Sequential assimilation of remotely sensed water stages in flood inundation models

The hydrologically and hydraulically relevant variables (e.g. soil moisture, flood extent, water stage) and basin characteristics (e.g. topography, surface roughness) that can be obtained from radar remote sensing are useful for implementing, calibrating and evaluating both rainfall-runoff and flood inundation models. To achieve a time continuity that is crucial in monitoring applications but cannot be obtained by the sole use of remote sensing observations, the information extracted from the discrete Earth observation data can be used both as parametric input and as time-varying state and flux data in coupled hydrological and hydraulic models. This paper focuses on the sequential assimilation of remotely sensed water stages into a 1-D flood inundation model (HEC-RAS). Through the integration of radar imagery of flood events with high precision digital elevation models, inundation depths can be extracted from remote sensing observations. The methodology consists of adjusting simulated water surface lines by comparing modelled water stages with those that are derived from remote sensing observations, thereby increasing the overall accuracy and reliability of flood predictions at subsequent time steps. The potential of the proposed methodology is illustrated by a well-documented flood event of the Alzette River (Grand-Duchy of Luxembourg).