Background on The GPS/MET Experiment

The GPS radio occultation technique was tested for the first time with the launch of the GPS/MET mission from April 3, 1995 to March 1997. GPS/MET is an experiment managed by the University Corporation of Atmospheric Research (UCAR) ( Ware et al., 1995) and consisted of a 2 kg GPS receiver piggybacked on the MicroLab I satellite which had a circular orbit of 730 km altitude and 60 degree inclination. The GPS receiver was a space qualified TurboRogue ( Meehan et al., 1992) capable of tracking up to 8 GPS satellites simultaneously at both frequencies transmitted by GPS. Under an optimal mode of operation, the GPS receiving antenna boresight was pointed in the negative velocity direction of the LEO providing 100-150 globally distributed setting occultations per day. Tens of thousands of occultations were collected and can be used to assess the accuracy and potential benefit of the GPS radio occultations.

Under ideal conditions, when a LEO tracking GPS has a 360 degree field of view of the Earth's horizon, about 750 occultations per LEO per day can be obtained. However, side-looking occultations (GPS-LEO link > 30 degree from velocity or anti-velocity of LEO) sweep across a large horizontal region, and the spherical symmetry assumption used in the abel inversion becomes inaccurate. Discarding side-looking occultations, one LEO provides up to 500 occultations per day. GPS/MET however tracked only setting occultations using an aft-looking antenna, therefore, reducing the maximum number of occultations to 250 per day. Moreover, due to a limited on-board memory and insufficient coverage from ground GPS stations, required for calibrating the GPS satellites' clocks, the maximum number of occultations obtained from GPS/MET varied between 100-150 occultations per day.

Early Results from GPS/MET

Results presented here correspond to a subset of data from GPS/MET collected on May 4 and 5 of 1995. [See Hajj et al., 1995; Kursinski et al., 1996; Ware et al., 1996; Leroy, 1997; or /absolute_urlbrowse the Neutral Atmospheric Data Browser for more examples]

Click on image to see the full-size version

The figure above compares a GPS/MET retrieved temperature profile, representative of high latitude conditions, with radiosonde data and model analyses. This profile from northern Canada has surface temperatures below freezing and a sharply defined tropopause near 8 km. Cold, dry conditions allow temperatures to be derived accurately almost to the surface. Temperature agreement between the retrieved profile, the model profile and a nearby radiosonde profile is excellent, with differences smaller than 1 K through most of the troposphere. Near and above the tropopause, temperature differences are comparable with the variability between the radiosonde and the model analysis. The striking agreement with the radiosonde in resolving the sharp tropopause and the sudden lapse rate change below 3 km illustrates the sensitivity and vertical resolution of the occultation technique.

Click on image to see the full-size version

The second example shown above illustrates the sensitivity of radio occultation to atmospheric waves with an occultation in the south Pacific (7.9 deg. S, 167.5 deg. E). The occultation is compared to a radiosonde profile obtained from a ship at a 350 km distance from the occultation location, and the ECMWF model analysis. Close agreement in both amplitude and phase with a radiosonde sounding implies that the wave has been resolved by the occultation measurement and that the horizontal wavelength is large (>>350 km).

A statistical comparison of retrieved profiles with the ECMWF analyses is shown below. The ECMWF analyses are one of the best available global analyses of atmospheric temperature structure below 30 mbar, and comparison against them has become an increasingly popular method for evaluating the accuracy and resolution of observational results. However, it should be noted that the occultations are sensitive to vertical structure not resolved by the analyses. The figure below shows temperature difference statistics for all profiles retrieved on May 4 and 5, 1995 with the exception of one 5-s outlier over the Tibetan plateau. In order to eliminate temperature retrieval errors due to water vapor, tropospheric temperatures exceeding 250 K have been excluded from the comparisons. The three panels in the figure display temperature difference statistics for the northern high latitudes, the tropics, and the southern high latitudes. There are approximately 30 profiles within each latitude zone, widely scattered in both location and time.

Click on image to see the full-size version

It is clear from the figure that agreement between the two data sets in the northern hemisphere is impressive with mean differences of generally less than 0.5 K and difference standard deviations of typically 1 to 2 K. It should also be remembered that these differences include retrieved vertical structure that is not resolved by the ECMWF analysis, especially above 100 mbar. This agreement is particularly significant because the ECMWF analyses are expected to be most accurate in the northern hemisphere. Although both radiosonde and TOVS data are assimilated into the ECMWF model, the analyses are expected to be less accurate in some regions of the southern hemisphere due to the sparse distribution of radiosondes. Southern hemisphere radiosondes cluster over a few land masses whereas the occultations fall mostly over the ocean. The figure also shows that in the southern hemisphere, both mean temperature differences and standard deviations increase at lower altitudes. As the occultation retrieval process has little dependence on latitude, the good agreement in the northern hemisphere suggests that the larger systematic and random differences at southern latitudes originate in the analyses rather than in the retrieved profiles.