Methodology

• The work packages, correspond to the three challenges identified for the project (C1-3, section a), with the objective to design, implement, improve and evaluate key features of integrated forecasting chains meeting operational needs.
• WP1 - Coupling models and scales to reach users needs. Integrated tools with multi-scale forecasting capabilities, from small scales (< 1km2) to catchment scales (env. 2500km2)will be obtained by coupling short range precipitation forecasts, spatially distributed scalable and regionalized hydrological and high resolution hydraulic models including hydrological processes;
• WP2 - Reducing uncertainties with new data and methods. Improved accuracy of high resolution chains, computationally efficient over large domains, will be reached by performing innovative methods: offline-online strategies, flow model reduction combined with statistical, DA and machine learning.
• WP3 - Better involving end-users. Impact modeling and expert specifications-evaluations for the forecasting chains will be achieved thanks to the expertise of project end-users at multiple scales and related to the chosen cases studies, rich datasets, damage-risk models.

Figure 1. (Left) Overall objectives of the MUFFINS project and new insights. (Right) Main study catchments in the French Mediterranean region (magma), Aude catchment (green), fall-back/additional case.

An end users group will include the national SCHAPI and regional SPC MedEst flood forecasting center, the national reinsurance company CCR, departmental rescue service (SDIS) and local basin organisms. They will help defining specifications for forecasting chains design and testing on "their" catchments but also help to enrich datasets and access existing hydraulic models from engineering studies (ex. local land use planning studies).

The case studies will be located in the French Mediterranean region, and especially in the Var, Bouches du Rhône and Alpes-Maritime departments (Fig 1) which corresponds to the territory of SPC MedEst. Indeed, several recent flood events offer a particularly favorable context in terms of data availability (including damage data) and applicability of the multi-scale approaches. These events include the June 2010 and October-November 2018 floods in the Argens watershed, the January 2014 floods in the Gapeau watersheds, the October 2015 and October 2020 floods in the Alpes-Maritimes region. In order to foster fruitful collaborations between project partners and propose high quality research and demonstrators, efforts of the MUFFINS project will be concentrated on (but not limited to) the Gapeau and Argens, and next Huveaune watersheds representing very interesting cases with multi-scale flood issues.
The list of events, existing dataset (in particular from the PICS project - https://pics.ifsttar.fr/) and modeling goals will be refined at the beginning of the project in consultation with end-users (WP3).

Several models are used in the project, including distributed hydrological models (SMASH, MARINE, CINECAR), 2D hydraulic models (DassFlow, Floodos, Telemac) as well as data assimilation algorithms, machine learning and processing tools.