The invention discloses a novel landslide risk evaluation method, and the method comprises the following steps: S1, carrying out the statistics of effective rainfall, extracting impervious surface data, and building a water seepage capability model according to the quantitative relation among the rainfall, the impervious surface data and an infiltration relation; S2, taking the gradient, the slopedirection, the water seepage capability index, the vegetation index and the road influence as parameters, and selecting related data of a plurality of points where landslide occurs and does not occurto train the model; and S3, carrying out model construction by using a Logistic regression function to obtain landslide risk divisions, and dividing the landslide risk levels into five risk levels, that is, an extremely low risk level, a low risk level, a medium risk level, a high risk level and an extremely high risk level. According to the method, the rainfall parameters are introduced into theinfiltration relation model, the corresponding landslide risk regionalization map is displayed in real time according to different rainfall, and a certain guiding effect is provided for landslide prevention. Compared with a traditional slope model and a change model for monitoring vertical displacement and horizontal displacement of a region, the landslide risk evaluation model is more suitable for large-scale real-time monitoring.