Radiance Validation for the
MOPITT Instrument
The validation phase of the MOPITT
mission is well underway. Two classes of products from MOPITT will
become available. The Level-1 Product consists of the sixteen geolocated
radiances measured by the instrument at each pixel. Of these, twelve
are sensitive to CO, and four to CH4. The Level-2 Product consists
of the retrieved trace gas quantities corresponding to each MOPITT pixel
(in addition to several retrieval byproducts, including surface temperature).
Since the Level-1 Product is a required input to the Level-2 Product,
it must be validated first. The MOPITT retrieval algorithm is based
on the maximum likelihood technique, and makes heavy use of a fast operational
radiative transfer model known as MOPFAS. The degree to which this
model accurately calculates the MOPITT radiances is a main determinant of
the accuracy of the MOPITT Level-2 Product.
MOPITT Radiances
The twelve CO-sensitive MOPITT
radiances include four solar-channel signals (in the 2.3 um band) and eight
thermal-channel signals (in the 4.6 um band). The thermal-channel signals,
which are the subject of the current study, are evenly divided between 'Average'
and 'Difference' signals. 'A' signals are sensitive mainly to the spectral
regions between the CO absorption lines, and are therefore very insensitive
to CO concentration. Conversely, 'D' signals are sensitive only to
spectral regions very close to the absorption lines and are therefore very
sensitive to varying CO. Both the 'A' and 'D' thermal-channel radiances
are generally very sensitive to surface temperature.
Radiance Validation Processing
The goal of this research project
is to compare satellite-measured radiances with model-calculated radiances
in situations where the most relevant inputs to the operational forward radiative
transfer model can be accurately specified. Quantitative agreement
between the measured and modeled radiances over widely varying atmospheric
conditions is essential to producing reliable trace-gas retrievals.
Initially, we have chosen to validate the MOPITT thermal-channel radiances
using the following sources as inputs to the forward radiative transfer model.
- Meteorological data
is taken from NCEP, and is temporally and spatially interpolated to the
individual MOPITT pixels.
- Surface temperature
, to which all of the MOPITT thermal-channel signals are very sensitive,
and which exhibits large variability in space and time, is retrieved from
the MOPITT 'A' signals. Generally, this forces agreement between the
calculated model 'A' signals and the observed satellite 'A' signals.
- The input CO profiles
are generated by combining the CMDL in-situ profiles (measured from aircraft)
for the lower- to mid-troposphere (p > 350 mb) with mixing-ratio models
to represent the range of expected CO uncertainty in the upper troposphere
and lower stratosphere.
For one in-situ validation profile,
all the available MOPITT clear-sky data within a given temporal and spatial
'window' centered on the time and location of the in-situ profile are collected.
For each MOPITT pixel found, the in-situ profile is fed to the forward radiative
transfer model (MOPFAS) along with the relevant ancillary data for that pixel
(described above). This produces two datasets, representing the observed
(satellite) radiances and a corresponding set of model-calculated radiances.
Each set of radiances corresponding to one in-situ validation profile is
averaged to produce a mean observed radiance and a corresponding mean modeled
radiance.
Validation Profile Data
In support of MOPITT, NOAA's Climate
and Monitoring Diagnostics Laboratory (CMDL) is currently engaged in a program
to make regular in-situ CO profile measurements at five sites: Carr,
Colorado; Harvard Forest, Massachusetts; Poker Flats, Alaska; Molokai, Hawaii;
and Raratonga in the Cook Islands. In addition, CMDL is often involved
in mission-specific field campaigns, such as SAFARI-2000 (which took place
in August-September, 2000, in Southern Africa).
During October, 2000, NCAR personnel
obtained access to a Citation aircraft, which is able to fly (and make in-situ
measurements) considerably higher than the aircraft generally used at the
five CMDL anchor sites. The aircraft acquired four useful in-situ profiles
over Colorado and California during this period. Validation radiances for
these profiles are particularly interesting because they include data at
upper levels where the mixing ratio can only be estimated (by extrapolation)
for the standard CMDL anchor sites. These data are most important for
validating the 'D' signals for channels 3 and 7, which are most sensitive
to CO in the upper troposphere and lower stratosphere.
Validation Quality Issues
Altitude Limits of CMDL Aircraft
The MOPITT weighting functions
exhibit significant sensitivity to CO throughout the troposphere and into
the lower stratosphere (up to ~ 100-200 mb). In contrast, the CMDL
aircraft from which the in-situ profiles are recorded typically have an upper
altitude limit of about 350 mb. To fill this 'data void', the in-situ profiles
are extended by interpolating between the monthly mean climatological CO
concentration (from model output) at 100 mb, and the highest actual in-situ
level (typically between 350 and 400 mb). Low and high scaling factors
are then applied to this extrapolated part of the profile to represent the
expected range of CO variability, which could cause error in the validation
radiances.
The upper altitude limit for
the Citation aircraft is above the 200 mb level. For the in-situ profiles
measured from this platform, the validation radiance errors associated with
the high-altitude 'data void' should be expected to be significantly less
than for profiles with lower altitude limits.
Temporal and Spatial Matchup
Quality
Generally, MOPITT observations
used for validation will be displaced relative to the in-situ observations
both spatially and temporally. For the validation radiances reported
here, only MOPITT pixels recorded within 3 hours and within 300 km of the
in-situ data are retained. Nevertheless, significant validation radiance
errors might still occur if the measured in-situ profile is influenced by
localized phenomena (e.g. plumes from agricultural or industrial emissions).
The validation radiances for the SAFARI-2000 experiment (during which there
was significant biomass burning activity) are probably most susceptible to
errors of this type.
Results
Validation Results for CMDL
Anchor Sites
Thermal-channel validation radiances
for the five CMDL anchor sites and SAFARI-2000 profiles are shown in
Figure 1. As expected, satellite and modeled radiances for all
four thermal-channel 'A' signals agree very well. More importantly,
however, the satellite and modeled radiances for all four thermal-channel
'D' signals also agree very well. Generally, the agreement for the
'D' signals is highest for channels 1 and 5, which are most sensitive to
CO in the lower and middle troposphere, and somewhat less for channels 3
and 7, which are relatively more sensitive to CO in the upper troposphere
and lower stratosphere. This is most likely due to the fact that the
upper part of the assumed in-situ profile (obtained by extrapolation) is
much more prone to error than is the lower part of the profile (see section
above on 'Altitude Limits of CMDL Aircraft'). Most importantly, no significant
bias is apparent in any of the thermal-channel A or D signals.
Validation Results for Citation
Profiles
Validation radiances for the four
profiles measured from the Citation aircraft are shown in
Figure 2. In comparison with
Figure 1, it is apparent that the the observed and modeled radiances
agree somewhat better for these profiles than for the CMDL anchor site profiles.
It is believed that the better agreement for these profiles is directly attributable
to the inclusion of in-situ measurements of CO at upper levels unattainable
with the aircraft used at the CMDL anchor sites.
Summary
- Radiance validation is an essential
prerequisite to validation of the MOPITT CO and CH4 trace-gas retrievals.
- Current focus of MOPITT radiance
validation is on CO thermal-channel 'A' and 'D' signals.
- Radiances are validated by
feeding forward radiative transfer model 'known' inputs, and comparing model-calculated
radiances with actual satellite radiances.
- Radiance validation errors
(unrelated to the performance of the forward radiative transfer model) may
arise from temporal and spatial displacements between the MOPITT pixels and
CO in-situ data, and from the lack of in-situ data for the upper troposphere
/ lower stratosphere region.
For more information, contact
Merritt Deeter.
7 Mar 2001