Uncertainty Reduction in Fluvial Flood Re-analysis by Assimilating SAR-derived Flood Extent Maps
Abstract
The occurrence and intensity of flooding have increased in recent decades, especially in the context of climate changes. Relevant comprehension and assessment of flood hazards has thus become a crucial necessity. Flood simulation and forecast capability have been greatly improved thanks to advances in data assimilation (DA). Conventional DA methods combine in-situ gauge measurements with numerical models to correct the hydraulic states and reduce the uncertainties in the model parameters and inputs (e.g. friction, inflow discharge). However, they depend strongly on the quality and density of observations, thus requiring leveraging other data sources to improve the performance of flood simulation and forecast. In this work, the boundary conditions for the high-fidelity and local-scale hydrodynamics model Telemac 2D (T2D) are provided by the large-scale river network hydrologic model RAPID (Routing Application for Parallel Computation of Discharge), over the Garonne Marmandaise catchment. Major sources of uncertainties in this chained hydrology-hydraulic lie in the RAPID hydrographs and the T2D parameters (e.g. friction, topography), thus requiring a time-varying and sequential correction from DA analysis. The chained model allows for extended lead time forecasts that overcome the limits of forecast when using observed only gauge measurements. The present work focuses on the assimilation of 2D flood observations derived from remote-sensing images. A Random Forest algorithm is applied to derive flood extents from Sentinel-1 SAR or Sentinel-2 optical images. The resulting binary wet/dry maps are then expressed in terms of wet surface ratios (WSR) over selected subdomains of the floodplain. This ratio is assimilated jointly with in-situ water-level observations to improve the flow dynamics within the floodplain. An Ensemble Kalman Filter with a dual state-parameter analysis approach is implemented on top of the T2D model to control friction coefficients, inflow discharge corrective coefficients, and the hydraulic states within floodplain subdomains. The observation operator associated with the WSR observations, as well as the dual state-parameter sequential correction, was validated in the context of twin experiments allowing the evaluation of DA performance in a controlled environment.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.H46D..08N