|From||Vivien Mallet <Vivien.Mallet@inria.fr>|
|Date||Mon, 11 Apr 2011 17:58:19 +0200|
PhD Thesis: Coupling Ensemble Forecast and Data Assimilation -- Application to Air Quality Simulation Nowadays, air quality forecasts are carried out using chemistry-transport models. Based on meteorological forecasts, these models compute pollutant concentrations (like ozone over Europe) for a few days ahead. Shortcomings in the forecasts originate from the high uncertainties in the input data to the models (meteorological fields, emissions, ...) and in the physical formulation of the models (turbulence, chemistry, ...). In such a context, forecasting should not rely on a single model. Instead, a forecast should be based on an ensemble of models that should account for all uncertainty sources. In order to reduce the uncertainties, data assimilation methods take advantage of ground-based and satellite observations. These assimilation methods actually merge the information contained in a numerical model and the information brought by the observations, so as to produce the estimate of the model state that minimizes the error variance. Several such methods are appropriate for high-dimensional systems like air quality models. They naturally apply to a single model. Meanwhile, better forecasts have been produced by ensemble methods in which the forecasts of several models are linearly combined. The weights of the linear combination may be determined by machine learning algorithms, based on past observations and forecasts. This approach is often referred to as ensemble forecasting. The objective of the PhD thesis is to develop methods that combine the two approaches: data assimilation and ensemble forecasting. A proper theoretical framework will be needed for these new methods. Application to air quality forecast will probe their efficiency. The first research advances in this direction are extremely promising. Further information: http://www-rocq.inria.fr/clime/jobs/2011/ensemble_assimilation-en.pdf Contact: Vivien Mallet (firstname.lastname@example.org, +33 01 39 63 55 76)
Go to: Periods · List Information · Index by: Date (or Reverse Date), Thread, Subject or Author.