Use of GPS/MET refraction angles in three-dimensional variational analysis

The Spectral Statistical Interpolation (SSI) analysis system of the National Centers for Environmental Prediction (NCEP) is modified to include GPS/MET data (meteorological data from the Global Positioning Satellite system) using a GPS ray-tracing operator. The new system is tested by incorporating 30 actual GPS/MET observations of refraction angles obtained during the GPS/MET experiment. This is the first time that real radio occultation refraction angles and refractivities have been incorporated into a three-dimensional variational analysis system. We examine the magnitude and the vertical distribution of the analysis adjustments that result from using refraction-angle observations in the NCEP SSI analysis system. The average magnitudes of the adjustments in the temperature and specific-humidity fields are approximately 0.4 degC and 0.6 g kg(-1), respectively. Individual changes can be as large as 4 degC and 4 g kg(-1), respectively. The greatest adjustments to the temperature occur in the middle and upper troposphere and stratosphere, while the major changes in specific humidity occur in the lower troposphere. An assessment of the impact of the GPS/MET observations on the analysis, verified by conventional (mostly radiosonde) data, is difficult because of the small number of GPS/MET data used. Nevertheless, it is found that, even over data-rich regions (regions containing many radiosonde observations), and even when the verification data were the radiosonde data themselves, the use of GPS/MET refraction angles makes a slight improvement, overall, to the analysed temperatures and winds. The impact on the water-vapour analyses, again as measured against radiosonde data, is mixed, with improvements shown in some layers and degradation in others. Compared with the background field, the use of refraction angles from one occultation results in an analysis whose simulated refraction angles are much closer to the withheld GPS/MET refraction angles at the two nearby occultation locations, and whose temperature and moisture profiles are also closer to those resulting from the direct assimilation of the two withheld occultations. Although the forward model used in this study, with the ray tracing being carried out in a two-dimensional plane, is much cheaper than a more accurate three-dimensional forward model, it is still quite expensive. In order to further reduce the computational requirement for the assimilation of GPS/MET data, we test a scheme in which the GPS/MET-retrieved refractivities (instead of refraction angles) are used above a selected height for each occultation. These heights are determined objectively based on the departures from spherical symmetry of the model held. It is shown that the mixed use of GPS/MET refraction angles and refractivities produces an analysis result similar to the one using refraction angles alone, while the computational cost is reduced by more than 30%.