Tropospheric Ozone Monitoring with IASI/MetOP Using a Self - Adapting Regularizati on Method

Tropospheric ozone is a key species for tropospheric chemistry and air quality. Its monitoring is essential to quantify sources, transport, chemical transformation and sinks of atmospheric pollution. Accurate data are required for understanding and predicting chemical weather. Space-borne observations are very promising for these concerns, especially those from IASI/MetOp. However, their sensitivity near the surface remains limited and advanced retrieval methods are needed to access to the information from the lowest troposphere. Ill-conditioning is a well-known issue of the retrieval of vertical atmospheric profiles. It produces oscillations in the retrieved profiles beyond the error margins defined by the mapping of the measurement noise onto the solution. Tikhonov regularization is often used to improve the conditioning of the inversion. As for any regularization scheme, a crucial step is the choice of the strength of the applied constraint. This choice depends on the measurement errors and on the sensitivity of the measurements to the target parameters at the different altitudes. For this reason a self-adapting and altitude-dependent regularization scheme is likely preferable over a fixed strength determined apriori, on the basis of sensitivity tests. Such a scheme was already introduced in 2009 and applied to atmospheric profiles retrieved from MIPAS/ENVISAT. The implementation of this method on nadir IASI retrievals required the appropriated definition of the target function used to optimize the constraint for lower tropospheric retrievals. The challenge for this new retrieval algorithm is to limit the use of a priori constraints to the minimal amount needed to perform the inversion. Since the sensitivity of the observations to the ozone amount in the lowest layers depends on the atmospheric and surface conditions, it is crucial for the inversion algorithm to tune accordingly the contribution of the a priori information. We apply the method first on simulated observations of tropospheric ozone for August 20th, 2009 over Europe. A first evaluation of the method is discussed in the paper. Significant improvements in terms of degrees of freedom (DOF) for the solution are achieved with a 15% increase on average. The error estimate during the retrieval is in better agreement with the true error, calculated as the difference between the retrieved ozone and the true ozone. The spatial distribution and the dispersion of the error are better described. Finally, a first attempt to apply the method to actual IASI measurements is presented.