Smoothing radio occultation bending angles above 40 km

The 'statistically optimal' approach to smoothing bending angles derived fr om radio occultation (RO) measurements is outlined. This combines a measure d bending angle profile with an a priori or background estimate derived fro m climatology, in order to obtain the most probable bending angle profile. However, the method is only optimal if the error statistics of both the mea sured and background profiles are known and applied accurately. In this wor k it is shown that correlations in the background estimate have a significa nt role in determining the degree of smoothing in the solution. We find tha t smooth profiles, consistent with the measured values, can be derived if t he correlations are approximated analytically with a Gaussian, assuming a s cale length of 6km. In regions where the observed and background error leve ls are comparable, the solutions take the general shape from the background estimate, centred on the observation data. The effects of correlated obser vation errors are also considered. It is shown that the quality of the temp erature retrievals can be significantly affected by the choice of climatolo gy used for background estimate.