The paper tackles an interesting and challenging problem of flood description within the context of uncertainties.
Flood risk assessment requires the prediction of the water depth which is estimated here using 2D shallow water equation where the inputs of the model are the topographical description of the area considered and the hydraulic conditions.
The numerical method for the resolution of the shallow water equation is first detailed and each run reveals time demanding for the presented application.
In the present work, the hydraulic conditions are those of a previous flood event while the topographical description is defined using high resolution classified topographic data that are available for the area.
The modeling of uncertainties for the application represents an essential part of the study.
Indeed, the definition of the topography (terrain and structures), used as an entry in the shallow water equation,
depends on the resolution of the topographic data, the level of elevation information considered for modeling the surface and also on the resolution of the mesh used for the numerical computation of the water depth.
These three parameters are modeled as continuous or categorical ordered random variables.
The paper aims at providing a general scheme for uncertainty analysis using on the one hand high resolution classified topographic data and on the other hand deterministic and statistical available codes, in the view of sensitivity analysis in order to reduce the model. It focuses therefore on a first proposal for the modeling of input uncertainties and on the construction of the design of experiments of a challenging application.
Mathilde Chevreuil (Universite de Nantes, GeM)