SeaWiFS Reprocessing #2 Results

As promised in our August announcement of the global reprocessing, we are posting more detailed information about the analyses that went into the current reprocessing of SeaWiFS data.

Specifically, the information includes the original reprocessing announcement, the article we published in the September issue of Sea Technology and the following three suites of analysis results:

  • Changes in Level-2 Processing: Wayne Robinson
  • SeaWiFS Match-up Results: Brian Schieber
  • SeaWiFS Calibration and Match-up Results (with global clear-water analyses): Gene Eplee - updated 23 November 1998

We hope this information helps you to utilize SeaWiFS data in your research and in understanding the process by which we have tried to improve the data products. Feel free to download this information. All we ask is that you acknowledge the SeaWiFS Project Office as the source.

A Proposed Analysis of the Time-Dependent Changes in the Radiometric Sensitivities of SeaWiFS Bands 1 through 6

by Bob Barnes

Download the paper, figures and tables in PDF format.

Changes in the Level-2 Processing

by Wayne Robinson

  • On average, changes have been made at the rate of 1 change every 10 days since last reprocessing started (2 Jan 98).

  • Major changes were:
    • better tests to reduce rogue chlorophyll points
    • new transmittance tables to reduce nLw variation across scan line
    • better computation of nLw, using the atmospheric transmittance instead of Rayleigh alone
    • Improved, noise resistant dark count calculation
    • more accurate F0 calculation used
    • use a tighter grid of anchor points, more accurate parameter derivation
    • fix calibration algorithm to compute the gains correctly
    • reduce whitecap radiance effect
    • change code to allow more high chlorophyll data to be binned in the level-3 product
    • include a wider range of aerosol models -- new oceanic aerosol models added
    • write out the actual epsilon derived in the aerosol correction algorithm for epsilon <> model range
    • new chlorophyll algorithm resulting in increased low chlorophyll (<.03 ( and chlorophyll decreased high m^3) mg>1. mg / m^3)

Figures

8-day tbin new software
This is an image of the 8-day timebin of SeaWiFS data for the period of March 14 - 21, 1998 using the new processing. The largest change that can be seen in the new reprocessing is that the low chlorophylls in the open-ocean areas have been significantly increased. New areas of retrieved chlorophyll (especially in the Baltic Sea) are evident, contributing to the 5% increase in the total # of bins containing data.
8-day tbin old software
This is an image of the 8-day timebin of SeaWiFS data for the period of March 14 - 21, 1998 using the old software (used in the 1st reprocessing). Many regions of open ocean had chlorophyll of .01 -- the lowest chlorophyll that can be retrieved and an unreasonable value for such a large area.
old / new chlor histograms
This is a histogram of the chlorophyll for the timebin of March 14 - 21, 1998. The dotted line is for the old processing while the solid line is for the new processing. It clearly shows the increase in low values that has resulted from the changes.
old / new chlor scatter
This is a scatter diagram showing a bin-by-bin comparison of the chlorophyll using the old (x axis) and new (y axis) processing for the 8-day timebin of March 14 - 21, 1998. In addition to the increase of low-chlorophyll bins is a general decrease of the high chlorophyll values.
Gulf and Florida band 8 and new chlor image
These 2 images show the band 8 and true color images for the Gulf of Mexico on May 20, 1998 at 1803 UTC. 2 areas of thin or sub-pixel clouds and bright aerosols can be seen in the top-left and botton-center of the image. These areas caused the cloud mask to be turned on in the chlorophyll data. The effect became more severe during the 2nd reprocessing as a result of the correction of band 8 for degradation.
Gulf and Florida new/old chlor image
These 2 images show an example of the chlorophyll generated using the old and new processing for the Gulf of Mexico on May 20, 1998 at 1803 UTC. The low chlorophylls of around .02 or .03 in the Gulf (magenta) have increased to .10 and high chlorophylls closer to the coast have decreased as a result of the new processing.
old / new nlw_443 histograms
This is a histogram of the 443 nm water leaving radiance for the timebin of March 14 - 21, 1998. The dotted line is for the old processing while the solid line is for the new processing. A small, .02 mw / etc increase of the nLw can be seen.
old / new nlw_443 scatter
This is a scatter diagram showing a bin-by-bin comparison of the 443 nm water leaving radiance using the old (x axis) and new (y axis) processing for the 8-day timebin of March 14 - 21, 1998.
old / new nlw_555 scatter
This is a histogram of the 555 nm water leaving radiance for the timebin of March 14 - 21, 1998. The dotted line is for the old processing while the solid line is for the new processing. A significant, .06 mw / etc increase of the nLw is seen. In addition to decreasing the occurence of negative water-leaving radiance, the new processing has resulted in a more reasonable value of the nLw in this band.
old / new nlw_555 scatter
This is a scatter diagram showing a bin-by-bin comparison of the 555 nm water leaving radiance using the old (x axis) and new (y axis) processing for the 8-day timebin of March 14 - 21, 1998.
before/after change to whitecap algorithm
These 2 images show an example of the change that happened when the magnitude of the whitecap correction was reduced by 75%. This region off the West coast of South Africa was very clear in the visible but a large area was masked out (black central area) because the radiance in bands 6, 7 and 8 were going negative after the rayleigh radiance was subtracted. It was found that the winds were about 15 m s-1 in this area and the whitecap radiance was a significant fraction of the total radiance. When the whitecap correction was decreased, chlorophyll retrievals were possible again.

SeaWiFS Matchup Results

by Brian D. Schieber

MOBY Matchups
This figure shows the matchup comparisons between in_situ optical measurements of water-leaving radiance (Lw) and the corresponding value as determined by SeaWiFS for the Marine Optical BuoY (MOBY) site. One-to-one (solid) and regression (dotted) lines are shown in each window. Regression statisics are shown for the first panel (Lw412) only.
Water-leaving radiance (Lw) Matchups
This figure shows the matchup comparisons between in_situ optical measurements of water-leaving radiance (Lw) and the corresponding value as determined by SeaWiFS. The individual experiments are listed by name in the first panel and referenced by number in all four panels. One-to-one (solid) and regression (dotted) lines are shown in each window. Regression statisics are shown for the first panel (Lw412) only.
Chlorophyll Matchups
Correlation of in_situ measured chlorophyll (fluorometric or HPLC method) versus the algorithm OC2v2 (Ocean Color 2, version 2) applied to SeaWiFS radiances. The one-to-one (solid) and data regression (dotted) lines are shown.

SeaWiFS Calibration and Matchup Results

by Gene Eplee

updated 23 November 1998

Plot of total lunar radiance versus time (November 1997 - November 1998)
The integrated lunar radiances for each band are normalized to a common observing geometry (a moon-sun distance of 1 A.U., a sensor-moon distance of the semi-major axis of the moon's orbit, a phase angle of 7 degrees) and to the radiances at the first lunar calibration. In order to reduce the scatter in the data due to an imperfect normalization over viewing geometry, the total radiances have also been normalized to the values in band 5. The resulting plot shows the degradation in the response of bands 7 and 8. Two degradation regimes are evident: a regime from September 1997 through July 1998 where the response in band 7 decreased by ~1.5% and the response in band 8 decreased by ~4.5%, and a regime since July 1998 where the rate of degradation has decreased substantially. Piecewise linear functions have been fit to the band 7 data for each of these regimes to compute the temproal correction to band 7 for the current vicarious calibration.
Plot of solar calibrations versus time
The solar radiances for each band are normalized to a common earth-sun distance of 1 A.U., corrected for seasonal variations in the position of the solar image on the diffuser, and normalized to the radiances at the first solar calibration. Two trends are apparent in these plots: a degradation in the reflectance of the diffuser with time as photolyzed organic materials, outgassed from the spacecraft, condense onto the surface of the diffuser and the degradation in the response of bands 7 and 8.
Plot of solar calibration band ratios versus time
The band pair ratios show that the degradation of the diffuse reflectance decreases with increasing wavelength. The degradation of the response of band 8 with respect to band 7 is apparent in the band 7/band 8 ratio. This ratio, when corrected for the time drift in band 7, is used to compute the temporal correction to band 8 for the current vicarious calibration. Six piecewise linear time corrections are computed for band 8.
Plot of time-corrected solar calibrations versus time
The normalized solar radiances have been corrected for the time drifts in bands 7 and 8. The remaining trend in the data is due to the degradation of the diffuser reflectance.
Plot of time-corrected solar calibration band ratios versus time
The data for bands 7 and 8 have been corrected for their respective time drifts. The corrected band 7/band 8 ratio is essentially constant with time.
Plot of SeaWiFS/MOBY matchups versus scan angle
The vicarious calibration adjusts the system gain of SeaWiFS to maximize the agreement between water-leaving radiances measured by SeaWiFS and MOBY for band 1-6 over the 53 matchup data sets. There are not any trends in the matchups, which indicate that the atmospheric correction algorithm does not have any residual effects as a function of scan angle.
Plot of SeaWiFS/MOBY matchups versus time
There are not any trends in the SeaWiFS/MOBY matchups, which indicate that the time corrections for bands 7 and 8 are working as expected.
Results of the SeaWiFS Vicarious Calibration
The table shows that SeaWiFS and MOBY radiances which results from the vicarious calibration, along with vicarious gain for each band that has been incorporated into current calibration table.
Plot of mean global clear water radiances versus time for bands 1-5
Clear water is water with a depth greater than one km and with a chlorophyll concentration of less than 0.15 mg/m^3. The filled circles and asterisks represent the mean radiances computed for the current (August 1998) reprocessing. The solid lines represent the mean radiances computed for the February 1998 reprocessing. Some of the improvements in the data quality resulting from the current reprocessing are evident in comparing the two sets of time series.
Plot of mean global clear water epsilons versus time
The filled circles represent the mean epsilons computed for the current (August 1998) reprocessing. The solid line represents the mean epsilons computed for the February 1998 reprocessing. The trend with time in the February data set has been removed from the current data set by the temporal corrections to bands 7 and 8.
Plot of mean global aerosol optical depths at 865 nm for clear water versus time
The filled circles represent the mean optical depths computed for the current (August 1998) reprocessing. The solid line represents the mean optical depths computed for the February 1998 reprocessing. The decrease in optical depth with time in the February data set has been removed from the current data set by the temporal corrections to bands 7 and 8.
Plot of mean global clear water chlorophyll versus time
The filled circles represent the mean chlorophyll computed for the current (August 1998) reprocessing. The solid line represents the mean chlorophyll computed for the February 1998 reprocessing. The decrease in chlorophyll with time in the February data set has been removed from the current data set by the temporal corrections to bands 7 and 8.
Results of the SeaWiFS Clear Water Analysis
The table compares global, clear water, normalized water-leaving radiances computed for SeaWiFS with mean normalized water-leaving radiances measured by MOBY.