COUPLED DYNAMICS/CHEMISTRY DATA ASSIMILATION AND OZONE STUDIES

In recent years several research groups have demonstrated that data assimilation techniques can be very successfully used for analysis of atmospheric chemical observations. Fundamentally, all data assimilation techniques use either variational or sequential technique. While both were proven to yield the same analysis under special conditions, we believe that the sequential approach might be more appropriate for some practical applications as it allows researches to gain additional insights through explicit computation and examination of (approximate) error covariances and biases. Proper analysis of these quantities helps to attend to concerns of "contaminating" data with "models" often arising when assimilation results are presented without corresponding statistical indicators of their value. Most research efforts conducted so far on assimilating observations of chemical composition of the atmosphere using the sequential approach fall into two categories. Research efforts in the first category concentrate on and attempt to take into account spatial correlations in distributions of atmospheric chemicals while ignoring the cross-species correlations. An advantage of this approach is that it can be used, albeit with some limitations, to fill in temporal and spatial gaps between sparse and irregular observations. Those in the second category focus on the cross-species correlation, usually in the Lagrangian framework, and largely ignore the spatial correlations. An advantage of this approach is that concentrations of some non-observed species can be constrained, at least to some extend, by the observed ones. We review selected results obtained with both approaches in studies of atmospheric chemistry and, particularly, ozone distribution and discuss development of a coupled three-dimensional data assimilation system that accounts for both spatial and cross-species correlations.