A high-resolution bathymetry database for global reservoirs using multi-source satellite imagery and altimetry


Despite the importance of precise bathymetry information for reservoirs, there is a dearth of a consistent reservoir bathymetry database which is locally practical and globally available. Here we developed an approach to generate bathymetry for global reservoirs with a 30 m resolution by combining multiple satellite altimetry datasets and Landsat based water classification datasets. The satellite altimetry data include ICESat/GLAS Global Land Surface Altimetry dataset and radar altimetry datasets reported by Global Reservoir and Lake Monitor (G-REALM) and Hydroweb, and the Landsat based water classification datasets comprise the Surface Water Occurrence (SWO) provided by Global Surface Water Explorer (GSWE) and the improved version of Monthly Water History (MWH) provided by Zhao and Gao (2018). First, Area-Elevation (A-E) relationships for the identified reservoirs were derived using two methods according to the altimetry data source. For the ICESat/GLAS data, the A-E relationship was established by pairing the elevation tracks with the SWO image. While for the G-REALM and Hydroweb datasets, A-E relationships were obtained by connecting the monthly elevation with the monthly water surface area provided by the improved MWH. Then, the A-E relationship was in turn applied to the SWO image to obtain the bathymetry values for the dynamic area. Finally, if the remotely sensed bathymetry cannot represent the full bathymetry in terms of the storage, an extrapolation method was adopted to help achieve the full bathymetry. The remotely sensed bathymetry results were primarily validated against (1) A-E and V-E (Volume-Elevation) relationships for fifteen reservoirs, with RMSE values of elevation from 0.10 m to 1.99 m and NRMSE values of storage from 0.50% to 5.07%, and (2) survey bathymetry values for four reservoirs, with R2 values from 0.82 to 0.99 and RMSE values from 0.13 m to 2.31 m. While for the projected bathymetry, it has relatively large uncertainties and errors from the validation results of Lake Mead against A-E and V-E relationships, and the survey bathymetry. The resultant database includes 180 global reservoirs (with a mean R2 of A-E relationship 0.86), representing a total volume of 3030.69 km3 (48.91% of the total global reservoir capacity), which will contribute to various studies and applications such as global hydrological models and water resource managements.

Remote Sensing of Environment (to be submitted)