Abstract: This study aims to use local polynomial interpolation (LPI) to build DEM model of debris flow gully which is caused by earthquake, analyze the effect of interpolation parameters on interpolation precision, and extract the optimum interpolation parameters. The debris flow gully of Xiejiadian landslide mass in Yinchanggou, Pengzhou City was used as an example, mean error (ME), root mean square error(RMSE), maximum positive error and negative error (MPE and MNE) and the percentage of positive errors and negative errors (PPE and PNE) were used as accuracy evaluation indexes of the interpolated DEM, the accuracy of DEMs that interpolated was analyzed by different powers (P), kernel functions [K(x)], bandwidth (H) and sector types, and the optimum interpolation parameters were contrastively analyzed and extracted. The results showed that with the other parameters being defaulted, the interpolation accuracy was the highest when interpolated with P=2; the imitative accuracy was the best when interpolation K(x) was Exponential; the interpolation accuracy of four different sector types in LPI interpolation process had little difference; the optimum choice of H value was 2 in interpolation; the positive errors and negative errors of all experimental results with various interpolation parameters in LPI were approximately half-and-half distributed. The result of DEM model built by LPI met the accuracy of large scale DEM in Xiejiadian debris flow gully, the parameters of P, K(x) and H had great effect on interpolation accuracy, but sector type had little effect on interpolation result, the optimization of interpolation parameters could contribute to the DEM modeling in debris flow gully terrain.