Changes in land use types in typical areas of Yinchuan City from 1990 to 2020 were explored, and the evolution process of landscape patterns represented by impermeable surfaces was analyzed. Based on the coupling model of hydrology and hydrodynamics and the random forest model, landscape pattern indexes that significantly affect the maximum submergence depth and area of waterlogging under different return period design rainfall scenarios were identified, and the reasons were analyzed from the perspective of runoff generation mechanism. The results indicate that the impervious surface landscape pattern indexes that significantly affect waterlogging in typical areas of Yinchuan City are patch density, fractal dimension, and Euclidean nearest neighbor distance. In the design rainfall scenario with shorter return periods (520a), patch density has the highest impact on surface inundation, while in the design rainfall scenario with longer return periods (50200a), the effects of fractal dimension and Euclidean nearest neighbor distance are more significant.