Merits of Data Assimilation on Improving Flood Forecasting - A case study of Ohio Cannelton-Newburgh
- 1CECI UMR5318 CNRS-CERFACS, Toulouse, France (ricci@cerfacs.fr)
- 2Luxembourg Institute of Science and Technology (LIST), Esch-sur-Alzette, Luxembourg
- 3Centre National d’Etudes Spatiales (CNES), Toulouse, France
- 4Collecte Localisation Satellites (CLS), Toulouse, France
- 5Ohio River Forecast Center, NOAA NWS, Wilmington, OH, USA
- 6Ohio State University, Columbus, OH, USA
Flooding is represented with a 2D hydrodynamic model over a reach of the Ohio river between Cannelton and Newburgh locks and dams. The geometry of the river and floodplain was provided by the National Oceanic and Atmospheric Administration (NOAA), based on U.S. Army Corps of Engineers (USACE) survey channel data merged with United States Geological Survey (USGS) LiDAR in the overbank regions. The description of hydraulic structures from USACE and in-situ water depth measurements from USGS stations were also used. Working from the 1D HEC-RAS model from Ohio University that covers a much larger area, the friction for our 2D local model was set uniformly over the floodplain. These values were further calibrated to 45 m1/3s-1 over the river bed and 17 m1/3s-1 with in-situ water depth measurements from USGS stations at Cannelton, Owensboro, and Newburgh over high flows periods in 2022 and 2023.
The performance of the model was first assessed for the significant flooding event in February 2018, with RMSEs of the order of a few tens of centimeters. Remote-sensing (RS) products provided by satellite missions such as Sentinel-1 SAR, Sentinel-2 optical and Landsat-8 optical imagery undoubtedly offer opportunities to improve our ability to monitor and forecast flooding. For this study, the performance of the TELEMAC-2D (www.opentelemac.org) Ohio model was improved with the joint assimilation of in-situ and remote-sensing data within an EnKF framework that accommodates 2D RS-derived observations alongside with in-situ water level time-series. The RS-derived flood extent maps are expressed in terms of wet surface ratios (WSR) in selected subdomains of the floodplain. The assimilation of in-situ data reduces the RMSE to tenths of a centimeter. Ongoing work on the assimilation of WSR aims at improving the dynamic of the floodplain. This 2D Ohio model will serve as a demonstrative test case for the FloodDAM-DT (https://www.spaceclimateobservatory.org/flooddam-dt) prototype dedicated to flood detection, mapping, prediction and risk assessment.
How to cite: Ricci, S., Nguyen, T. H., Piacentini, A., Rodriguez-Suquet, R., Pena-Luque, S., Bonassies, Q., Fatras, C., Astifan, B., Davis, R., Durand, M., and Coss, S.: Merits of Data Assimilation on Improving Flood Forecasting - A case study of Ohio Cannelton-Newburgh, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-17653, https://doi.org/10.5194/egusphere-egu24-17653, 2024.