|
|
7.
ATOVS Level 1b Product Processing Algorithms |
The data calibration and retrieval algorithms are documented
in the Product Generation Specification (PGS) documents [RD13],
[RD15], [RD17]. The operational processing algorithms for Level 1a are summarised
below.
1) Like the AVHRR IR channels, AMSU-A is calibrated by viewing
on-board calibration targets and cold space to determine gains
and offsets for each channel.
2) Then, a number of corrections are applied:
- A detector non-linearity correction
is applied to the measured signals based upon pre-flight measurements of
the instrument response, and the actual instrument temperature.
- A warm bias correction is also added
to the measured calibration target temperature for all 15 AMSU-A channels.
This is based upon pre-flight measurements of the difference between its
effective radiating temperature and its measured temperature. It is a function
of instrument temperature and the phase-locked loop oscillator (PLLO). It
corrects for contamination by radiation originating from the enclosure of
the black body target, which is influenced by the spacecraft temperature
and the Earth limb radiance.
- In the case of channels 9 - 14, there is a primary and a secondary phase-locked
loop oscillator (PPLO#1 and PLLO#2, respectively). Therefore there are two
different values of the warm bias correction for these channels.
- The cold bias correction linearly shifts
temperature to produce a space-view value consistent with the known radiative
background. It corrects for contamination of the measured background radiation
by radiation originating from the spacecraft and the Earth limb when viewing
space.
- A limb adjustment is applied to the
brightness temperatures provided to AVHRR. Reported Level 1 radiances and
ATOVS Level 2 products are based upon non-limb-corrected data.
- It is applied to the AMSU-A data used for the total column water vapour
content determination over sea, that is then analysed in the cloud detection
routine. The correction is based partially upon the cosine of the satellite
viewing angle. More details can be found in §5.2.2 of the AMSU-A PGS
[RD13].
- The AMSU-A stand-alone retrieval uses the same limb correction as for
AVHRR cloud detection.
- An antenna efficiency correction is
also applied to the radiances.
- A lunar intrusion correction is applied.
A model is implemented of the relative position of the moon with respect
to the field of view. Cold space counts are ignored when the moon lies within
a certain angular range; temporally interpolated cold space counts from
non-contaminated scenes are employed in their stead (see §5.1.2.32
of [RD13] for more detail).
Geolocation is calculated separately for each pixel, on the assumption of nominal
scanning properties and knowledge of the satellite's position.
1) Radiance calibration involves corrections for detector non-linearity,
warm bias, cold bias and lunar intrusion, as for AMSU-A.
2) An antenna efficiency correction is applied since the deployment
of PPF v4.0.
3) Provision has also been made for possible scan-dependent
limb corrections, but currently these will not be applied.
4) The PRT count-to-resistance conversion is carried out based
upon three constant (reference) resistors that are measured as well as the PRTs.
Geolocation is calculated separately for each pixel.
Radiance calibration is channel dependent:
1) For the IR channels (1-19):
- For gain and offset
determination:
- The cold space signal is averaged
over 48 of the 56 samples comprising one scan line. The first eight
samples are unusable since the measured radiance is still contaminated.
- All 56 samples from viewing the warm
target are used.
- Samples falling out of the "3σ"
interval of the set of measurements are not used.
- A non-linearity correction is
applied when converting counts into radiances. This correction is based
upon pre-launch measurements.
- In practice, NOAA does not apply any non-linearity
correction to their HIRS data, and EUMETSAT may not exercise this option
either. However, provision has been made in the operational processor
to implement such a correction if eventually deemed necessary.
- Calibration coefficients are
interpolated between calibration events to the specific Earth scan line.
2) Visible channel radiance calibration (channel
20) is based upon pre-launch measurements, is linear as a function of counts,
and will be updated in-orbit based upon opportunistic measurements (e.g. using
desert scenes).
Cloud coverage for HIRS (and for the Level 2 product) is determined
based upon AVHRR scenes analysis within the HIRS field of view (FOV) ellipses.
AVHRR data is mapped onto the HIRS FOV to determine the fraction
of cloudy scenes.
Geolocation is calculated separately for each pixel.