Lake Level Reconstructed From DEM-Based Virtual Station: Comparison of Multisource DEMs With Laser Altimetry and UAV-LiDAR Measurements

Although the traditional water-level observation has been improved by wide application of satellite altimetry, the acquisition of fully-covered, long-term water level is impeded by the inadequacies of radar/laser altimetry sensors. Alternatively, the water level of lakes in various sizes can be obtained by combining multitemporal lake shorelines and superimposed topographic information [e.g., digital elevation model (DEM)]. However, the quantification and reduction approach of the uncertainty of water levels reconstructed from the topographic data remains largely unexplored. Therefore, this study aims to develop an improved DEM-based method for reconstructing long-term water levels for ungauged lakes. Before this, we first assessed the characteristics of vertical height uncertainties of DEMs varying with topographic slopes. Assessment results for DEM evaluated based on ICESat-2 show that higher spatial resolution DEMs (Shuttle Radar Topography Mission (SRTM) DEM and AW3D DEM) achieve a higher vertical accuracy with the mean absolute error (MAE) of 3.67 and 3.46 m, respectively, while the mean error will decrease to 1.8 m upon the slope below 1°. Our study confirms that terrain relief exerts strong influences on the DEM data quality. As the DEMs show higher error in steeper terrain, the proposed method first selects the lake bank sites at gentle slopes as the virtual stations and then reconstructs water-level series by superimposing lake shorelines over DEMs at the virtual stations. Results revealed that the reconstructed lake levels show strong correlations with the Hydroweb water level from altimetry data with an R² close to 0.90. Furthermore, we found that unmanned aerial vehicle (UAV) Light Detection and Ranging (LiDAR)-based DEM is expected to have much better performance in the virtual station method.