August 2004 Project Update

[genecarlfeldman]

Ocean Color Community:

As you know, former SeaWiFS/SIMBIOS Project group at NASA/GSFC was designated as the NASA ocean color discipline processing lead early this year and assumed the operational processing of MODIS Aqua data on February 1.  Since then, we have made a number of refinements in the MODIS Aqua calibration, data processing, and data distribution for the reduced MODIS product suite.  We have made formal reports on our status at the NASA Ocean Biogeochemistry Program annual meeting in April and at the MODIS team meeting in July and have posted the results of our analyses and documentation on the new OceanColor website at

http://oceancolor.gsfc.nasa.gov

specifically under the "Validation" and "Documentation" sections.  The Project has also sent occasional messages out to the community on various topics.  However, it has been suggested that we provide a more frequent and routine update to the community and we also believe is indeed necessary.  Therefore, we will provide a brief monthly report to you.  The update will include short summaries of various analyses and evaluations we have conducted in our attempts to improve the MODIS Aqua data quality, assist the algorithm working groups (AOP, chlorophyll, etc.) in testing algorithms for possible future incorporation into the operational processing, and other particularly noteworthy items.  We will provide links to analysis results and encourage you to review them and contact the Project staff member associated with each analysis if you have questions, comments, or suggestions for additional analyses. We also encourage people to make use of the online Ocean Color web forum ( http://oceancolor.gsfc.nasa.gov/ ) that has been quite helpful in keeping the ocean color community informed and for responding to questions and issues that have arisen.

To date, we have conducted 19 time series tests on the MODIS Aqua data that involve changes in the calibration, masking algorithms, etc.  These usually involve a reprocessing of both the SeaWiFS and MODIS data sets to ensure consistency.  The results of these tests are posted on a Project web site and a short summary of the objectives and conclusions from each is also posted there for your convenience.  Our usual approach is for the Project calibration and validation group to meet, usually on a weekly basis, to discuss issues, results, and future analysis suggestions with each analyst presenting their analysis with the overall group objective of being as scientifically critical as possible.  Also, the Project staff has also been working closely with the MODIS Characterization and Support Team (MCST) to refine the lunar and solar calibrations and provide the MCST feedback on possible refinements based on our analyses.

Chuck McClain

August 2004 Project Update
Highlights:  At the MODIS team meeting, we presented SeaWiFS-MODIS Aqua comparisons that indicated that there remained a seasonally varying difference (~ ±10%) in normalized water-leaving radiances (LwnâEUR(TM)s) at high latitudes and that the MODIS Aqua LwnâEUR(TM)s have a slight downward trend of about 3%/yr relative to SeaWiFS.  These two issues are the primary focus of our analyses.  We are also looking at stripping in the MODIS imagery, but have nothing specifically to report this month.  With regard to bio-optical algorithms, we have been working with Ken Carder (and staff) and Barney Balch on chlorophyll and calcite algorithm implementations, respectively.  The Carder algorithm will be used for NPP/VIIRS operational processing under the NPOESS program and our group is involved through the NPP science team in algorithm evaluations.  Since the MODIS team meeting, Bryan Franz has been modifying the MSL12 code to handle more than 8 bands so that the MODIS fluorescence line height bands can be processed.

Two of the action items that came of the team meeting have been addressed.  The first was a suggestion by Janet Campbell that a bias between the SeaWiFS and MODIS LwnâEUR(TM)s could produce the observed seasonal variation observed.  An analysis was completed that demonstrated to JanetâEUR(TM)s satisfaction that a bias was not present, a least to the degree that would generate the variation (item 6 below).  The other was a suggestion by Howard Gordon to mask one side of the Aqua data where the Rayleigh polarization is most pronounced to see if that removed the seasonal variation.  A Rayleigh âEURoedegree of polarizationâEUR flag was implemented and a test conducted (item 2 below).  One other pertinent test was conducted that compared SeaWiFS and MODIS Aqua total radiances (Lt) measured at nearly identical sun-sensor viewing geometries and times.  This occurs occasionally.  The results (item 5 below) basically show that without the MODIS polarization correction, the seasonally variation in Lt occurs in all bands.  With the polarization corrections applied, the seasonal variation occurs primarily in the NIR bands.  This would imply that the seasonal difference in LwnâEUR(TM)s at high latitudes could be an artifact of polarization sensitivity propagating through the atmospheric correction.  This mechanism needs to be further quantified.

Specific Reports:
1. Sean Bailey (sean@seawifs.gsfc.nasa.gov)
*Calcite ImplementationThe 3-band calcite algorithm was implemented within msl12.

http://oceancolor.gsfc.nasa.gov (/DOCS/ScienceTeam/Updates/Calcite_implementation.pdf)

*FLH, IPAR, ARP and CFE ImplementationWith recent modifications to MSl12 to allow processing of band 14 (678nm), FLH was implemented.  Efforts are underway to calibrate band 14.   Implementation of IPAR, ARP and CFE are currently being investigated.  The first cut at a table describing the tests we've run is here:

http://oceancolor.gsfc.nasa.gov (/cgi/aquatests.cgi)

2. Bryan Franz (franz@seawifs.gsfc.nasa.gov)
* Temporal Trend Analyses: Temporal trend analyses have continued, with focus on determining the source of seasonal differences between MODIS/Aqua and SeaWiFS which increase with latitude (solar zenith). Results of this testing are posted here:

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

The tests labeled AT16-AT19 were performed since the MODIS Science Team Meeting. We attempted to identify sources of contamination in the SeaWiFS or MODIS datasets which might produce differences between sensors as a function of solar zenith angle. The trends that show the problem most clearly are those derived by averaging data in latitudinal zones, e.g.:

http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_seawifs_at15_st06/comp_zonal_nlw_mission.html)

In test AT16 we tried masking all glint, using an ultra conservative threshold on the Cox & Munk glint coefficient. This reduced the discrepancy in aerosol optical thicknesses, but it had no significant effect on the nLw trends.
http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_seawifs_at16_st06/comp_zonal_nlw_mission.html)

In test AT17 we tried lowering the MODIS cloud threshold. Although we operationally use the same cloud algorithm for MODIS and SeaWiFS (surface + aerosol reflectance test at 869nm), there is a discrepancy between mean AOT of MODIS and SeaWiFS, e.g.:

http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_seawifs_at15_st06/comp_global_aot_mission.html)

which may be an indication that more clouds get through for MODIS. In AT17, we reduced the cloud threshold to force the AOTs to match,
http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_seawifs_at17_st06/comp_global_aot_mission.html)

but this also had no significant effect on the seasonal differences of the zonal nLw trends.

We added a new product to MSl12 which is the Rayleigh degree-of-polarization (dop), and created a new HIPOL Level-2 flag which is set when the dop exceeds a threshold. In test AT19 we tried masking all data with dop > 0.5 (50% polarized). This has the effect of masking most data above ~50-deg solar zenith and a significant fraction of the western scan edge. Still, enough data remains to derive trends and look for sensitivities. The resulting zonal nLw trends indicated an improvement in the behavior of the most highly polarization-sensitive 412-nm band. By removing the highly polarized observations from the means, the relative differences between MODIS and SeaWiFS at 412 now look more consistent with the relative trends in the 443 and 488 bands. However, the seasonal variability in the relative differences still remains.

http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_seawifs_at19_st06/comp_zonal_nlw_mission.html)

The conclusion from AT19 is that we still have some problems correcting the response to highly polarized light for the most sensitive band, but the connection between the seasonal differences between MODIS and SeaWiFS and polarization sensitivity remain unclear (note:  there is no correction for aerosol polarization in the atmospheric correction algorithm).

Test AT18 was a change to the binner which we needed to evaluate. As you know, it is possible to get negative nLw retrievals, and we don't necessarily want to mask the data simply because the 412 or 667 band went negative. For SeaWiFS we average negative nLw as zero. For MODIS we were averaging negative nLw "as is".  We changed the MODIS binning to set the lower-limit on nLw to zero. As expected, this had no other effect but to raise the mean nLw for MODIS at 412 in coastal waters. See:

http://seawifs.gsfc.nasa.gov (/staff/franz/l3trend/modisa_at15_at18/comp_global_nlw_mission.html)

Of course, the "right" thing to do with negative retrievals is up for debate. In my opinion it would be more correct to flag and mask extreme negatives, but allow some negatives to be included in the binned average. The distribution of clear ocean retrievals in the red, for example, can be expected to have a mean near zero and a lower tail which cross into negative retievals, and the present scheme used for SeaWiFS will skew that distribution and bias the Level-3 means. But the problem then becomes: how negative is too negative?*

New Products:
We now have the ability to produce fluorescence line height. The new version of MSL12 with this capability is being tested and debugged, and we are beginning to look at the calibration of the red bands.We will be adding ipar, arp,and cfe as well. In conjunction with this change, and to better support future sensors, the aerosol correction code was completely restructured and rewritten (actually in C). The revision will allow for greater flexibility and improved maintainability of the Level-2 processing code. Due to the extensive nature of the changes, considerable testing must still be done before the new MSL12 code can be put into production or released in SeaDAS.

3. Gene Eplee (eplee@seawifs.gsfc.nasa.gov)
* MODIS Aqua Lunar Data Analysis:We have begun a detailed analysis of the MODIS Aqua lunar calibration data for the 1-km reflective solar bands. Our initial approach is to sort the data for a given calibration into composite lunar images for each detector and mirror side. Our goal is to develop a SeaWiFS-style time series of disk-integrated lunar reflectances (for the bands which do not saturate when viewing the Moon) from which we hope to gain further insight into the radiometric stability of the instrument over time.

4. Gerhard Meister (meister@simbios.gsfc.nasa.gov)
* Polarization characterization: The only design difference between MODIS Aqua and MODIS Terra regarding the polarization sensitivity is the first dichroic. A comparison between the prelaunch polarization characterization measurements of Aqua and Terra shows significant differences between the two instruments http://oceancolor.gsfc.nasa.gov (/DOCS/ScienceTeam/Updates/aqtepol.html). It remains unclear whether these differences are due to the dichroic, due to measurement uncertainties, or other factors. The polarization ray-tracing group headed by Eugene Waluschka is expected to provide guidance regarding this question.  However, so far, the ray-tracing modeling and the prelaunch polarization characterization data have puzzling inconsistencies.

5. Ewa Kwiatkowska (ewa@simbios.gsfc.nasa.gov)
* TOA MODIS and SeaWiFS matchups Matchups are based on a spatially and temporally limited time-series composed of coincident pixels imaged by both sensors with very close viewing and solar geometries. Temporal distribution of average MODIS-to-SeaWiFS TOA-radiance ratios is calculated. To analyse the impact of MODIS polarization correction, ratios are computed before the correction is applied and after.  As mentioned above, the initial analysis has yielded some insight into MODIS polarization characteristics.

http://seawifs.gsfc.nasa.gov (/staff/ewa/TOAmatch/TOAintro.html)

Other activities:
A MODIS validation approach has been developed which matches pixel L2 products and parameters against averaged L3 data which can originate from either  MODIS or  SeaWiFS.  The method has been used to estimate the impact of different BRDF corrections, MODIS RVS, detector-to-detector variabilities, and processing parameters, such as cloud albedo threshold and glint coefficient.  Also, a procedure that matches MODIS TOA pixel radiances and the radiances predicted from SeaWiFS water-leaving radiance brought to the TOA using MODIS solar and viewing geometries and sensor-optional aerosol characteristics has been developed.  The algorithm has been applied to study MODIS polarization sensitivity, RVS, and other parameters and is used in MODIS and SeaWiFS vicarious calibration.

6. Fred Patt (fred@seawifs.gsfc.nasa.gov)
At the 2004 MODIS Science Team meeting, it was postulated that the seasonal cycle seen in the MODIS-Aqua/SeaWiFS nLw rations might be due to a calibration bias. An analysis was suggested to test this idea by plotting MODIS vs. SeaWiFS nLw values. This was performed with monthly SMI files from June and December 2003. The analysis showed that the variations are real, systematic effects and not a statistical artifact. The full analysis can be seen at:

http://oceancolor.gsfc.nasa.gov (/DOCS/ScienceTeam/MST_Jul2004/CalibrationBias.pdf)