Abstract: 【Objectives】The physical-based stability models, such as traditional Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability Model (TRIGRS), have been widely used for predicting the regional rainfall-induced accumulation landslide hazard. However, the linearization processing of the Richards equation leads to deviations in the solution of rainfall infiltration from the actual situation, thereby resulting in low accuracy of regional landslide hazard prediction. 【Methods】 By combining the high-precision advantage of the Hydrus software in solving the Richards equation, a quantitative rainfall-induced accumulation landslide hazard prediction model has been proposed based on the Regional Hydrus Model (RHM). First, the precise pore water pressure is calculated by solving the Richards equation using the Hydrus software. Then the Factor of Safety of each grid within the region is calculated by coupling with the infinite slope model. Finally, the regional landslide hazard is classified to five levels (very high, high, moderate, low, very low), and a comparative analysis is carried out with the traditional TRIGRS model using Receiver Operating Characteristic (ROC) accuracy, actual landslide distribution, and overlap rate of terrain interval distribution. The landslide event in Mibei Village, Guangdong Province from June 10th to 11th, 2019 is selected as an example. 【Results】The results indicate that: (1) the ROC accuracy of the RHM and TRIGRS models is 0.86 and 0.80, respectively, the proportion of actual landslides in the very high and high landslide hazard levels is 38.6% and 49.4%, respectively, indicating that the prediction performance of the RHM method is superior to that of the TRIGRS model. (2) The distribution overlap rate of RHM in key terrain intervals such as elevation, slope and accumulation layer thickness is significantly higher than that of the TRIGRS model, enabling it to more accurately identify the correlation between terrain factors and landslides. (3) The landslides in the study area are controlled by terrain and driven by heavy rainfall, mainly developing in the steep slope zone of valleys with elevation of 300~400 m and slope of 20~35°. The very high and high landslide hazard areas are distributed in strip-like patterns along the steep slopes of the valleys, while the very low and low landslide hazard areas are located on gentle slopes and terraces. 【Conclusions】By employing high-precision numerical solutions for the Richards equation, the RHM overcomes the limitations inherent in the linearized analytical solutions of TRIGRS model. It provides an effective tool for high-precision rainfall-induced accumulation landslides.