Abstract: Areas with different levels of susceptibility to landslides, rainfall warning thresholds vary widely, to improve the pertinence and accuracy of landslide monitoring and early warning. Based on field geological survey and landslide susceptibility condition analysis, this paper combines information quantity model and analytic hierarchy method to carry out susceptibility evaluation, and then analyzes the correlation between landslide occurrence probability and cumulative rainfall, and conducts landslide rainfall early warning threshold model research by zoning. The results showed that: Slope, elevation, geological structure, engineering rock group and water system are the main geological conditions of landslides in Longling County; The area of non-prone areas in Longling County was 14.33 km², the area of low prone area was 1053.87 km², the area of medium prone area was 1471.65 km², and the area of high prone area was 254.73 km²; The rainfall warning time was determined for a single day and the previous 3 days, and the rainfall warning threshold was refined by district; Compared with the rainfall warning threshold model applied to the landslide monitoring and early warning in Longling County, the number of early warning information decreased by 70, the accuracy of early warning was increased by 14.4%, and the effective early warning of landslide in Shijiazhai Formation of Huba Village, Zhen’an Town was realized. This paper provides a good reference method for determining the regional landslide rainfall warning threshold.