Stochastic precipitation generator over the whole of France

Spatio-temporal models -

Supervisors: Denis ALLARD (INRAE) - Lionel BENOIT (INRAE)

To perform simulations of precipitation fields over large domains, as is the case here with the simulation of precipitation fields at the national scale, non-stationary approaches that are efficient enough to deal with large grids must be considered. The classical approach consists in dividing the simulation domain into sub-domains of modest sizes considered as stationary. However, the segmentation into subdomains, the links between subdomains, as well as the simulation of precipitation in areas with strong gradients (typically foothill regions) constitute methodological obstacles that are difficult to overcome by confining oneself to existing simulation tools. To overcome these difficulties, the goal of this project is to develop a stochastic precipitation generator taking advantage of the recent development of sufficiently flexible mathematical tools to directly simulate highly non-stationary random fields of very large size. The favored mathematical tool will be the SPDE approach and its spatio-temporal extensions (Lindgren et al. 2011, Pereira et al. 2022, Clarotto et al. 2023). The advances to be made during this project consist in adapting these methods to the case of precipitation (non-Gaussian variable, very asymmetrical with a large quantity of zero values), both for the simulation and for the estimation of the model parameters. stochastic.