Abstract: Taking Liuyang river basin as the research area, the hydrological mutation test of runoff sequence data from 1961 to 2015 was carried out. The SWAT model driven by China Atmospheric Assimilation Driven Data Set （CMADS） was used to simulate monthly runoff from 2009 to 2015. The SUFI-2 algorithm was used to analyze parameter sensitivity and uncertainty. The results show that the runoff series of Liuyang river basin changed abruptly in 1987, which was not consistent. SWAT model driven by CMADS can better simulate the runoff process of Liuyang river basin under the condition of inconsistent hydrological sequence, and the five indicators of determination coefficient, Nash coefficient, deviation percentage, root mean square error and standard deviation were within the error range. Among the 11 parameters related to runoff, SCS runoff curve coefficient, groundwater time delay, soil available water, base flow α-factor and soil saturated permeability coefficient are the most sensitive. Validation period and rate of periodic P-factor were greater than 0.5, R-factor were less than 1.0, simulation uncertainty was small. The research results can provide scientific reference for water resources management and regulation and distribution in Liuyang river basin, and provide strong support for the coupling between subsequent models.

Key words: CMADS, SWAT model, runoff simulation, Liuyang river basin