Abstract:

Grassland Livestock Intensity(GLI) refers to the number of various types of livestock raised per unit area, and is an important indicator for evaluating the ecological status and management of grasslands. Excessive GLI may lead to a series of ecological and environmental problems, such as grassland degradation, soil erosion and biodiversity reduction, so research on estimating the GLI and guiding reasonable grassland use can maintain the sustainable development of grassland ecosystems. The traditional way of estimating GLI is time-consuming and labour-intensive, and it is difficult to directly estimate the effect of grazing on the GLI. In this study, we used the grazing quantity to indicate the GLI as the research object, and used a Bayesian network model to estimate the GLI within a kilometre grid in the Basin of Kherlen River by considering the causal relationship between environmental influences, such as soil properties, vegetation, topography, river network density and road density, and the GLI of the 113 bags in the Basin of Kherlen River in 2021. In 2021, five types of livestock, including horses, camels, cows, goats, and sheep, were grazed in the Basin of Kherlen River. After conversion, a total of 10821500 sheep were distributed among 113 bags, showing significant spatial heterogeneity. The study showed that topographic elevation (DEM), river network density, vegetation index (NDVI) and fine-grained soil accumulation density directly affected the GLI, with NDVI having the most significant effect. The prediction results of GLI showed that the maximum number of sheep could be up to 53,480 and the minimum was 0, with an average of 115 sheep per square kilometre. The model accomplished accurate prediction of GLI with an accuracy of 84% for the training data and 87% for the test data in cross-validation.

Key words: the Basin of Kherlen River, Bayesian network, livestock intensity