Predictive Model of Rainfall-Induced Landslides in High-Density Urban Areas of the South Primorsky Region (Russia)

Vladivostok city and its surrounding areas have previously experienced rainfall-induced landslides, which have caused significant casualties and damage in high-density urban areas. As a result of anthropogenic factors, steep slopes in some areas reach 90°, which significantly affects the slope stability. The authors collected all available historical data about landslide incidents in the study area. A predictive model was derived using logistic regression and data on antecedent rainfall, cumulative precipitation, and daily rainfall intensity. The resulting model has relatively low precision and recall, which may reflect the lack of slope material parameters. Nonetheless, the balanced accuracy of 78% allows rainfall to be considered the most important causative factor of slope instability. The main advantage of the predictive model lies in its simplified mathematical expression and input rainfall data set based on measurements from one station with 24-h granularity. These results show promise for the further implementation of the model for the purpose of early warning.