Spatial and temporal dynamics of livestock grazing intensity in the Selinco region: Towards sustainable grassland management

Grazing pressure in the Selinco region exhibits significant complexity and variability. To optimize sustainable grassland management in the region, understanding the spatial and temporal dynamics of grazing intensity is crucial. Previous studies have shown the substantial potential of machine learning algorithms in accurately modeling grazing intensity. Consequently, this study predicted grazing intensity for the period 2001-2020 on a 500 x 500 m grid, using variables such as climate, topography, vegetation, anthropogenic disturbances, water sources, and roads as inputs to the machine learning model, with livestock density at the township level in suitable grazing areas as the output variable. Subsequently, the spatial and temporal dynamics of grazing intensity from 2001 to 2020 were synthesized and analyzed. Of these, the Support Vector Machine Regression (SVMR) model exhibited the best fit, with a mean absolute error (MAE) of just 0.068 SU/km2 and a coefficient of determination (R2) of 0.98. The results indicate that grazing intensity shows a spatial pattern of being high in the southeast and low in the northwest, with significant regional differences. The average annual trend declined by -0.006/10a, indicating a continuous downward trend; however, a substantial increase is anticipated in the future, as suggested by a mean Hurst index value of 0.39. Additionally, the grazing pattern in the northern region, including the Qiangtang Nature Reserve, remains relatively stable. However, it is worth noting that the southeastern region exhibited a significant exacerbation of grazing patterns during 2015-2020, necessitating focused attention. These findings are crucial for making informed decisions regarding grassland-livestock balance management and can aid in optimizing resource allocation and ecological conservation strategies. Future research will utilize higher resolution geographic data to enhance the accuracy and efficiency of grazing intensity models.