Figure 4.2: Level 0 to Level 1a processing steps
Process Level 1a to Level 1b
The Level 1b processing function calculates the bending
angle and the impact parameter from the instrument-corrected occultation
The Level 1b processing function performs occultation
isolation to combine GRAS data for each occultation with the auxiliary
data required to retrieve the bending angle profile. The pivot GPS satellite
and the fiducial station supporting differencing schemes (for clock correction)
have to be selected before all auxiliary data for each occultation can
The Level 1b PPF performs several corrections to the
measurement data before the actual bending angle retrieval is performed.
The phase residual, which is to a good approximation the phase delay introduced
by the atmosphere, is calculated by removing the geometrical distance
between the transmitter and receiver antennas from the measured phase.
This requires determination of the true reception and transmission times
and interpolation of the satellite state vectors into these times. The
corrections for relativistic effects are mostly included into the synchronisation
of the measurement time stamps with the reference time provided by the
GRAS GSN because the relativistic effects are included in the clock offset
estimates calculated in the GPS and GRAS/Metop NRT POD. The only relativistic
effect not included in the clock offset estimates is the variation in
the apparent velocity of light because of the gravitational field of the
Earth (Shapiro effect). This effect is taken into account in the determination
of the transmission time and geometric path removal.
After the removal of the geometric path the measured
phase residual is still wrapped around 2π. The unwrapping of the
phase is combined in this algorithm.
After the relativity correction a cycle slip detection
and correction function is applied to the phase residual data.
The Level 1b PPF corrects the data provided by the Level
1a function for clock drifts on board the GPS satellite and, if necessary,
the GRAS instrument. The Level 1b processing function obtains, via the
PGE, for each of the ground stations supporting differencing the Sounding
Support Data (SSD). GSN also provides an estimate of the Tropospheric
Zenith Delay (TZD) for each fiducial station and local surface level meteorological
observations (if available). TZD has to be mapped to the elevation of
the occulting and pivot GPS satellites by the Level 1b PPF.
|No differencing (ND)
||All clocks in the observation system are considered sufficiently
stable and no clock correction is required. Clock biases are removed
by using bias estimates from POD.
|Single differencing 1 (SD1)
||GPS clock is considered stable and only the impact of the GRAS clock
instability is corrected for. The differencing is performed between
links A and D in the figure below.
|Single differencing 2 (SD2)
||GRAS clock is considered stable and the impact of the GPS clock
instability is corrected for (current baseline scenario). The differencing
is performed between links A and B in the figure below.
|Double differencing 1 (DD1)
||All observation system clock errors are corrected for (GPS, GRAS,
fiducial stations). The differencing is performed between all measurement
links in the figure below.
|Double differencing 2 (DD2)
||Similar to DD1, but two ground stations are used. One station tracks
the occulting GPS satellite (GPS-1 in figure below) and the other
tracks the pivot satellite (GPS-2 in figure below). The advantage
is that neither station has to have visibility to both GPS satellites.
The disadvantage is that the ground station clock errors are not removed.
Figure 4.3: Measurement links used for clock corrections
The baseline scenario for the GRAS PPF is clock correction
with SD2. DD1 and DD2 are considered as fall-back options in the case
that SD2 cannot provide good product accuracy. ND and SD1 are optional
differencing methods that may be applied depending on the GPS clock characteristics.
The use of additional GPS satellites in the clock correction will introduce
noise on the bending angle, thus ND will be used if the GRAS clock is
found to be stable enough. The actual applied clock correction can be
found in the Level 1b data products.
In deriving the total bending angle, the Level 1b processing
function assumes a locally spherical atmosphere. The errors introduced
by this assumption are reduced by applying a correction for the Earth’s
oblateness. The Level 1b processing function computes correction parameters
for this purpose.
The derived phases of the occultation data are corrupted
by high-frequency noise. The Level 1b processing function therefore low-pass
filters the derived phase data. The filtering function is based on Savitzky-Golay
(see [SCD9], [SCD10]).
The Level 1b processing function computes the Doppler
shift (as a time derivative) for the phase residual observations in the
occultation. It retrieves the bending angle as a function of the impact
parameter by using the Geometrical Optics (GO) approximation. Additionally,
Wave Optics (WO) processing is applied to parts of the measurement, using
phases and amplitudes to derive a bending angle profile. GO is applied
to the whole measured profile and WO to the lower part of the profile
to detect and remove the impact of atmospheric multipath.
The frequency-independent neutral bending angle is computed
by correcting for ionospheric dispersion, by applying a linear combination
of the bending angles at two frequencies. Bending angle bias is calculated
and a correction is applied if necessary. The Level 1b processing function
also derives the total electron content (TEC) along the ray path. Error
characterisation is performed for all Level 1b products. For the raw sampling
mode the Level 1b processing algorithm is to be defined.
Further information is available at [RD5].
A functional decomposition of the GRAS L1a processor is shown in the