Séminaire
Data Assimilation (DA) is the procedure through which observations and model are merged to get an improved and homogeneous estimate of a system of interest. In numerical weather and oceanic predictions DA has dramatically contributed to enhance the forecast skill and its application to coupled Earth System Simulators is nowadays regarded with strong interest as a unified initialization approach across all forecast horizons from weeks to seasonal and decadal.
The talk will provide an introduction of the context and rationale behind DA theory and will describe the main methodological options arisen in geosciences in the last decades with a focus on the theoretical and mathematical challenges encountered. As an example of a new promising field of application of DA a new approach allowing for near real time, systematic causal attribution of weather and climate related events is described. The method is purposely designed to allow its operability at meteorological centers by synergizing causal attribution with DA. It is shown how casual attribution can be obtained as a by-product of the statistical inference undergone when doing the assimilation of data. The theoretical rationale of this approach is explained along with the most prominent features of a DA-based detection and attribution procedure. The proposal is illustrated in the context of the 3-variables Lorenz model and is compared with standard methods for detection and attribution showing promising performance.
The method stresses on the concept of model evidence, and open questions on how to compute and interpret the response to forcings whose effects one wants to contrast. A discussion on the lines taken to address these issues is provided. Practical obstacles that need to be addressed to make the proposal readily operational within weather forecasting centers are finally laid out.