In this framework, our efforts have been directed in three main axes: i) studying the relation between the occurrence of extreme events and their large-scale dynamical or thermodynamic drivers, ii) providing robust estimates of their return times in a changing climate iii) improving the representation of observed extreme events in climate simulations via statistical downscaling.

Extreme Events and large-scale circulation drivers

By adapting the theory of recurrences in dynamical systems to atmospheric fields, we have derived metrics that capture the predictability and the persistence of daily atmospheric fields. We have found that blocking patterns (associated to cold spells and heat-waves) afford low predictability and persistence, whereas extratropical storms are associated to patterns with higher predictability [9,10,11]. 

The ESTIMR team has also developed a strong expertise on European heatwaves. Analogues of circulation have been used to investigate the properties of observation datasets during recent events [5,7]. We have investigated the atmospheric dynamics during European heatwaves and shown how this dynamics has evolved since the beginning of the 20^th century [1]. Other case studies were investigated in an attribution context [16].

We have developed tools to analyse extra-tropical storms [6]. This methodology was combined with a weather generator based on analogues to create an efficient storm generator.

Statistical Extreme Value Analysis

A wide range of statistical advances were made to take into account issues linked to the specific extremal features in geophysical time series: change point detection in extremes [12], spatial clustering of extremes [4], climate records detection, rainfall extreme modelling [14], extreme weather generators [13], snow avalanches modelling [8] and post processing of forecasts [15]

Downscaling techniques for Extreme Events

Many impact studies or climate investigations require simulations at a much higher spatial resolution than that provided by GCMs, especially for some extremes (e.g., wind, precipitation). Downscaling of GCM simulations by dynamical (RCMs) or statistical models (SDMs) is thus often required. ESTIMR contributes to both approaches with a focus on statistical modelling of extremes and their spatial dependencies in SDMs [2,3,17].