Hello,
      I've been making inter-comparisons among the different atmospheric correction schemes that can be implemented through SeaDAS on MODIS AQUA images. For some of my images, I see that when the NIR bands are used for aerosol model selection, the output (L2) data product has '0' (zero) reflectance for all pixels for the 748 and 869 nm bands whereas the reflectances in the other bands look reasonably good. I'd like to know what causes the reflectances to be zeroed out at 748 and 869 nm bands. Is it a bug in the software or does it have something to do with how the aerosol contribution is factored out of the data product?? An example would be the image with the ID A2008108.1050

Thanks,
Wes

Wes,

With the NIR-based atmospheric correction, Rrs(NIR) is zero by definition, as the Rayleigh-subtracted signal is attributed to aerosol alone. In more turbid or moderate to high chlorophyll concentration, this is not a valid assumption, so we make a correction based on bio-optical modeling.  We model the NIR contribution.  This correction does not get computed below chlorophyll levels of 0.7 mg/m^3, so you will see Rrs(NIR)=0.0 when chl <= 0.7, and non-zero (modeled) Rrs(NIR) when chl > 0.7.  In either case, we are not really retrieving Rrs(NIR).

-- bryan

Bryan,
        Thanks for your reply. From what I understand, pixels for which the estimated chl is <= 0.7 mg/m3 will have zero reflectance at 748 and 869 nm. This image (A2008108.1050) has appreciable levels of chlorophyll-a as estimated by the OC-3 algorithm (as high as 184 mg/m3 for my study area - Lake Kinneret) and the pixels corresponding to such high chl have zero reflectance at 748 and 869 nm (in fact all the pixels in the image). I've had images in which some of the pixels have zero reflectance at 748 and 869 nm (due to what you've described). I don't know if something else is going on with this image because the entire image (I took a subset) is blacked out at 748 and 869 nm.

Regards,
Wes

Bryan,
        I apologize. There are indeed some non-zero values. I got fooled by the visual display. Due to large number of zeroes, the display looked all black, and on moving the cursor, it looked like the pixel values were all zero. I just computed min/max statistic for the band and stretched the histogram for the display. I see that there are many non-zero values. Sorry for the trouble.

- Wes