|Title||Background error statistics for aerosols|
|Year of Publication||2006|
|Authors||Benedetti, A, Fisher, M|
|Secondary Title||Technical Memorandum|
As part of the European initiative for Global Monitoring for Environment and Security (GMES), the Global and regional Earth-system Monitoring using Satellite and in-situ data (GEMS) project has been funded to provide forecasts and analysis of atmospheric constituents such as aerosols, and greenhouse and reactive gases. In this context, the development of an aerosol analysis system based on fourdimensional variational (4DVar) assimilation was started in March 2005 at the European Centre for MediumRange Weather Fore-casts (ECMWF). Among the main building blocks of the system, an error background covariance matrix for aerosol mixing ratio was built based on background statistics constructed from aerosol forecasts. These statistics were generated from the differences between the 48h and 24h forecasts of aerosol mixing ratios for seasalt, desert dust and continental particulate, using the NMC method. This method has been widely used by many Numerical Weather Prediction Centres to build background error statistics for state vari-ables. This is the .rst time, to our knowledge, that it is applied to a novel variable such as aerosol mixing ratio, in the context of a global model. Aerosol assimilation systems have been successfully run with ei-ther observationallyderived or modelprescribed error covariance matrices with preassigned correlation lengths. In this study, the aerosol vertical and horizontal structures of these error correlations as derived with the aid of a state-of-the-art model are investigated and discussed in depth. Successively these error correla-tions are modelled through a spectral/wavelet approach and used to construct a background error covariance matrix. Application of this background matrix in the context of a single observation 4DVar experiment produced successful preliminary analyses of total aerosol mixing ratio.