IMPROVING THERMODYNAMIC RETRIEVALS FROM IASI DATA USING REALISTIC OZONE AND OZONE-SENSITIVE CHANNELS

Abstract : Hyperspectral infrared sensors onboard polar-orbiting satellites provides 70% of measures used in the Numerical Weather Prediction (NWP) global model ARPEGE (Action de Recherche Petite Échelle Grande Échelle) of Météo-France, where IASI (Infrared Atmospheric Sounding Interferometer) represents 46%. The infrared passive sounding is sensitive to surface parameters and numerous atmospheric constituents. The atmospheric temperature information is retrieved from the channels which are sensitive to gases the distribution of which is known. Most of algorithms for infrared satellites measures use carbon-dioxyde (CO 2 ) sensitive channels to retrieve the temperature information. Part of the infrared spectrum are also sensitive to ozone (O 3 ) but are not currently used in the NWP models of Météo-France. In the current version of the assimilation in the ARPEGE model, the gase concentrations used for the radiance simulations are constant in space and in time. A study conducted in 2015 showed that using realistic ozone information from the Chemistry Transport Model (CTM) MOCAGE (Modèle de Chimie Atmosphérique A Grande Echelle) of Météo-France as input of the radiative transfer model improved the temperature retrievals from the infrared satellite measures. This presentation will describe the channel selection of IASI ozone sensitive channels to improve the retrievals of temperature and humidity profiles in the NWP model. There are several methods to select a set of channels to improve the atmospheric profiles such as the Jacobian and iterative methods (Rabier and Fourrié 2001). In our case, we have used DFS and sum of relative error reduction for temperature and humidity. Several settings of observation error covariance matrix have been evaluated such as the iterative Desroziers methods. We have also used different variances from operational NWP and calculated from the simulations in order to set-up the observation error covariance matrix. Results are very promising with sum of relative error reduction method, especially using the Desroziers technique (Desroziers, 2005).