Numerous methods of data assimilation algorithms have been developed in recent years, primarily elaborating on the feasibility to include observations over a time span, that is, space-time data assimilation methods. Most prominently, the four-dimensional variational data assimilation algorithm (4D-var) recently introduced in chemical data assimilation holds promise to exploit observations with beneficial impact for analyses of non observed species, which are chemically closely related to observed constituents. The principal drawback of these methods is the high computational expenditure.

The rationale of the data assimilation efforts at the University of Cologne is to identify a consistent chemical state of the European troposphere, combining actual measurements, a priori knowledge from climatologies or prior simulations and a comprehensive chemistry transport model. The practical use is the identification of optimal initial values for ensuing forecast runs.

While data assimilation is widely associated with identification of initial values or state analysis, space-time assimilation algorithms can also be taken for the optimization of other modeled parameters, like emission rates, deposition velocities, boundary values and others. The present implementation, for which results will be given, allows for the optimization of initial values and emission rates.

Special emphasis will be placed on the 'a priori knowledge,' the proper formulation of which ensures that the analysis results are also controlled by climatological data as far as introduced. This item includes the question of the design of the background error covariance matrix.

The underlying model is the EURopean Air pollution Dispersion (EURAD) model (CTM2), of which the adjoint version is based on the RADM2 gas phase mechanism. Results will be illustrated in more detail by a case study. A first application of the 4--dimensional variational technique to a real case study of an ozone episode during August 1997 will be presented. A number of about 400 measurement stations, mostly confined to central Europe, is available.

Generally, a significant performance improvement can be claimed not only during the assimilation interval from 6 to 12 GMT, but also for the following day. The favourable performance degrades however, presumably due to misspecified emission rates, coarse resolution, erroneous meteorological data and further biased parameters.

Back to Workshop Schedule

Back to Atmospheric Modeling

1999-2000 Reactive Flows and Transport Phenomena

Back to top of page