The SWIFTS project aims to develop innovative methods for combining big data, derived from satellite Earth observation, and hydrodynamic simulations to improve flood inundation modelling at local to regional scales. The project focuses on urban and peri-urban areas and relies on the advanced exploitation of SAR, optical and topography data to characterise complex inundation flows. The main motivation is to simultaneously improve observed and simulated products via advanced machine learning and data assimilation methods in order to reduce related uncertainties. SWIFTS will therefore contribute to tackle three main scientific challenges: (i) to improve topography and model friction parameterization in hydrodynamic models via the advanced use of photogrammetry, interferometry and land-use classification on images mainly provided by the Pleiades and Sentinel satellite missions, (ii) to further develop SAR image classification algorithms, including machine learning and interferometry techniques, for flood extent mapping in urban and peri-urban areas as well as exclusion area mapping where the SAR data doesn’t enable floodwater detection, (iii) to further develop methods for assimilating flood extent maps into hydrodynamic model as front information, probabilistic and binary flood maps along with the exclusion layers. SWIFTS will rely on big data and high performance computing for machine learning, high resolution hydrodynamic modelling and ensemble-based data assimilation. To embrace both local and regional scales a traditional high-resolution hydrodynamic modelling approach will be complemented with a larger scale modelling approach including porosity concepts. To demonstrate and evaluate the developed approaches, the project will use as test cases a well-gauged French basin and a more poorly-gauged Cambodian basin where most part of the information will come from satellite Earth observation.