Estimating Flood Discharges in Reservoir-Regulated River Basins by Integrating Synthetic SWOT Satellite Observations and Hydrologic Modeling: Fid-Bau Portal

Estimating Flood Discharges in Reservoir-Regulated River Basins by Integrating Synthetic SWOT Satellite Observations and Hydrologic Modeling

Yoon, Yeosang / Beighley, Edward / Lee, Hyongki / Pavelsky, Tamlin / Allen, George

Abstract

Lakes and reservoirs are widely used for water supply and flood control, especially during large storm events. In hydrologic modeling applications, accounting for the regulated behavior of reservoirs distributed throughout a river system poses a significant challenge, especially during flood events, when detailed reservoir operation rules and strategies are implemented. Building on this problem, this study addresses this question: Can we model reservoir water storage changes and outlet discharges based on satellite measurements of river water surface elevation and inundated areas especially during the flood event? A method is presented and evaluated using synthetic observations as a proxy for measurements from the forthcoming surface water and ocean topography (SWOT) satellite mission. The May 2010 flood event in the Cumberland River Basin is used as a case study. Based on synthetic SWOT observation, time series of water storage changes are generated and evaluated for eight reservoirs. As expected, although SWOT will provide relatively high temporal resolution measurements (i.e., three or four times per repeat cycle) compared with current point-based satellite altimeters, it provides only a 5% chance of direct observation of the 2-day flash flood event. To overcome this limitation, a new algorithm using the continuity equation and ensemble Kalman filter is presented to augment SWOT-estimated water storage changes for times between SWOT overpasses at eight reservoirs located throughout the watershed. The algorithm provides accurate storage changes with a 9% normalized RMS error (NRMSE). SWOT-estimated storage changes and reservoir routing are integrated into the Hillslope River Routing model to estimate inflows and outflows for the reservoirs. The average NRMSE for reservoir outflow is 28%, which is a 64% improvement compared with not including reservoir routing. At the watershed outlet, which integrates all eight reservoirs, the NRMSE and peak discharge error ( $E Q_{P}$ ) for discharge estimates are 9 and 28%, respectively. The approach, which is entirely based on remotely sensed data and is potentially applicable in global scale models, reduced NRMSE by 73% and $E Q_{P}$ by 85% compared with the simulated discharges, without accounting for reservoir routing.