CFMDG: a Coastal Flood Modelling Dataset in Gâvres (France) to support risk prevention and metamodels development

Along most of the coastal areas, detailed coastal flood observations (e.g. inland water depths) are scarce, and when they are available, this for a limited number of events. Given recent scientific advances, coastal flooding events can be properly modelled, even in complex environments and under the action of wave overtopping, and thus provide detailed information. However, such models are computationally expensive, which prevents their use for instance for forecasting and warning. At the same time, metamodelling techniques have been explored for coastal hydrodynamics and have shown promising results. Metamodels are functions that aim to reproduce the behaviour of a “true” model (e.g., a numerical hydrodynamic model) for given input variables (for instance, offshore conditions). Within the RISCOPE research project (https://perso.math.univ-toulouse.fr/riscope/) aiming at exploring to which extent such metamodelling techniques may allow to forecast coastal floods with a good accuracy, a simulated flood database has been built for the site of Gâvres (France), characterised by a significant effect of wave overtopping processes.

The CFMDG dataset compiles a set of post-processed coastal flood simulations on the site of Gâvres. The dataset includes 250 scenarios. Each scenarios is defined by 6h time series centered on high tide, with one time series per forcing variables. The forcing variables (called X) are: local relative mean sea-level, tide, atmospheric storm surge, the offshore wave characteristics and the offshore wind. These scenarios combine past real (flood and no flood) events in the 1900-2021 time span with extreme statistics based events, and some complementary fictive events. The post-processed outputs (called Y) includes, for each scenario, the maximal flooded area (m²) and the maximal water depth (m) in each of the 64 618 inland model grid points.

The modelling chain that allowed building this dataset relies on the joint use of a spectral wave model (WW3) to propagate the waves to the coast, and a non-hydrostatic wave-flow model (SWASH) to simulate the nearshore hydrodynamics and the flooding. The spatial and temporal resolution of the SWASH configuration validated on the Gâvres site are respectively 3 m and more than 10Hz. All the results are obtained for a Digital Elevation Model corresponding to the 2018 configuration of the site.   

Such type of dataset is of use for local knowledge, risk prevention, metamodel testing/training, and local coastal flood forecast. 

Part of this dataset has already been used in (Idier et al., 2021; López-Lopera et al., 2021; Betancourt et al., 2022), to develop metamodels and set up a coastal flood forecast and early warning prototype.

We hope and expect that making this dataset accessible will trigger further developments/investigations for improving risk knowledge on the considered site as well as methodological developments on machine-learning/metamodel-based techniques to support flood forecast.

The table below summarizes the variables contained in the dataset, for each scenario.