SCIENTIA GEOGRAPHICA SINICA ›› 2023, Vol. 43 ›› Issue (3): 398-410.doi: 10.13249/j.cnki.sgs.2023.03.003

Previous Articles     Next Articles

Mapping the multi-temporal grazing intensity on the Qinghai-Tibet Plateau using geographically weighted random forest

Li Lanhui1,2(), Huang Congcong1, Zhang Yili2(), Liu Linshan2, Wang Zhaofeng2, Zhang Haiyan2, Ding Mingjun3, Zhang Huamin3   

  1. 1. School of Computer and Information Engineering, Xiamen University of Technology, Xiamen 361024, Fujian, China
    2. Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
    3. School of Geography and Environment, Jiangxi Normal University, Nanchang 330022, Jiangxi, China
  • Received:2022-06-20 Revised:2022-08-20 Online:2023-03-30 Published:2023-03-20
  • Supported by:
    Strategic Priority Research Program of Chinese Academy of Sciences(XDA20040201);the Second Tibetan Plateau Scientific Expedition and Research(2019QZKK0603);National Natural Science Foundation of China(42101099)


Accurately quantifying the spatiotemporal pattern of grazing intensity on the Qinghai-Tibet Plateau (QTP) is crucial to improving our understanding of the driving mechanism of alpine grassland change, and is of great significance for maintaining regional ecological security and promoting sustainable development policies. Based on the data of livestock inventory at the end of the year and environmental covariates (e.g. population density in pastoral areas, growing season NDVI, annual precipitation, annual mean temperature, and settlements), the gridded grazing intensities on the QTP in 2000, 2010, and 2020 were simulated by geographically weighted random forest (GRF), and the regional differences in the interpretability of environmental variables were then analyzed. The results showed that grazing intensity maps predicted by the GRF could mirror the spatial distribution of grazing intensity on the QTP, compared with the classic random forest model, the R2 was higher, and both the mean absolute value error (MAE) and root mean square error (RMSE) were lower. The grazing intensity was generally higher in the southeast and lower in the northwest of the QTP. The areas with grazing intensity of less than 25 sheep units/km2 in the northwest part accounted for about half of the QTP. Compared with 2000 and 2010, the grazing intensity of the QTP in 2020 showed a trend of overall decrease but local increase. For example, compared with 2010, the areas of grazing intensity decreased by higher than 1 sheep unit/km2 in 2020 accounting for 61.69% of the pastoral area. The population density in pastoral areas was the most important factor explaining the spatial heterogeneity of grazing intensity, and its relative importance was higher in the western QTP and lower in the eastern QTP. On the contrary, the relative importance of both precipitation and growing season NDVI was higher in the northwest part and lower in the southeast part of QTP. Our results provide scientific references for sustainable grassland management and ecological safety barrier construction on the QTP.

Key words: grazing intensity, geographically weighted random forest, population density in pastoral areas, the Qinghai-Tibetan Plateau

CLC Number: 

  • S17