Ocean Color Reprocessing 2009.1
A follow-up reprocessing to the R2009.0 Reprocessing of SeaWiFS was initiated in late Fall 2009. Subsequent to Reprocessing 2009.0, a minor error was discovered in the update to the iterative model used for deriving the water-leaving radiance contributions to the near-IR. This NIR Lw algorithm and the changes made for Reprocessing 2009.0 are described here:
The problem and solution:
1) The revised model for total absorption at 670-nm was specified incorrectly, such that predicted absorption was too high in highly productive waters, thus over-predicting backscatter and thereby removing too much water signal from the NIR channels. The net result is to remove too little aerosol in the visible and thus under-predict chlorophyll by 5-15% in these waters (chl >> 1 mg/m3). This problem was corrected, and coastal retrievals are now better than ever before, relative to field measurements.
2) The iteration control was often reaching maximum iterations of 10, which is a failure condition that is masked in Level-3. This was improved by a revised iteration reset procedure. As before, the first iteration assumes zero water-leaving radiance in the NIR (all residual radiance is aerosol). If that first pass fails, the iteration is now reseeded by assuming zero aerosol radiance in the 670-nm reflectance (all residual radiance is water). This strategy more often allows the model to iterate to convergence, generally within 4-6 iterations total. Non-convergence is still flagged in Level-2 and masked at Level-3.
Evaluation of the solution:
Impact of the NIR Lw fix on global retrievals is only significant in eutrophic waters, with mean increase in chl of 6%. See:
It was previously shown that the updated NIR Lw model, as originally implemented for the R2009.0 Reprocessing of SeaWiFS, resulted in a substantially improved agreement in the turbid and highly productive waters of Chesapeake Bay. The regional Chesapeake Bay time-series was re-evaluated to verify that improved agreement with field measurements was still demonstrated. The link below compares the results of the R2009.1 configuration to that of R2007 (SeaWiFS Reprocessing 5.2).
In summary, we are getting more "valid" retrievals than ever before. SeaWiFS Chesapeake Bay chl match-ups with field measurement are very good (better than ever). Reduction in negative radiances at 412 is still substantial, relative to 2007 processing. There are no major changes to previous statements or conclusions regarding the results of R2009.0.
The reprocessing of MODIS-Aqua was initiated after the issues with the R2009.0 reprocessing were discovered, so MODIS-Aqua Reprocessing 2009.1 includes all changes described under R2009.0, as well as the R2009.1 correction to the NIR Lw algorithm.