Atmospheric Correction Issues for Future Consideration

Revised Solar Irradiance Model

The current solar irradiance model used in the algorithms is based on Neckel and Labs, which was published in 1984. Recently other projects (e.g., MODIS) have switched to a more recent model developed by Thuillier (1998). Also, Thuillier has been working on additional refinements to his model, and a new version was published in 2001. The SeaWiFS project has received a copy of the latest Thuilier spectrum (2002) to be published soon. The effects of a model change are under analysis. The decision has been made to postpone implementaion of the Thuilier spectrum until the latest spectrum is officially published and its effects on SeaWiFS products are fully quantified.

Alternative aerosol model selection algorithm

The current aerosol model selection algorithm is designed to choose two models from the full set (currently twelve) and interpolate linearly between them, based on the ratio of the Band 7/Band 8 aerosol reflectances. This has problems under several commonly encountered conditions: 1) Two models have nearly identical ratios for certain geometries; 2) The aerosol reflectances for the other bands change discontinuously when the model pairs change; 3) The uncertainty in the ratio, due to the radiometric resolution, is large for small IR band values, causing unnecessary model switching. This last problem is largely addressed by the epsilon smoothing algorithm described in 1.5. However, alternative model selection algorithms have been discussed to reduce the impact of all of these problems, by utilizing more than just two models to compute the aerosol reflectances.

The first scheme proposed was a Gaussian weighing of multiple models, with the width of the Gaussian determined using the estimated uncertainty in epsilon. Some tests of this scheme did not work well, and the suspected cause was non-uniform distribution of the models in epsilon-space.
A scheme has been proposed which would involve a weighted linear fit to multiple models in the vicinity of the computed epsilon. The output reflectances would then be determined from this line. (Note that the current algorithm is essentially the simplest form of this scheme, in that linear interpolation is a special case of a linear fit using two data points.) The weighting scheme is TBD, but would also use information about the epsilon-distribution of the models and the uncertainty of the computed ration. This would reduce or eliminate the impact of both (1) and (2), since the abrupt changes of model pairs would be replaced by smooth variations in the model weights.

N02 correction

At the 5th SIMBIOS Science Team Meeting (15 - 17 January 2002, Baltimore, MD), it was suggested that NO2 absorption may not be insignificant. The current atmospheric correction algorithm does not consider NO2 absorption. The topic bears furture study.

Pressure dependence on the band 7 oxygen absorption correction

The current oxygen absoprtion correction for band 7 is atmopheric pressure independent. The pressure effect on this correction is small and likely within the uncertainty of the correction. The effort involved has been determined to be unwarrented for this reprocessing.

f/Q correction

We have been working with David Antoine and Andre Morel on this issue since October of 1999. Our approach has been to evaluate the effect of the correction on reducing viewing-angle dependencies in the normalized water-leaving radiances. Thus far, we have been unable to demonstrate that the algorithm will improve the quality of the SeaWiFS products. The collaboration with Antoine, Morel, and Gentilli has provided confirmation that the latest updates to their code have been correctly implemented in the SeaWiFS processing code. The results of the analysis of the effect of the f/Q correction on SeaWiFS nLw and chlorophyll a retrievals is provided below.