September Ocean Color Announcement

[genecarlfeldman]

Ocean Biology Processing Group Monthly Progress Report: September 2004

The group continues to pursue improvements in the SeaWiFS-MODIS/Aqua
comparisons primarily through time series tests, analyses of the solar
diffuser and lunar calibration data in collaboration with the MCST, and
refinements in the interpretation/application of the prelaunch MODIS/Aqua
characterization data. Overall, these comparisons show that the binned
water-leaving radiance products in the open ocean are within 10%. However,
some seasonal and regional patterns remain, e.g., a latitudinal difference
that shifts seasonally. Differences at the 10% level are not particularly
serious, but are not within our goal of 5%. The fact that the patterns are
coherent spatially, temporally, and spectrally implies that a deterministic
explanation/correction should be possible. These differences are somewhat
mitigated in the band-ratio bio-optical algorithms because of the spectral
coherence. At level-2, stripping remains an issue and corrections at the
detector level are being evaluated. The modification of MSL12 to be more
flexible, e.g., accommodate more than eight bands, and efficient has been
completed. This has allowed us to incorporate the fluorescence bands and
implementation of the fluorescence line height product is progressing
forward in collaboration with Ricardo Letellier. Below are reports by the
staff on their accomplishments since the last report in August. Please
feel free to contact us if you have questions, comments, and suggestions.

Chuck McClain and Gene Carl Feldman

Individual Staff Reports:

Gene Eplee (eplee@seawifs.gsfc.nasa.gov)
MODIS Aqua Dark Count Analysis
We have begun an analysis of the MODIS Aqua dark count which is subtracted
from the 1-km reflective solar band data as part of the level-1b conversion
process. Operationally, the dark count is computed on a per-scan-line basis
from the space-view data and subtracted from all of the earth-view pixels
in the corresponding scan line. On the day side of an orbit, earthshine
within the instrument cavity contaminates the dark count by amounts that
vary from band to band, ranging from approximately one count in band 8 to
approximately four counts in band 16. The earthshine contamination varies
with the position of the spacecraft in the orbit and with time of year.
This contamination represents an effect in band 8 of > 0.1% of the
top-of-the-atmosphere radiance in band 8, of ~0.2-0.3% of the TOA radiance
in band 15, and of ~0.4% of the TOA radiance in band 16. The impact of the
earthshine on the retrievals of water-leaving radiances are being
investigated. Several approaches to correct for the earthshine are also
being evaluated.

Ewa Kwiatkowska (ewa@simbios.gsfc.nasa.gov)
TOA MODIS-Aqua and SeaWiFS matchups
Since MODIS-Aqua and SeaWiFS have been concurrently on orbit, there have
been a number of days for which both sensors imaged coincident ground
coverage with very close viewing and solar geometries. A time-series of
these overlapping pixels has been created and top-of-the-atmosphere (TOA)
and water-leaving radiances (nLw) have been matched between the
sensors. Within this data set, the only discrepancies between both
instruments originate from sensor-specific calibration and algorithm
artifacts. The benefit of the TOA matchups is that they are independent of
atmospheric correction and surface BRDF uncertainties. Also, most of the
nLw uncertainties should affect both sensors in the same manner. On the
other hand, the data set is limited. The matching overpass and geometry
between the sensors have only occurred on several dates through the span of
both missions and the derived pixels fall within a narrow path at around
50-degrees latitude North and within a confined slot of the MODIS scan line
around the pixel 200.
Further investigations into the data set focused on deriving a strategy for
converting TOA radiances from SeaWiFS to MODIS bands. If such conversion is
done reliably, the matchups can provide an independent assessment of
relative sensor calibrations. A theoretically derived TOA radiance spectrum
was used as a basis for the conversion. Next to the TOA data, water-leaving
radiances were matched and plotted across time. Following studies included
deriving MODIS detector and mirror side dependencies in the TOA data in
comparison with SeaWiFS, and TOA and water-leaving radiance correlations
with the geometry, including solar zenith angle.
The research on the data set is being described in a paper and there is a
web site containing some of the results:
http://seawifs.gsfc.nasa.gov (/staff/ewa/TOAmatch/TOAintro.html)

Level-2 data validation
A multi-sensor L2-data validation capability was developed which
complemented the standard L3 time series matchups. L2 pixels are matched
against averaged, multi-day, L3 bins where the L3 data can originate from
the same sensor as the L2 or other sensors. The method proved useful in
estimating impacts of different BRDF corrections, MODIS RVS,
detector-to-detector variabilities, and processing parameters, such as
cloud-albedo threshold and sun-glint coefficient. The validation software
is currently being ported to OBPG.

Bryan Franz
MSl12 Development:
A new version of MSl12 (v5.0) is in testing. This update includes major
changes
such as a complete rewrite of the atmospheric correction code (in C) and
reformatted aerosol and Rayleigh tables. All aerosol model information has
been incorporated into a single HDF file per model, and the model suite can
now be specified at run-time. The code has been expanded to better support
MODIS sensor capabilities, including the fluorescence bands, and many of
the derived product algorithms have been generalized for all supported
sensors. Significant work was done on the Carder and GSM01 bio-optical
models and the 3-band calcite algorithm. We have also added support for the
2-band calcite algorithm to facilitate evaluation of the competing calcite
approaches. MSl12 v5.0 includes full implementation of the fluorescence
line height and related algorithms for MODIS, including ARP (absorbed
radiation by phytoplankton) and CFE (chloropyll
fluorescence efficiency). An algorithm for IPAR (instantaneous
photosynthetically available radiation), based in part on the Carder (ATBD
20) algorithm, was also implemented for all supported sensors. Substantial
effort was made to verify the IPAR, ARP, and CFE implementations.

MODIS/Aqua Calibration and Algorithm Evaluations:
http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_testing.html)
Level-1B LUT changes
Two alternative LUTs were evaluated. The first, test ID AT21, was an update
by MCST which incorporated the reanalysis of the lunar calibration
time-series (which alters the cross-scan dependence and thereby the
temporal dependence away from the solar diffuser AOI). The second, test ID
AT20, was a LUT created by the OCDPG which attempted to first remove the
apparent beta-angle correlation for the solar diffuser measurements and
then refit the solar diffuser trends to a seawifs-like exponential (see
details from Gerhard Meister). The results from AT20 indicate that
significant improvement in the predictive capabilities of the derived SD
trends could be obtained by first removing the beta-angle
correllations. Compare these global trends, for example:
http://seawifs.gsfc.nasa.gov(/staff/franz/l3trend/modisa_seawifs_at18_st06/comp_global_nlw_mission.html)

http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_seawifs_at20_st06/comp_global_nlw_mission.html)

MCST has since identified the beta-angle correlation as a
detector-dependent vignetting effect associated with the diffuser screen
(based on re-analysis of the screen-up vs screen-down data from the yaw
manuever), and a new detector-specific (though still band averaged)
vignetting function is in development for future evaluation. The beta-angle
correlations, however, suggest that the problem is also band-dependent, due
to the relative position of the bands on the focal plane (i.e., a geometry
effect, not a spectral effect). We are also working with MCST to develop a
band-specific vignetting function, if possible, but the information is
limited due to the fact that the ocean bands saturate when the diffuser
screen is raised.

Straylight effects
Based on information gathered from pre-launch measurements and models (see
details from Gerhard Meister), it has become apparent that the MODIS design
is susceptable to substantial straylight leakage from neighboring bright
sources. Since we do not at this time have any method to correct for this
effect, we have attempted to mask the pixels adjacent to bright targets.
Using the filter capabilities of MSl12, we have simply dilated the cloud
and HILT (saturation) flags using a 5x3 dilation kernal (i.e., flag +/-
2-pixels along scan and +/- 1 pixel along track from any pixel which
experiences saturation in one of the ocean bands or exceeds the cloud
threshold in the 869nm channel). This was test ID AT22. With the exception
of some improvement in the 412nm band at high solar zenith, the impact to
temporal trends relative to SeaWiFS was not significant, but the additional
masking substantially reduced the long-standing
discrepancy between SeaWiFS and MODIS aerosol optical thickness retrievals.
Compare:
http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_seawifs_at18_st06/comp_global_aot_mission.html)

http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_seawifs_at22_st06/comp_global_aot_mission.html)

This result is very encouraging, as the AOT discrepancy was very puzzling
given that we use the same cloud screening and thresholds for both sensors.
It should also be noted that SeaWiFS already incorporated a similar level
of straylight masking, as well as a straylight correction beyond the mask.
We are looking at possible approachs to develop a straylight correctioon
for MODIS. We will also try a more agressive masking (7x5) and repeat the
temporal evaluations. Of course, the downside of masking is the reduction
in coverage, which becomes significant in the persistently cloudy, high
latitude regions.
c) Temporal change in standard products The current standard production
parameters are equivalent to test ID AT11. In the most recent round of
testing, it has become apparent that the MODIS products are now drifting
significantly away from SeaWiFS, so a reprocessing or, at the least, a LUT
update will need to be considered in the near future.

Striping effects
The change to detector-dependent corrections for the solar diffuser
vignetting function should contribute to reduced striping, and we will be
evaluating that in the next month. We are also looking at the variation in
the relative spectral response between detectors, which may indicate the
need for detector-specific, band-averaged solar irradiances and Rayleigh
optical thicknesses. There is a particularly large difference in the RSR of
the 1st detector of the 412nm band.

Jeremy Werdell (jeremy@simbios.gsfc.nasa.gov)
SeaBASS
SeaBASS data submission statistics for 2004 were provided to NASA HQ. A
collaboration between the NASA OCDPG and CHORS (San Diego State University)
has reinitiated the submission of overdue SIMBIOS-era HPLC-derived pigment
data to SeaBASS. Details regarding these new, globally-distributed data are
available online at:

http://seabass.gsfc.nasa.gov (/cgi-bin/seabass_news.cgi)

Algorithm evaluation
In preparation for the addition of new data products to operational
MODIS/Aqua data processing, and the assemblage of OCRT Working Groups, a
large scale review of existing semi-analytic reflectance models was
initiated (for example, that of Garver and Siegel 1997 and Maritorena et
al. 2002, informally referred to as GSM01). Analyses and results will be
posted online in the near future.

Sean Bailey (sbailey@seawifs.gsfc.nasa.gov)
Relative Spectral Response:
The RSR spectra and band pass averaged quantities for the sensors
msl12 is capable of processing were compiled and published at
oceancolor.gsfc.nasa.gov/RSR_tables.html.
All RSRs were put into a common format, covering the wavelength range of 380
-
1100nm in 1nm increments. The source data for the band averaged quantities
were also put in this format. Both the RSRs and the band averaged source data
are posted.

Glint Estimate:
The effect of switching the glint probability distribution from the Cox-Munk
to the Ebuchi-Kizu function was investigated. It was found that the end
result on derived water-leaving radiances was minimal. An investigation into an
alternative glint estimate based on measured radiances in the NIR was begun.

SeaDAS Support Group
The SeaDAS Development Group continues to make good progress in its efforts
to support the Ocean Color Community. They recently released support for
the Fedora Core 1 and 2 operating system platforms, and are getting close
to releasing support for the Mac OS-X operating system. Preliminary testing
on the Mac port has shown identical results for all but one of the
processing programs, and debugging continues. Mirror sites for the
distribution of SeaDAS are also in the works, with sites in Australia and
the UK almost ready to go online.