Numerical tools for Data-driven approaches


Alain Crave
Thomas Corpetti


Today, many acquisition systems offer extremely rich observations related to water resources (remote sensing, radar, sodar, lidar, …) with high quality of temporal and spatial precision. Associated measurements can either be direct (e.g. in situ velocity probes) or indirect (backscattered electromagnetic or acoustic signal, images, …).  In any cases, the amount of data to process opens new researches on numerical and modeling tools to deal with a better:

  • Representation of data: denoising, upscaling, …
  • Interpretation of data: how to accurately synthetize multivariate data and extract relevant characteristics?
  • Inversion of models: how to retrieve physical quantities from surrogate and noisy data?
  • Simulation of data: building synthetic databases, lacking data, …

In this session we solicit contributions in domains related to computational methods for physical data processing dealing with any field of the water cycle, which includes among others: data assimilation, model inversion, multivariate signal processing, non-linear machine learning approaches (kernel methods, deep learning).

Comments are closed