Reconstruction of hydropower reservoir inflows using SWOT imaging radar altimeter

Abstract:

Radar imaging altimeters, like the future SWOT mission, have the potential to generate water elevations in large inland water bodies such as hydropower reservoirs. In this study, a series of synthetic SWOT water elevation images of the Eastmain-1 reservoir, a large water body with surface area of 608 km² located in Northern Quebec, were produced using the Centre National d’Études Spatiales (CNES) Large Scale hydrology simulator (SWOT-LS). The simulation period spanned from May 1^st to September 30^th 2015, during which the reservoir was filled with snowmelt runoff. To that end, a transient 3-D hydrodynamic model of the reservoir was built from which the water elevation outputs were forced into SWOT-LS to generate 29 images corresponding to 4 distinct SWOT orbits. The resulting level 2 water elevation pixel clouds images were upscaled to a 1-km resolution to increase altimetric accuracy while still being commensurate with reservoir surface area for storage calculations. Since each created image only covers a portion of the reservoir surface area, from 12 to 96% depending on selected orbit, a linear model was developed to extrapolate SWOT surface water elevations to the area of the reservoir not imaged by the satellite. The synthetic images were compared with the ‘truth’ from the 3-D model and were used to reconstruct inflows to the reservoir during that period. The reservoir inflow time series were compared to inflows reconstructed using water elevation measured with a level meter.

Results indicate that the lowest altimetric accuracy of the SWOT derived reservoir averaged water elevation never exceeded 4 cm mean square error, which is well below the 10-cm target for 1-km resolution products, and mean RMSE was 1.5 cm. The storage volume above the minimal reservoir exploitation level obtained from the SWOT images closely matched the volume computed from the storage-elevation curve of the reservoir with level meter measurements. The difference in volume never exceeded 15 hm³, which represents 0.4% of the total storage volume above exploitation level. Finally, the reconstructed inflows to the reservoir generated by using SWOT water level observations closely followed inflows retrieved using level meter observations, suggesting that the future SWOT mission would prove useful for estimating inflows to large reservoirs.