Abstract: Loess is widely distributed in the Northwest Plateau of China. One-third of the landslides in China occur in the loess area. Shallow loess landslides are especially widespread and frequent geological disasters, causing serious casualties and huge property damage. Under rainfall and loading, loess is prone to structural collapse and strength reduction. Therefore, shallow loess landslides distribute widely and occur frequently. Usually, rainfall and earthquakes are the frequent and active triggers for loess landslides. In recent years, a large number of loess landslides have been induced by the coupling of rainfall and earthquakes on the Loess Plateau. Although the coupling effect of earthquake and rainfall will seriously aggravate the instability probability and disaster risk of shallow loess landslides, there is still a lack of quantitative disaster evaluation research on such landslide events. This study chose the shallow loess landslide as the research object in the Dashagou catchment of Lanzhou city. The rainfall penetration model was integrated into a three-dimensional deterministic model of the loess slope, and the stability of the shallow loess landslide was evaluated in the study area with different rainfall and seismic coupling effects. The confusion matrix and the receiver operating characteristic (ROC) curve were used to evaluate the results of the stability evaluation prediction. Results of this study reveal that the integration of a three-dimensional deterministic model of rainfall infiltration and earthquake effects has a good impact on the stability evaluation of shallow loess landslides at the watershed scale. Moreover, this model can be used as a tool for the assessment and early warning of rainfall and earthquake-induced loess landslides. The employment of the three-dimensional deterministic model considering a complicated slope and rainfall situation has great significance in the acquisition of results that are more accordant with the actual situation. It is of great reference value to strengthen the spatiotemporal disaster assessment and prediction of loess landslide disasters under different scales of extreme events.