Abstract: A major challenge in flood routing research is to calculate the flood routing of rivers accurately and efficiently in catchments without high-precision topographic data. In this paper, the Digital Elevation Model(DEM) of river channel from Yebatan to Batang on Jinsha River was reconstructed based on ASTER GDEM V2 terrain data with a spatial resolution of 30m. The 11· 03 Jinsha River dam break event in 2018 was simulated by using a GPU accelerated surface water flow and transport model(GAST), for which the simulation results were compared with the measured flood process. The results showed that: (1) Different spatial interpolation methods have a large effect on the description of river channel terrain characteristics, and the river channel terrain generated by the spline interpolation and the linear interpolation based on the latitude gradient of the river trend(trend smoothing method) can better express the river landform. (2) The simulation accuracy of flood routing in which topography was reconstructed by trend smoothing method increases as the section distance increases, and the effect of simulation using only the inlet and outlet section elevation is the best among different section distances; (3) The flood simulation accuracy increases with the increase of river terrain digging depth, in which river terrain was simulated by trend smoothing method, and NSE and R2 can reach to 0.937 and 0.976 separately when the digging depth is 50 m.