Flood hydrograph estimation using an adjoint shallow-water model

• Data assimilation was used to estimate a Tone River flood discharge hydrograph.
• We developed an adjoint shallow-water model for hydrograph estimation.
• Assimilated data comprise a time series of water levels measured along the river.
• We verified the hydrograph by comparison with data from aerial photographic analysis.
• Results show that our method reasonably corrected the observed discharge.

This paper describes the estimation of the flood discharge hydrograph for a compound open channel floodplain using data assimilation. The hydrograph was produced iteratively using an adjoint shallow-water model with time-series data of observed water levels. Data assimilation was applied to a flood event that occurred in mid-September 1998 along a 20 km stretch of the lower Tone River in Japan. Both the estimated hydrograph and the flow field simulated using the hydrograph were verified by comparison with the results of aerial photographic analysis. Results show that the hydrograph based on the stage–discharge was overestimated, but it was reasonably corrected by the data assimilation technique used for this study. In addition, the resulting relation during the flood event exhibited a hysteresis loop characteristic, which is typical of compound open channel flows during flooding.