the QTA in HYDE3.2 for 1900–2000 are close to those of the Li-dataset, whereas HYDE3.1 underestimated and SAGE overestimated the cropland areas significantly. Spatially, HYDE3.1, HYDE3.2, and SAGE have large uncertainties, which cannot reflect the distribution of cropland on the QTA and its changes for 1900–2000 well, and too much cropland is allocated to southeastern Tibet. HYDE3.1 and HYDE3.2 overestimated the pasture area and its distribution on the QTA significantly. The distribution of pasture in SAGE showed overall an agreement with the spatial pattern for grazing intensity, but changes in grazing intensity for 2000–2010 was not reflected in SAGE. The FAO pasture definition and estimates and the method of using population as a proxy for pasture area are not appropriate for the QTA. Methodology which uses the pasture inventory data to calibrate satellite-based grassland maps to obtain the current pasture maps may also not be appropriate because of the lacking differentiation between natural and anthropogenic grasslands in remotely sensed data. More regional level land use estimates with concise definitions, define the land use more clearly, and stratification reconstruction based on differences in agro-climatic conditions and resource endowments may be used to improve global maps.