Abstract: The process of flash flood sediment yield and transport is controlled by multiple factors such as slope erosion, river transport, sedimentation, and re transportation, and has significant nonlinear characteristics. Existing simulation methods often use a single "flow sediment transport" relationship or a unified conversion coefficient for overall fitting, which makes it difficult to distinguish the differences in sediment production sediment transport relationships under different flood levels, and is easily dominated by low flow samples, leading to instability in simulating high-level floods; However, fully mechanized models have high parameter requirements and are limited in their application in engineering. In response to the above issues, a two-stage model of "slope sediment yield river transport" was constructed using the Shouxi River Basin in Sichuan Province as the research object. In the first stage, the Modified Universal Soil Loss Equation (MUSLE) was used to calculate the daily sediment yield of each flood driven by rainfall; In the second stage, the graded comprehensive transfer coefficient (TSC) method is proposed to characterize the transport, sedimentation, and retransport processes in river channels. Floods are divided into four levels I to IV, and a quantitative conversion relationship between sediment production and outlet sediment transport is established through grading. The research results indicate that the MUSLE model can effectively characterize the slope sediment yield dynamics of single floods in mountainous watersheds on a daily scale; On this basis, the graded TSC method further converts daily sediment discharge into export sediment discharge, with high simulation accuracy during the calibration period (Nash efficiency coefficient of 0.940, goodness of fit R² of 0.942), and also maintains good accuracy during the validation period (Nash efficiency coefficient of 0.818, goodness of fit R² of 0.860). The simulated residuals are generally symmetrically distributed around the zero line, with no obvious systematic deviation. The absolute value of the residuals shows a reasonable amplification trend with the increase of measured sediment discharge, and the model maintains a robust characterization of the cross scale sediment transport process in the watershed. This method introduces flow classification to characterize the magnitude differences in the relationship between yield, transportation, and sediment. Without significantly increasing the demand for mechanism parameters, it improves the structural clarity, physical interpretability, and robustness of the model for simulating high-level floods. It can provide reference for early warning and forecasting of flash flood and sediment disasters in mountainous areas, as well as comprehensive watershed management.