New SeaWiFS Calibration
The ambient temperatures at which the SeaWiFS focal planes operate are affected by the solar insolation on the spacecraft. Examination of the SeaWiFS lunar calibration time series shows an annual periodicity in some bands which corresponds to the variation in focal plane temperatures as the Earth-sun distance changes over the course of a year. The SeaWiFS counts-to-radiance conversion incorporates corrections for variations in the radiometric response of the instrument with changing focal plane temperatures. The prelaunch corrections do not fully account for the variation in radiometric response with changing focal plane temperatures observed over the course of a year. The SeaWiFS CalVal Team has used the periodicities in the lunar time series to compute revised focal plane temperature correction factors for the SeaWiFS counts-to-radiance conversion. These revised temperature correction factors have been used to reprocess all of the lunar calibration data. The resulting time series no longer show the annual periodicities.
The revised lunar calibration time series has been fit with a single decaying exponential function of time for each band. These fits have been incorporated into the calibration table which will provide the time correction for the fourth reprocessing of the SeaWiFS mission data set. The new calibration table must be run with the updated version of MSl12 which incorporates the exponential form of the time corrections.
The revised lunar calibration is discussed in detail in the December 2001 issue of Applied Optics in Barnes et al. , "Calibration of SeaWiFS. I. Direct techniques."
For one analysis of a Repro #4 mission test, mean global Lwn in bands 1-6 were computed from the 8-day timebin files for clear water (water with a depth of at least 1 km and with a chlorophyll concentration of less than 0.15 mg m^(-3)). The mean radiances in clear water should be constant with time if the radiometric calibration of the instrument is stable. This analysis showed decreases in the radiances for bands 1-6 of several percenet over the course of the mission. Figure 1 and figure 2 show the clear water time series for bands 3and 5 in red. Band 3 exhibited a change of -2% over 1000 days and -3.4% over the mission (a rate of change of -1.9e-5 per day). Band 5 exhibited a change of -2.6% over 1000 days and -4.4% over the course of the mission (a rate of change of -2.5e-5 per day). These results imply that there is a small residual decrease in the radiometric response of SeaWiFS with time that is not corrected by the current lunar calibration-based time correction of the SeaWiFS data. With 4 1/2 years of data available, small residual drifts can become significant.
The SeaWiFS CalVal Team, in analyzing the lunar data to derive the corrections for the instrument, references the lunar calibration time series to the mean of the observations for bands 3 and 4. This normalization is done to reduce the systematic noise in the lunar observations which arises from our lack of understanding of how lunar libration affects the long-term measurements of the moon. The initial assumption of the CalVal Team was that the mean of bands 3 and 4 did not chage with time. The clear water analysis shows that this assumption is incorrect.
Recently, a comparison of the first 50 SeaWiFS lunar measurments with the newly completed lunar model of Hugh Kieffer (USGS, Flagstaff, Arizona) has verified that bands 3 and 4 have small drifts with time. The comparisons show that, on average, the radiometric responses of bands 3 and 5 have been decreasing at a rate of 0.35% per 1000 days or 0.60% over the course of the mission. The effect of this time drift on the Lwn would be an order of magnitude larger.
To correct the residual drift in the SeaWiFS time correction, the CalVal team has applied a time correction of -3.5e-6 per day the the mean of bands 3 and 4 before the mean is used to normalize the lunar time series. The overall time corrections for bands 1-8 have been recomputed and the vicarious calibration of SeaWiFS has been run again.
The Repro #4 mission test was rerun with the updated time correction and the 8-day timebin clear water analysis was rerun. This analysis showed no significant changes in bands 1-6 over the course of the mission. Figure 1 and figure 2 show the updated clear water time series for bands 3 and 5 in blue. Band 3 exhibits a change of +0.16% over 1000 days and +0.27% over the course of the mission (a rate of change of +1.5e-6 per day). Band 5 exhibits a change of -0.38% over 1000 days and -0.64% over the course of the mission (a rate of change of -3.5e-6 per day). Any residual time drift in the instrument calibration has been reduced by a factor of 8-10.
As part of an ongoing intra-agency collaboration, the MOBY Project has contracted with NIST to perform a set of instrument characterization measurements for stray light within the MOBY spectrometers. At this time, the MOBY Project plans to reprocess the MOBY time series. The results of this reprocessing for approximately half of the existing MOBY data were made available at the end of December 2001, with the remainder expected in February 2002. The analysis of the data provided to date has resulted in revised vicarious gains, which provide water-leaving spectral radiances that are higher for SeaWiFS bands 1 and 2 (412 and 443 nm nominal center wavelengths) and are lower for SeaWiFS band 5 (555 nm nominal center wavelength), compared with the current MOBY results.
The currently available set of stray light corrected MOBY data has produced 163 points used in the calibration process. The corrections were applied using a second generation model derived by Steve Brown and Carol Johnson of NIST. Future refinements of this model and the addition of the remainer of the MOBY deployments will be incorporated as they become available, but in all probability should not significantly alter the results.
MOBY derived Lw values are computed by propagating the Lu measurements made at depth to the surface using Kl. Measurements of Lw computed from the Lu measurement at top arm (1.5m) have been compared to those computed from the Lu measurement at the middle arm (5m). The majority of these measurement pairs are in good agreement. However, divergence between the two Lw's on the order of 5 - 30% can occur. These are likely the result of unfavorable sky and/or sea conditions. As an additional exclusion criterion, the CVT has implemented the following:
- Data from both the top and middle arms must exist.
- The combined uncertainty in the Lw measurements for all bands from both arms must be less than 10%, as determined by the square root of the sum of the squared absolute percent differences for bands 1-5, corrected for the cosine of the solar zenith angle.
After applying this additional criteria, the total number of MOBY matchup points used in the calibration was reduced to 45. Currently, straylight corrected MOBY data are only available for both arms for deployments 212 - 219, Feb 2000 to present (April 2002).
Significant changes to the vicarious calibration procedure for SeaWiFS have been implemented for Reprocessing #4. These changes are summarized below.
The SeaWiFS matchup data consist of 101x101 pixel extracted scenes centered on the MOBY location. The size of these extracts ensures that stray light will be handled properly in the vicinity of MOBY. The analysis for each matchup is performed on 5x5 pixel subscenes centered on the MOBY location. The CalVal team checks each 5x5 pixel subscenes for data quality, screening out matchups for clouds and cloud shadows, stray light, sun glint, high satellite and solar zenith angles, radiances above the knee in bands 7 and 8, and aerosol optical depths greater than 0.1. All 25 pixels in a subscene are required to be clear for the matchup to be used in the vicarious calibration. As discussed above, the CalVal team has identified 45 SeaWiFS/MOBY matchups based on MOBY data quality. The quality constraints imposed on the SeaWiFS data has reduced the number of matchups to 23. The vicarious calibration used in Reprocessing #4 is based on 23 SeaWiFS/MOBY matchups selected from an original pool of 156 potential matchups.
Forward Vicarious Calibration
The forward vicarious calibration procedure compares normalized water-leaving radiances (Lwn) between SeaWiFS and MOBY to determine the vicarioius gains for SeaWiFS bands 1-6. For a given matchup, the Lwn for SeaWiFS is averaged over the 5x5 pixel subscene and this average value is ratioed to the MOBY radiance for that matchup. The mean SeaWiFS/MOBY radiance ratio is computed over all of the individual matchups. The vicarious gain for each band is adjusted iteratively over a number of runs of MSl12 on the matchup data set until the mean SeaWiFS/MOBY radiance ratio converges to unity. The forward vicarious calibration procedure allows the effects of iterative near-infrared correction algorithms to be incorporated into the vicarious gains.
The forward vicarious calibration procedure, as implemented for Reprocessing #3, is discussed in detail in Eplee et al. (2001), "Calibration of SeaWiFS. II. Vicarious techniques," Applied Optics, Vol. 40, 6701-6718.
Inverse Vicarious Calibration
The inverse vicarious calibration procedure compares top-of-the-atmosphere (TOA) radiances between SeaWiFS and MOBY to determine the vicarious gains for SeaWiFS bands 1-6. A flow chart for this procedure is shown in the vicarious_calibration_flow.gif . For a given matchup, the atmospheric correction parameters retrieved for the SeaWiFS data are used to propagate the MOBY-measured water-leaving radiances to the top of the atmosphere. The TOA radiances are averaged over the 5x5 pixel subscene for both the SeaWiFS and MOBY data. Vicarious gains for the matchup are computed from the ratios of the mean MOBY and SeaWiFS TOA radiances in each band. The overall vicarious gains are the means of the gains computed for the individual matchups.
The inverse vicarious calibration procedure generates gains which agree with those produced by the forward calibration to within 0.04% for the same calibration conditions. At the same time, the inverse calibration offers a number of advantages over the forward calibration. The inverse calibration runs more quickly than the forward calibration, since the gains are computed directly from a single run of MSl12 rather than interactively from a number of runs of MSl12. The inverse calibration allows the effect of an offset in the vicarious calibration to be investigated. Additionally, the inverse calibration preserves spectral shape information in the matchups by computing a set of gains for each matchup, while the forward calibration computes the gains from the mean of the matchup ratios, thus diluting the spectral shape information. The only drawback to the inverse calibration procedure is that it cannot incorporate the effects of iterative near-infrared correction algorithms. Consequently, the inverse calibration procedure can only be used when the near-infrared correction algorithms have been disabled or have been enabled with a ramp-in at low chlorophyll concentrations.
Calibration Changes during the Reprocessing #4 Algorithm Analysis
Throughout the algorithm analysis performed for Reprocessing #4, the CalVal team updated the vicarious calibration as specific changes were implemented in the processing which would impact the derivation of the vicarious gains. The updates to the calibration are summarized as follows.
Baseline - Current Operational Processing:
The vicarious gains for the operational processing has been in place since Reprocessing #3. The calibration was performed in the forward direction with the Siegel near-infrared correction algorithm enabled. The calibration used the operational calibration table with the piecewise quadratic time corrections. The calibration data set was comprised of 125 SeaWiFS/MOBY matchups spanning the mission through March 2000.
Case 1 - MOBY Stray Light Correction:
The first update to the calibration incorporated the straylight-corrected MOBY data, as discussed above. The calibration was performed in the forward direction with the Siegel near-infrared correction algorithm enabled, using the operational calibration table. The calibration data set was comprised of the 23 SeaWiFS/MOBY Reprocessing #4 matchups discussed above.
Case 2 - Updated Calibration Table:
The second update to the calibration incorproated the calibration table with the exponential time corrections and the revised focal plane temperature corrections. The calibration was performed in the forward direction with the Siegel near-infrared correction algorithm enabled, using the 23 Reprocessing #4 matchups.
Case 4 - Miscellaneous Changes:
The third update to the calibration incorporated the ramp-in of the near-infrared correction at low chlorophyll concentrations. The calibration was performed in the inverse direction with the near-infrared correction disabled, which is functionally equivalent at MOBY to having the ramp-in of the correction enabled. The calibration used the exponential calibration table and the 23 Reprocessing #4 matchups.
Case 9 - Reduction in Aerosol Model Selection Noise:
The fourth update to the calibration incorporated the epsilon-targeted smoothing of the band 7 radiances. The calibration was performed in the inverse direction with the near-infrared correction disabled, using the exponential calibration table and the 23 Reprocessing #4 matchups.
Case 10 - Add Fresnel Transmittance Correction:
The final update to the calibration incorporated the correction to the water-leaving radiances for the Fresnel transmittance through the water-atmosphere interface. The calibration was performed in the inverse direction with the near-infrared correction disabled, using the exponential calibration table and the 23 Reprocessing #4 matchups.
Vicarious Gains for Reprocessing #4
The calibration scenario for Case 10 is the scenario that will be used for Reprocessing #4. As was discussed above, the calibration data set is comprised of 23 SeaWiFS/MOBY matchups selected from an original pool of 156 potential matchups. Time series of the TOA gains for bands 1-6 are shown in Fig. 1-6. The 163 potential matchups in each band are shown in blue and the 23 matchups used in computing the overall vicarious gains are shown in red. The data quality screening appled to both the MOBY and SeaWiFS data has reduced the scatter in the gains used to computing the overall vicarious gains.
|Operational (repro#3) Gains w/MOBY correction||1.0182||1.0009||0.9698||0.9871||0.9948||0.9583||0.9460||1.000|
|Exponential Cal Table Gains||1.0130||0.9964||0.9629||0.9821||0.9913||0.9566||0.9380||1.000|