Ocean Color Reprocessing 2009The OBPG is working toward a reprocessing of all ocean color missions currently supported, including SeaWiFS, MODIS on Aqua and Terra, OCTS, and CZCS. Over the past year, the team has been working to identify and evaluate potential processing changes. These changes include sensor-specific calibration and characterization updates, sensor independent algorithm revisions, updated ancillary data sources, and product format and content changes. Major algorithm changes include new aerosol models and model selection methodology, new Rayleigh tables, updated out-of-band corrections and band-averaged absorption and scattering coefficients, additional correction for NO2 absorption, and updates to the standard Kd and chlorophyll algorithm coefficients. Description of these changes are listed below. Additional details and results will be added as they become available.
Reprocessing of SeaWiFS is complete. The ocean color products from SeaWiFS Reprocessing 2007 (a.k.a., R5.2) are no longer available. SeaWiFS R2009 global ocean color and land products and merged LAC (MLAC) files have been regenerated. Reprocessing of MODIS/Aqua is expected to occur sometime in late 2009. For comment or discussion on the reprocessing plans and results, please refer to the Reprocessing 2009 forum.
Calibration and characterization
SeaWiFS instrument calibration
The need for a change to the SeaWiFS radiometric/temperature calibration was indicated by both a downward trend in the green band nLw in recent years and a deterioration of the agreement between the current calibration and the lunar measurements. The revised calibration includes the following changes: 1) The counts to radiance conversion coefficients were reverted to the values measured by the instrument builder during the initial prelaunch calibration of SeaWiFS; 2) The reference temperature for the temperature corrections to all bands has been changed from 20 C to 16 C; 3) The temperature corrections for Bands 1 through 5 have reverted to the prelaunch measured values; 4) A single temperature epoch has been used for Bands 6 and 7, and two epochs have been used for Band 8 with the second epoch starting in mid-2005; 5) The radiometric calibration function has been changed from dual exponentials in time to a single exponential in time plus a linear term; and 6) The Gain 3 drift correction has been revised and applied to Band 7 alone.view additional details
MODIS Aqua instrument calibration
MODIS Aqua suffered from a clearly errneous downward trend in the global average 412nm water-leaving radiances. The scan angle dependence of the 412nm water-leaving radiances showed large variations towards the end of the mission. Furthermore, the global average of the fluorescence line height product showed an unrealistic downward trend over the mission. The water-leaving radiances of the red bands decreased strongly with increasing scan angle. Two new approaches largely removed these deficiencies: 1) The MODIS Aqua bands were vicariously calibrated using SeaWiFS as a truth sensor. The full correction is applied to the 412nm and the 443nm band, whereas only the scan angle dependency is applied to the ocean color bands from 488nm to 678nm. 2) The temporal variation of the response versus scan for the ocean color bands from 667nm to 870nm is modeled with a new algorithm (developed by MCST) that uses nonsaturated pixels of the lunar measurements.view additional details
The Marine Optical BuoY (MOBY) has served as the source for the in situ data used in the vicarious calibration of the visible bands. Since the last reprocessing, in addition to data from more buoy deployments, the OBPG has obtained the full-resolution radiometric data for MOBY and used these data to compute the water-leaving radiances. Previously, the water-leaving radiances produced by the MOBY Operations Team were used. Differences in the water-leaving radiances produced by the MOT and OBPG are small. The benefit of the OBPG doing the processing is to allow a finer grained control of the QA/QC for the data used.
The primary assumption with the NIR calibration method employed by the OBPG ( Franz et al. 2007) is that the aerosol type is known and constant for the site(s) used to calibrate the NIR bands. An examination of the temporal stability of the AERONET-derived aerosol Angstrom exponent for the AERONET stations closest to the two NIR calibration regions revealed that the South Indian Ocean site had a distinct seasonality, whereas the South Pacific Gyre site exhibited no obvious seasonality. On the basis of this, we chose to limit the NIR calibration to the South Pacific Gyre site and to target the median Angstrom derived from the AERONET station on the island of Tahiti (~0.65). In past reprocessings, we had assumed a typical maritime aerosol with Angstrom of ~0.2.
The sources and methods for producing band-averaged atmospheric and in-water absorption and scattering coefficients have been updated for traceability and consistency across all sensors. This includes Rayleigh optical thicknesses, ozone absorption coefficients, NO2 absorption coefficients, and pure-water absorption and scattering coefficients.view additional details
The Rayleigh tables were recomputed for all sensors using the vector radiative tranfer code of Ahmad and Fraser (AF), as recently updated by Z. Ahmad. This was done to ensure consistency in the Rayleigh tables between sensors and to ensure consistency with the new aerosol models (see below). The AF code was also used for the MODIS Atmosphere products, so consistency across disiplines is an added benefit. The AF computations used the Rayleigh optical thicknesses and spectrally-dependent depolarization factors described above.view additional details
Aerosol models and model selection method
A new suite of aerosol models have been developed based on knowledge gained from many years of AERONET observations. The new models are based on observed size distributions and single-scattering albedos from AeroNET sites on various coastal and island locations. Radiative transfer computations were performed with the AF code described above, and 80 models were generated spanning 10 fine-to-coarse-mode size fractions and 8 relative humidities (RH). The model selection process was also adapted to utilize RH observations to limit the number of models to be considered.view additional details
NIR water-leaving radiance correction
The correction for water-leaving radiance in the NIR bands, as employed in past reprocessings, was inconsist in the inversion of remote sensing reflectance (Rrs) to absorption (a) and backscattering (bb) terms and the subsequent estimation for Rrs from a and bb. A revision to this algorithm was devised which now consistently handles this forward and inverse translation. In addition, a revised spectral dependence on backscattering was implemented, whereby the wavelength dependent scattering estimate described in Gould et al. was replaced with the spectral backscattering estimate described in Lee et al. These updates significantly improved bio-optical retrievals in eutrophic waters, and substantially reduced the occurrence of negative water-leaving radiance retrievals.view additional details
Corrections were added to account for absorption by nitrogen dioxide (NO2), using the approach described in Ahmad et al. 2007 and a time-series of tropospheric and stratopheric NO2 retrievals.view additional details
Out-of-band Lw correction
A revised ocean reflectance model (ORM) was used to update the out-of-band corrections for water-leaving radiances. The ORM was updated to the version of Morel and Maritorena (2001) that is described in Werdell et al. (2007). In addition, the MODIS 551nm band was re-defined to be 547nm, greatly reducing the need for any out-of-band correction to this band.
The current whitecap correction uses a fractional coverage model that ramps up rapidly with increasing wind speeds. To prevent a severe over estimation of whitecap radiance, a threshold on wind speed was introduced (8 m/s) with reprocessing 2, along with a multiplicative reduction factor (0.4). The result is a nearly constant whitecap correction. A revised whitecap correction has been developed that uses a fractional coverage model that is less prone to over estimation. The wind speed threshold was increased to 12 m/s and the multiplicative reduction factor was eliminatedview additional details
Chlorophyll and Kd algorithms
The operational algorithms for deriving near-surface chlorophyll-a concentrations (OC; O'Reilly et al. 2000) and diffuse attenuation coefficients of downwelling irradiance at 490-nm (KD; Mueller 2000) were updated using in situ data from NOMAD version 2. NOMAD is a publicly available, global bio-optical data set constructed from data archived in SeaBASS. The data span a wide range of water types from coastal and offshore regions and were collected by participants in the NASA Ocean Biology & Biogeochemistry Program and by voluntary and international collaborators. Both algorithm forms describe the polynomial best fit that relates the log-transformed geophysical variable to a log-transformed ratio of remote-sensing reflectances. The polynomial form is consistent with past versions of OC, but replaces the power-law form employed by past versions of KD.view additional OC-related details view additional KD-related details
In Addition to Chlorophyll and Kd, a number of additional in-water derived products will be included in the standard Level-2
ocean color products suite. These include:
Ancillary data sources
The NCEP meteorological data used for the processing has been enhanced to include a stable surface relative humidity field, which will be needed for the new aerosol model selection algorithm. The primary ozone ancilary data source will be changed to use OMI ozone instead of TOAST ozone starting in 2006. NO2 data fields will be introduced into the operational processing in the form of monthly climatologies, to facilitate the new NO2 corrections. The sea ice flag will now be based on contemporaneous data from the NSIDC, which imposes an additional ancillary requirement on operational processing. Finally, ancillary file names have been standardized.view additional details
Masks and Flags
The coccolothophore flagging algorithm was re-tuned to improve detection. The sea ice flag, which was previously set using a fixed climatology, will now be set using co-incident data from NSIDC. Masking of high glint for SeaWiFS, which was disabled in a previous reprocessing, has been re-enabled for consistency with MODIS. The effect is small, as high glint is generally masked as cloud or high radiance. Similarly, the SeaWiFS straylight masking of Level-3 has been increased around bright targets. This has the effect of reducing bias in oligotrophic chlorophyll retrievals, and reducing speckling noise in the SeaWiFS Level-3 products, at the expense of coverage. Finally, a number of standard Level-2 flags have been eliminated, as they are not currently used for any operational masking or quality screening.view additional details
Level-1 file changes
The Level-1A files for MODIS will now be unaltered, thus containing the full band suite including higher-resolution land bands. This enables support of new products, but files will be larger.view additional details
Level-2 file changes
For the standard Level-2 ocean color files, some parameters were renamed, some were dropped, others were added. Of particular note, the nLw products have been superceded by reflectance products (Rrs). In the case of MODIS, the reflectance at 551 is now reported at 547, to reflect the true band center and associated change in out-of-band correction. The 678nm product was also added to MODIS. The fill-values have been standardized across products to simplify the identification of masked or failed retrievals. Meta-data attributes were added to each field to indicate the fill value. The longitude and latitude datasets are now provided at full resolution.view additional details
Level-3 bin file changes
The standard Level-3 ocean color products have been expanded to reflect the new Level-2 content. The binned files have been partitioned into product suites, many of which now contain only one product. This is a departure from the past, wherein all standard products were contained within the same logical or physical bin file.view additional details
Level-3 map file changes
For every product in the OC product suite, for every binning period, there is now a corresponding Level-3 global mapped product. These products now contain gridded floating-point arrays, rather than scaled integers.view additional details
Relative to previous reprocessings, the R2009 results show improved agreement in ocean color products between sensors and improved agreement in aerosol type and optical thickness retrievals relative to AERONET and relative to independent microtops measurements. Substantially improved agreement with in situ chlorophyll measurements in turbid and highly productive waters has also been demonstrated, and some instrumental instabilites in SeaWiFS and MODISA retrievals have been rectified.view additional details