The GPS meteorology (GPS/MET) experiment, led by the Universities Corp
oration for Atmospheric Research (UCAR), consists of a GPS receiver ab
oard a low earth orbit (LEO) satellite which was launched on 3 April 1
995. During a radio occultation the LEO satellite rises of sets relati
ve to one of the 24 GPS satellites at the Earth's horizon. Thereby the
atmospheric layers are successively sounded by radio waves which prop
agate from the GPS satellite to the LEO satellite. From the observed p
hase path increases, which are due to refraction of the radio waves by
the ionosphere and the neutral atmosphere, the atmospheric parameter
refractivity, density, pressure and temperature are calculated with hi
gh accuracy and resolution (0.5-1.5 km). In the present study, practic
al aspects of the GPS/MET data analysis are discussed. The retrieval i
s based on the Abelian integral inversion of the atmospheric bending a
ngle profile into the refractivity index profile. The problem of the u
pper boundary condition of the Abelian integral is described by exampl
es. The statistical optimization approach which is applied to the data
above 40 km and the use of topside bending angle profiles from model
atmospheres stabilize the inversion. The retrieved temperature profile
s are compared with corresponding profiles which have already been cal
culated by scientists of UCAR and Jet Propulsion Laboratory (JPL), usi
ng Abelian integral inversion too. The comparison shows that in some c
ases large differences occur (5 K and more). This is probably due to d
ifferent treatment of the upper boundary condition, data runaways and
noise. Several temperature profiles with wavelike structures at tropos
pheric and stratospheric heights are shown. While the periodic structu
res at upper stratospheric heights could be caused by residual errors
of the ionospheric correction method, the periodic temperature fluctua
tions at heights below 30 km are most likely caused by atmospheric wav
es (vertically propagating large-scale gravity waves and equatorial wa
ves).