Ocean Color Forum - Not logged in
Is there a way to implement a more robust cloud detection methodology than the threshold of a single band using L2gen?
I find that when using Hi-Res MODIS with the NIR/SWIR correction that extremely turbid waters are tripped by the cloud flag and not processed.
I am relatively new to using SEADAS and would like to implement something similar to what I think NRL uses for their APS (see below) .
From the APS Users Guide.
In addition to the standard NASA albedo theshold, n2gen provides
the reflectance ratio test. This test is part of the MODIS ATBD MOD35: "Discriminating
Clear-Sky From Cloud With MODIS". For MODIS 250m processing, it allows for full-resolution
cloud detection since the ratio test using the 1KM data.
The reflectance ratio test uses the 870 band divided by 660 band
(R0.87/R0.66). This test makes use of the fact that the spectral
reflectance at these two wavelengths is similar over clouds (ratio is
near 1) and different over water and vegetation. Using AVHRR data this
ratio has been found to be between 0.9 and 1.1 in cloudy regions. If
the ratio falls within this range, cloud is indicated.
Not without modifying the source code and recompiling l2gen.
However, you could test the effectiveness such a mask before
taking the time to modify l2gen by generating a L2 file with
rhot (TOA reflectance) then use the 'User-Defined Function' to
implement the algorithm.
I'm curious, have you experimented implementing the band-ratio cloud masking algorithm? How has it worked in your turbid study area?
The short answer is no. Our programmer looked over the code, and thought there were better things to do with his time. However, the surface reflectance land product MOD09 & MYD09- GA have this methodology implemented in their flags (1 km cloud mask), and you can couple this product with the GQ (250m bands 1 and 2) product and work out something better than the thresholding used in l2gen. The threshold cloud mask has got be one of the biggest problems for us coastal Hi-RES users, as the most turbid days are flagged as clouds.
You should check out, (Doxaran et al. 2009). He has a link to the pdf on his website. http://www.obs-vlfr.fr/LOV/OMT/fichiers_PDF/Doxaran_et_al_ECSS_2009.pdf