AC Using fix Angstrom

By adogliotti Date 2012-03-12 13:30

Hello,

I want to use a fix Angstrom coefficient to perform the atmospheric correction and band 2130 to extrapolate the aerosols. So I used the following options

aer_opt= -6
aer_angstrom= -0.1759
aer_wave_long= 2130

And when I look at the output, the Rrs_748 band is completely flagged. When I look at the log file I find that it says
"Aerosol selection bands 748 and 2130"
and after loading the aerosol models it says
"Using fixed aerosol model based on predefined Angstrom exponent Extrapolating from 2130.0 nm".

Then I tried the same as before but added

aer_wave_short= 2130

This time 748 band was not flagged, and the log file says

"Aerosol selection bands 2130 and 2130"

and

"Using fixed aerosol model based on predefined Angstrom exponent Extrapolating from 2130.0 nm".

Looking at the two outputs, the values of all the bands are different. So I was wondering if it's actually using the fixed Angstrom or keep using two bands to get the aerosol model.
Thanks in advance

Ana

By @bryan Date 2012-03-14 02:04

Ana,

Just a quick note to let you know that I I looked at this briefly and confirmed your findings. aer_wave_short "should" have no impact when aer_opt=-6, but it does. We'll have to investigate.

-- bryan

By @sean Date 2012-03-17 18:23

Ana,

There are two, somewhat related issues at play here.

1) l2gen will set the Rrs for the bands used in the aerosol selection to the fill value.
While 748 wasn't actually used in the aerosol selection when aer_opt=-6, the logic
that sets the Rrs fill value is separated from the aerosol selection process.

What actually happens is that Rrs isn't computed for the bands set as aer_wave_{short,long},
and since Rrs for all bands is set to the fill value prior to the calculation, Rrs_748 is not
computed when you don't change aer_wave_short from the default.

2) The glint correction has logic that attempts to prevent an "over-correction".
This logic will modify the resulting TLg if the ratio of TLg[aer_wave_short] to La[aer_wave_short]
or the equivalent ratio for aer_wave_long is above a predefined threshold.
When you change the aer_wave_short, the TLg calculation is therefore affected.
For the vast majority of pixels, this logic has no impact (i.e. the ratio is below the threshold).
However, where it is above the threshold, TLg changes, and every band will be different by
that amount.

So, if you want Rrs_748, change the aer_wave_short so that it will be computed, but you
should be aware that there will be minor differences when compared to a processing that
has aer_wave_short set to 748. (Well, individual pixels affected may have large differences,
but most pixels won't be affected)

We may modify l2gen to handle the first case better so Rrs can be computed if aer_wave_short
isn't really used in the aersol selection. The second issue of the glint correction may take some
consideration, though. I never really liked this approach to preventing "over-correction", but we
haven't had time to come up with something better.

Regards,
Sean

By adogliotti Date 2012-03-20 14:57

Thanks for your reply, now it's clear what's happening.

One more question which is closely related to this topic. I want to retrieve rhom_xxx (MUMM Rayleigh corrected reflectance). It corrects for whitecaps and glint, so it partially corrects the image using Stumpf et al. atmospheric correction. The problem is that this correction flags part of the my region because of the bands it uses. So I tried using aer_opt=-9 and the bands for the NIR-SWIR algo. Then I get rhom products and the region is not flagged, but slight different values for the rest of the pixels are retrieved. I guess this is due to differences in the glint correction as you mentioned.
I just want to check that what I've done is correct and Seadas is actually calculating rhom using the NIR-SWIR switching AC algo for the glint correction thus avoiding the flagging in my region.
Thanks again!

Ana

By @sean Date 2012-03-21 13:38

Ana,

The rhom product can be computed using any of the atmospheric correction options, each may (well, likely will) produce different results.
BTW, Stumpf et al. isn't an atmospheric correction, it's simply a method to address non-neglible NIR reflectance in the atmospheric correction
process - as does the MUMM approach.

Now, for the NIR-SWIR switching approach, where the algorithm chooses the NIR bands, the result should be identical to the default atm. corr.
algorithm, as the bands passed to the glint correction code are the same (when the NIR is selected). Now, where the SWIR is used, yes, the glint
correction may differ as I explained in the previous post.

Sean