In order to study the multidimensional dynamic evolution of soil erosion under the influence of multi-factors and spatial variability of the subsurface, to realize high-precision simulation of water and sand erosion processes in the watershed, to comprehensively analyze the spatial and temporal distribution characteristics of water and sand migration, and to predict the occurrence of erosion risk. Taking the existence of serious hydraulic erosion in the Laixi watershed as an example, the basin was accompanied by a wide distribution of soil erosion with multiple points, long lines, and extensive coverage. The existing models almost exclusively considered slope and fluvial erosion processes at the watershed scale or a basic combination of hillslope and watershed scales. It led to high uncertainty in the physical simulation of the soil erosion process. This paper proposes a coupled soil erosion monitoring system with different scale models, which comprehensively reflects the spatial and temporal dynamic migration process of the watershed and sediment in the watershed from the subsurface to the fine channel on the slope. Then it migrates with the watershed to the river channel for deposition from the surface, line, and point scales, following a hierarchical structure of surface-slope-river. Based on the watershed application of the multi-model coupling system, the results showed that, firstly, a single model demonstrated good applicability and accuracy in the Laixi watershed. The RUSLE model predicted erosion patterns that closely matched actual observations. The Nash-Sutcliffe Efficiency (NSE) and the coefficient of determination (R²) obtained from parameter calibration of the SWAT model exceeded 0.6, indicating good robustness of the simulated results. Secondly, intercorrelations between the feature inversions of the different erosion models all reached highly significant levels of correlation. The R² between the WEPP and SWAT models was 0.96, between the RUSLE and WEPP models was 0.77, and between the RUSLE and SWAT models was 0.58, and the analyses indicated that the physical processes of erosion at the surface, slope, and channel scales were tightly coupled. Finally, the whole process risk evaluation of soil erosion with multi-model coupling was more comprehensive than a single model. It facilitated the study of the multidimensional dynamic evolution of soil erosion under the influence of multiple factors and spatial variability of the subsurface. This enabled high-precision simulation of water and sediment erosion processes in the watershed, comprehensive analysis of the spatiotemporal distribution characteristics of water and sediment transport, and prediction of the regular occurrence of erosion risk.