Earthdata Forum

Dear colleagues,
in Kd_490 level 2 files I have moderately high outliers in an area which is mostly flagged by the cloud-ice mask. I am looking at a tropical area and therefore know that the area is covered by clouds. The high turbidity values are supposedly caused by stray light and shadow effects of the clouds. Please have a look at the attached image.
I would like to exclude these values from the band but am struggling to find a good way to do so. Applying a filter, like for example a mean 1.5 pixel radius does not deliver satisfying results. This must be caused by the fact that no-data values of the source band are not considered by the filter operation. So, should I simply take off a certain percentage of the highest values (e.g. the 10% of highest values)? But this would cause a potential loss of valid data.
Did anybody encounter a similar situation? What is the way to solve this? Are there any references to rely on?
Thanks a lot! Greetings, JC

See the help pages for the Mask Manager

You could apply the STRAYLIGHT flag as a mask.  In this image, the high Kd490 in the upper right portion of the image is also flagged as COASTZ (i.e. <30m) ...clear, shallow water will probably have a bottom reflectance component that will fool the Kd algorithm.  You may also want to mask HIGLINT, as that can also fool the algorithm.

Regards,
Sean

Hi Sean,
thanks for the advice! Is there a study/paper on the Kd algorithm failing in shallow waters available in addition to those mentioned here (https://oceancolor.gsfc.nasa.gov/atbd/kd_490/) under references?
Thank you! Greetings, JC

The issue is the impact of bottom reflectance of ocean color retrievals, not that the algorithm fails simply because the water is shallow.  There are plenty of articles on the impact of bottom reflectance (not necessarily specific to the retrieval of Kd)

Sean