MODIS-Aqua Reprocessing 2012.0
MODIS-Aqua Reprocessing 2012.0
For the previous reprocessings (R2010.0, R2009.1), the calibration of MODIS Aqua bands 8 and 9 (412nm and 443nm) was modified by the OBPG to adjust the temporal trends in the response versus scan angles (RVS). These temporal adjustments, which could not be fully characterized by the on-board (lunar, solar) calibration, were derived by cross-calibration with SeaWiFS (Meister et al. 2011). Unfortunately, the SeaWiFS mission ended in late 2010, due to a terminal spacecraft anomaly. The purpose of MODIS-Aqua Reprocessing 2012.0 is to implement a new instrument calibration that is fully independent of the SeaWiFS mission.
2.1. Instrument Calibration
The basis of this reprocessing is the MODIS Calibration Support Teams (MCST) look-up table (LUT) V18.104.22.168. This is the same LUT that is being applied for the Collection 6 reprocessing of MODIS Aqua Land and Atmosphere products (completed in early 2012 for L1B products, on-going as of this writing (April 2012) for L2 products; note that OBPG actually used the update V22.214.171.124 for reprocessing). It includes time-varying corrections for changes in response versus scan angle (RVS) for bands 8 and 9, which were derived using observations of desert sites. These desert-based corrections largely replicate what the OBPG had previously derived from cross-calibration to SeaWiFS, and previously applied in recent Aqua reprocessings. The results using this LUT alone, however, still showed significant artifacts (residual cross-scan variations and detector/mirror-side striping in derived products). To mitigate these effects, the OBPG developed a cross-calibration approach similar to Meister et al. 2011, but using global MODIS-Aqua Level-3 water-leaving radiance retrievals as the vicarious calibration source. This Aqua-to-Aqua calibration (see below, 'Additional Information') is used to derive temporal adjustments to RVS, relative to the MCST calibration, that significantly reduce residual striping and cross-scan artifacts.
2.2. Processing Algorithms
Processing algorithms are identical with the previous reprocessing (R2010.0).
The plot below shows the averages of the MODIS Aqua Rrs in global deep water as a function of time both for R2010.0 (AT78, solid lines) and R2012.0 (AT85, dashed lines).
The plots below show the change in chlorophyll in oligotrophic, mesotrophic, and eutrophic waters.
For a complete analysis, see:
4. Additional Information
4.1. Aqua-to-Aqua Calibration: Ocean Bands
An analysis of global water-leaving reflectances produced with only the MCST LUT (V126.96.36.199, not using cross-calibration coefficients) showed that both the beginning and end of scan calibration showed trends that did not agree with the center of the scan. The MCST calibration applied in the center of the scan (frames 300-1050) showed very good agreement with SeaWiFS trends, whereas the edges of scan where rather inconsistent, especially for bands 8 and 9. The OBPG decided to create an angle-restricted Aqua L3 data set using only scan angles corresponding to frames from 300 to 1050. The only purpose of the angle-restricted L3 data set is its use in a cross-calibration of Aqua to itself to derive cross-calibration coefficients as a function of scan angle (for all frames, 1-1354). The only parameters retrieved are the M11 (see Meister et al. 2011; M11 are the gain coefficients; the measured radiances are divided by M11 to obtain the corrected radiances). Optimized polarization coefficients were not retrieved because an independent analysis did not produce sufficient evidence of a temporal trend in the MODIS Aqua polarization sensitivity).
The resulting cross-calibration coefficients are shown below as a function of frame (or scan angle) for detector 1, mirror side 1. The colors in the plots below correspond to time: black is the beginning of the mission (2002), as time progresses the color changes to purple, blue, green, yellow, orange, and red (2012).
As expected, there are minimal corrections (values are close to 1.0; 1.0 corresponds to no change) in the center of the scan, but there are large corrections (up to 3.5%) at the edges of the scan, especially for bands 8 and 9. For bands 10-14, the corrections are less than 1.0% (all M11 are between 0.99 and 1.01). Note that most of the temporal change is at the beginning and end of scan (in the middle of the scan, only the red line can be seen, because it is plotted right over the other lines).
The application of the cross-calibration coefficients does not significantly change the temporal dependence of the global averages of ocean color products (see above for trend plots) relative to applying the new MCST LUT only. However, striping and RVS artifacts have been significantly reduced after application of the cross-calibration coefficients (plots are shown here). Note that the Aqua L3 products distributed by the OBPG are derived from the L2 products from all scan angles (frames 1-1354) after the cross-calibration coefficients shown above have been applied; the angle-restricted L3 data set is not distributed.
4.2. Aqua-to-Aqua Calibration: Land Bands
The cross-calibration coefficients were derived for bands 8-14. For the land bands (1-7), these coefficients cannot be derived at this moment. No corrections are applied for bands 5-7. For bands 1-4 (645nm, 859nm, 469nm, 555nm, respectively) a simple striping correction was derived. The goal of the approach is to transfer the detector and mirror side dependence of the top-of-atmosphere (TOA) radiances of the nearest ocean band (667nm, 869nm, 488nm, 547nm, respectively) to the land bands. For each detector and mirror side of these four bands, the ratios of the top-of-atmosphere (TOA) radiances to those of the nearest ocean band was calculated for all valid ocean color retrievals (excluding high AOT and high chlorophyll pixels) for several selected granules. The average of these ratios over all retrievals was calculated for each detector and mirror side. The resulting averages were then normalized to the mean of the averages of that band. The resulting corrections (M11) are shown below (coming soon). The corrections are less than +/-0.5% for the different detectors, no mirror side difference was detected for the land bands.
The corrections for the land bands were derived at 1km resolution. This means that e.g. for band 1 (a 250m resolution band with 4 detectors for every ocean band detector), the same correction coefficient is applied to detectors 1-4. The evaluation of the impact of the correction has also only been performed at 1km resolution so far.
The M11 for the land bands do not depend on scan angle and time. A global analysis of the resulting ocean color products (plots shown here) indicates that the corrections perform better in the second half of the scan than in the first half. The corrections seem to perform reasonably well for bands 3 and 4 (469nm and 555nm), although there is striping for those two bands, even in the globally averaged analysis results (plots shown here). Band 1 (645nm) shows large striping and residual RVS effects, the performance of band 2 (859nm) was not evaluated. The long term solution is to derive cross-calibration coefficients for the land bands in the same way as is done for the standard ocean bands (bands 8-14), which is one of the goals of the OBPG.