High-accuracy DEM construction method based on ICESat-2 and multisource DEMs

This study adopts a research method combining multi-source data fusion and the random forest machine learning algorithm to develop a high-precision elevation prediction model based on the random forest algorithm. This model integrates terrain data such as ICESat-2, AW3D30 DEM, COP DEM and NASA DEM, improving the accuracy of the Digital Elevation Model (DEM). The accuracy of traditional DEM data is limited. However, with the advancement of remote sensing technology, terrain data from different sources can be combined for prediction (such as ICESat-2, COP DEM, AW3D30, NASA DEM, etc.), thereby improving the prediction accuracy of DEM. In this study, the models of traditional DEM models (AW3D30 DEM, COP DEM and NASA DEM) were compared and analyzed, and a random forest elevation prediction model was constructed using multi-source data to obtain high-precision DEM. The results show that RF30 DEM can exhibit low errors under different terrain types and slope conditions. Even in complex locations such as wetlands and grasslands, this study outperforms other models.