Statistical modelling at different climatic and environmental scales

Statistical climatology develops original approaches to modelling climatic phenomena, their processes, their uncertainties, etc., based on statistical approaches and probabilistic concepts. This hybrid and relatively recent discipline now represents an undeniable force for understanding climate and environmental variability. In this inherently multidisciplinary context, my research has mainly focused on three interdependent areas: the characterisation and modelling of weather patterns, the development of statistical approaches to regionalization Euros 'downscaling'), and the modelling of extreme events. These areas are associated with different spatial scales and often different temporal scales: weather patterns and their properties provide information on large spatial scales by characterising atmospheric structures several hundred kilometres across, with persistence of several days. Statistical downscaling makes it possible to simulate climatic or meteorological phenomena on very small scales (i.e. very local, e.g. at the level of weather stations) by constraining them with various large-scale information. Extreme events, on the other hand, can be considered both on large spatial scales (e.g. heat waves or droughts) and on much more local scales (e.g. extremely intense rainfall, often brief, which can generate so-called flash floods, must be modelled at high resolutions, e.g. at catchment level, to be relevant. These three areas also provide valuable information for different time horizons: Weather patterns are generally studied for modes of variability in the present climate, but also make it possible to assess their potential changes in the future or since a more or less distant past climate (e.g. the last millennium). Similarly, statistical modelling at high spatial resolution can be used to study processes (continental, atmospheric) in the present climate, but also to make local projections of climate variables, which are necessary for studies and models of the impacts (ecological, hydrological, economic, etc.) of future climate change, or, for example, for model-data comparisons in the context of palaeo-climatic studies. Finally, while studies on contemporary extreme events are relevant for better characterising rare phenomena and better understanding our vulnerability to climate, they must also be deployed to define risk maps (e.g. return level map for a 100-year event, or a 1,000-year event), not only for the present but also and above all in the context of climate change (e.g. for the construction of buildings to protect against extreme climatic events), which is likely to change both the frequency and the intensity of these intense phenomena. Overall, understanding risk (climatic and environmental) and the associated uncertainties requires statistical concepts and estimates of the probabilities of various events. The role of statistical modelling is therefore central. My work provides a natural and operational link between climatology and other fields influenced/impacted by climate, by developing statistical concepts and tools that are made available to the entire climate and impact community