Abstract: Persistent rainfall over a region normally follows the occurrence of landslides.Limited by fixed revisit period, swatch width and possible cloud cover in the region at the moment of imaging, it is hard to obtain post-disaster terrain information with traditional optical stereo satellites through along-track measurement.In view of this, a novel landslide volume estimation procedure based on cross-track DSM extraction technique was proposed.To begin with, high-quality satellite images that meet the production requirements were screened and recombined.A secondary elevation correction method was used to further reduce the overall error before and after landslides, thus realizing the extraction of landslide volumes based on conventional optical satellites in complex mountainous environments without relying on in-situ control points.Selecting a case study at Shaziba landslide, Enshi, Hubei Province, which occurred in July 2020,a DSM was generated using cross-track stereo image pairs composed of two GF-2 satellite images from different periods after the landslide, which was combined with DSM generated by the ZY3-02 satellite before the landslide to extract the landslide volume.The results suggest that:(1) After the secondary accuracy correction of the external elevation data, the DSMs generated by the cross-track and along-track stereo measurement techniques had similar accuracies, and both met the mountainous area mapping requirements of 1∶10 000 DEM;(2) The extracted landslide volume was 2 962 800 m3 with highly accuracy, and the landslide's structural characteristics could be identified from the volume changes.As such, this study provides a constructive reference for the analysis of landslide structural characteristics based on remote sensing and conventional mountain terrain mapping.