Radiance measurements from the Stratospheric Sounding Unit (SSU) and the Advanced Microwave Sounding Unit (AMSU-A) are the primary source of information for stratospheric temperature in reanalyses of the satellite era. To improve the time consistency of the reanalyses, radiance biases need to be properly understood and accounted for in the assimilation system. The investigation of intersatellite differences between SSU and AMSU-A radiance observations shows that these differences are not accurately reproduced by the operational version of the radiative transfer model for the TIROS Operational Vertical Sounder (RTTOV-8). We found that this deficiency in RTTOV was mainly due to the treatment of the Zeeman effect (splitting of the oxygen absorption lines at 60 GHz) and to changes in the spectral response function of the SSU instrument that are not represented in RTTOV. On this basis we present a revised version of RTTOV that can reproduce SSU and AMSU-A intersatellite radiance differences more accurately.
Assimilation experiments performed with the revised version of RTTOV in a four-dimensional variational analysis system (4D-Var) show some improvements in the stratospheric temperature analysis. However, significant jumps in the stratospheric temperature analysis still occur when switching satellites, which is due to the fact that systematic errors in the forecast model are only partially constrained by observations. Using a one-dimensional retrieval equation, we show that both the extent and vertical structure of the partial bias corrections must inevitably change when the nature of the radiance measurement changes with the transition from SSU to AMSU-A. Copyright © 2009 Royal Meteorological Society