Page 1 of 1

AQUA MODIS Cloud Mask Useless

Posted: Mon Aug 08, 2022 10:01 am America/New_York
by captlen
After extracting turbidity (K490) data from the AQUA MODIS Data files. The cloud mask is in error. It is to intense creating cloud covered area that are actually clear. As an example, on 8-6-22 AVHRR imagery was clear all day, but the AQUA MODIS data after extraction showed major cloud cover making the image completely useless. The K490 cloud mask should be corrected.

Re: AQUA MODIS Cloud Mask Useless

Posted: Tue Aug 09, 2022 11:14 am America/New_York
by OB.DAAC - SeanBailey
No, the cloud mask is not in error. SST uses infrared wavelengths, while ocean color uses visible/NIR wavelengths. The two do not "see" the same thing with respect to what constitutes a cloud. SST will often be able to "see" through clouds that ocean color cannot. While the cloud mask threshold may be slightly conservative, it's not too bad. Sure, when compared to the SST scene it seems extreme, but most of what is wiping out the data in the Kd490 image you show is glint. In the high glint region, the bands used for the atmospheric correction of the ocean color data are saturated in MODIS and thus unrecoverable.

There is no simple tweaking that can be done to recover more ocean color data.

As an aside, the Kd490 product is not a measure of turbidity, it is a measure of the diffuse attenuation of light (at 490nm). It can be correlated with turbidity, but also can be completely uncorrelated. Turbidity is the result of suspended particulate matter, which will scatter light and thus increase attenuation (thus the correlation with Kd). However, waters with high colored dissolved organic matter content will also attenuate - but not scatter - the light. Waters with high CDOM are not necessarily turbid. In situ nephelometric measurements of turbidity have a detector at 90 degrees to the incident light source, and so only measure scattered light....we can't do that from space :D We *can* separate out absorption from scattering in the remote sensing reflectance signal using optical models (typically of the semi-empirical variety). The backscattered light can also be correlated with turbidity (and likely better than Kd).

Regards,
Sean