Inversion of Time series Elevation Changes in Open Pit Mining Areas Based on Adjacent Orbital Sentinel-2 Images

The ortho-rectified Sentinel-2 images have systematic offset errors in areas with large surface elevation changes in the cases of open pit mining, land collapse and glacier movement, which can lead to the misinterpretation of geological features classification from one scene image, or bias of deformation results from multiple images. Such systematic offset errors are mainly caused by the outdated DEM during the orthorectification processing. In turn, the systematic offsets can be applied to estimate the surface elevation changes. We firstly deduce the pixel offset of ortho-rectified Sentinel-2 image caused by outdated DEM based on imaging geometry, new DEM and projection equations. The sum of two offsets by two Sentinel-2 images from adjacent orbits are verified by the results calculated directly by Pixel Offset Tracking (POT) algorithm to the same images. The correlation coefficient R2 between two offsets amounts to 0.95, and the standard deviation of their difference is of 4.04 m. Secondly, the offsets by POT algorithm between two adjacent orbits are used to derive the time series elevation changes. Finally, the surface elevation changes are validated using the ICESat-2 point cloud and the SRTM DEM. Results show that the standard deviation of height change difference is around 4.13 m. This method is able to reveal the historical surface elevation changes related to the geophysical or anthropogenic process.