QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY Q. J. R. Meteorol. Soc. 135: 225–237 (2009) Published online 21 January 2009 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/qj.360 The potential of variational retrieval of temperature and humidity profiles from Meteosat Second Generation observations F. Di Giuseppe,* M. Elementi, D. Cesari and T. Paccagnella Servizio IdroMeteoClima-ARPA Emilia e Romagna, Bologna, Italy ABSTRACT: The quality of temperature and humidity retrievals from the infrared Spinning Enhanced Visible and Infrared Imager (SEVIRI) sensors on the geostationary Meteosat Second Generation (MSG) satellites is assessed by means of a one- dimensional variational algorithm. The study is performed with the aim of improving the spatial and temporal resolution of available observations to feed analysis systems designed for high-resolution regional-scale numerical weather prediction (NWP) models. The non-hydrostatic forecast model COSMO in the ARPA-SIMC operational configuration is used to provide background fields. Only clear-sky observations over sea are processed. An optimized one-dimensional variational set-up comprised of two water-vapour and three window channels is selected. It maximizes the reduction of errors in the model backgrounds while ensuring ease of operational implementation through accurate bias correction procedures and correct radiative transfer simulations. The 1Dvar retrieval quality is first quantified in relative terms, employing statistics to estimate the reduction in the background model errors. Additionally the absolute retrieval accuracy is assessed by comparing the analysis with independent radiosonde observations. The inclusion of satellite data brings a substantial reduction in the warm and dry biases present in the forecast model. Moreover it is shown that the use of the retrieved profiles generated by the 1Dvar in the COSMO nudging scheme can locally reduce forecast errors. Copyright c  2009 Royal Meteorological Society KEY WORDS 1Dvar; data assimilation; Meteosat Second Generation Received 13 December 2006; Revised 12 November 2008; Accepted 20 November 2008 1. Introduction The requirement to improve the prediction of severe weather events and localized heavy precipitation regimes driven by deep convection and complex orography has led many national forecast centres to increase the resolution of regional-scale models steadily, which poses unprece- dented problems for widely used variational data assimi- lation systems. On the one hand, analysis at 1 km resolu- tion requires dense and frequent observations to capture spatially incoherent and quickly evolving structures typ- ical of the meso-γ scales. On the other hand, the appli- cability of variational techniques such as 4Dvar imposes the linearity of the observation operator and the valid- ity of balances imposed by the large-scale flow (Rabier et al., 1998), which are violated at high resolutions. As the role of non-conventional observations such as radar or new high-resolution satellite platforms becomes cru- cial for regional modelling at the km scale, the suitabil- ity of 3Dvar/4Dvar analysis systems, which allow direct radiance/reflectivity assimilation, remains widely debated (Chevallier et al., 2003, 2004). ∗ Correspondence to: Francesca Di Giuseppe, Servizio IdroMeteoClima- ARPA Emilia e Romagna, Bologna, Italy. E-mail: fdigiuseppe@arpa.emr.it A possible way to overcome the linearity problem at high resolution while still retaining a variational approach is to perform an intermediate step in which part of the satellite information is transferred to prognostic vari- able increments, which can then be used as ‘pseudo- observations’ to be ingested into the assimilation system (Mar´ ecal and Mahfouf, 2000). In this case the assimi- lation is performed in two steps. The first step consists of applying a 1Dvar algorithm to the observed quantity seeking optimal, in a least-squares sense, model variables (e.g. temperature, humidity etc.) which fit within spec- ified model and observational errors. The second step incorporates the 1Dvar retrieved products as ‘pseudo- observations’ into the assimilation system itself. A weak- ness of this methodology is the creation of a correlation between these pseudo-observations and the model state (Moreau et al., 2004); moreover, observational errors for these new observations are often specified in an empirical fashion. Nevertheless the simplicity of the approach leads to some advantages. Firstly, it enhances the tractability of the assimilation problem for non-conventional obser- vations in those schemes such as the nudging technique, which would not permit raw radiance assimilation for example. Secondly, in a 1Dvar contest it is straight- forward to assess the impact of a specific observation by means of linear estimation theory (Rodgers, 2000) and /or by comparing the analyzed profiles with in situ Copyright c  2009 Royal Meteorological Society