Error Characteristics of Refractivity Profiles Retrieved from CHAMP Radio Occultation Data

We present results of an empirical error analysis of refractivity profiles based on CHAMP radio occultation data. We analyzed two seasons of observations, boreal winter 2002/03 and boreal summer 2003. The processing was performed with the WegCenter/CHAMPCLIM Retrieval version 2. The error statistics is based on comparison to reference profiles calculated from ECMWF analyses fields. Bias profiles and error covariance matrices are provided, the latter separated into standard deviation profiles and error correlation matrices. Since the error characteristics contain both the observational error of the retrieved data and the model error of the ECMWF analyses we performed an estimation of the ECMWF model error and separated the observation error. The relative refractivity bias of CHAMP radio occultation data with respect to ECMWF was found to oscillate around −0.4 % at 5–25 km globally. Wavelike structures apparent at high latitudes in Southern Hemisphere winter are mainly due to the representation of the polar vortex in the ECMWF analyses. The combined relative standard deviation was found to be 0.7–1 % at 5–25 km height globally, showing larger values in winter than in summer in the upper stratosphere at mid- and high latitudes. The global observation error for CHAMP refractivity was estimated to be 0.5–0.75 % at 6–30 km. The results are compared to the findings of Kuo et al. (2004) and to those of an end-to-end simulation study being the precursor of this work (Steiner and Kirchengast 2004, 2005). Based on the simulation study we provide simple observation error covariance matrix formulations for CHAMP refractivity for convenient use in retrieval algorithms and in data assimilation systems.