Statistical precipitation models modeling droughts and intense precipitation

The general objective in this thesis is to develop a stochastic approach (also called stochastic generator) to model the distribution of precipitation (and possibly temperature), ranging from intense droughts to heavy rain episodes, in the current and future climate at on a national scale (SAFRAN mesh, 8 km x 8 km) and on a daily time scale. A key methodological aspect is to develop simulation techniques based on the latest advances in univariate and bivariate space-time modeling, multivariate extreme value theory and expert aggregation (in the sense that a given climate model can be considered a special expert). This will be done through modeling of spatiotemporal stochastic generators, and downscaling approaches between observations (ERA 5 type) and regional models and CMIP 6 outputs. The issues of changes in spatial scales (downscaling), spatial non-stationarity and the impact of anthropogenic forcing as well as the choice of explanatory variables and/or relevant forcing will be at the heart of the applications.