Negative rhow (bit flag 3 value 8) for Sentinel-3 DSF

[patilla7]

Hello!

I am having the same problem with S3-OLCI images in inland water. I tried processing the images by setting 'l2w_mask_negative_rhow' to false. However, the results are still the same, with no Rrs values and 8 values, in some cases 1 and even 9. I am using the darkest DSF option + glint correction, fixed AOT, and no mask for high TOA values. I am currently validating several atmospheric correction algorithms (C2RCC, POLYMER, S3-OLCI-L2_WFR, and iCOR, these algorithms produced results in the pixels shown). I would like to display negative values as well, but I am unsure if enabling the L2W mask is recommended for this validation case. Do you have any advice?

Thanks in advance,
Patricia

[quinten]

Hi Patricia

Could you share the settings and inputfile for this example?

Quinten

[patilla7]

Hello Quinten,

Attached is a subset of the S3-OLCI-L1C product that I am using as an input file. The settings are as follows:

l2w_parameters = Rrs_*
rgb_rhot = False
rgb_rhos = False
map_l2w = False
dsf_spectrum_option = darkest
use_supplied_ancillary = True
dsf_residual_glint_correction = True
l1r_delete_netcdf = True
ancillary_data = False
dsf_aot_estimate = fixed
l2w_mask_high_toa = False
glint_force_band = 1020
glint_mask_rhos_wave = 1020
l2w_mask_negative_rhow = False


I corrected the image again using the same settings. However, this time I set the l2w_mask to false, and it finally gave results (negative Rrs values). I wonder if there is another masking option that could be hampering the results of showing negative Rrs values.

Thanks in advance,
Patricia

[quinten]

Hi Patricia

I separated the masking procedure from the l2w generation sometime in January, and the new masking procedure was not picking up the correct settings from the sensor defaults. For OLCI the mask buffering should be disabled, and this is what caused the disabling of the negative masking not to work. The latest commit should hopefully resolve this!

S3B_OLCI_2024_03_06_10_11_19_L2R_rgb_rhos.png (72.92 KiB) Viewed 81449 times

S3B_OLCI_2024_03_06_10_11_19_L2W_rgb_rhow.png (29.47 KiB) Viewed 81449 times

Note that it may be useful in this region to start using the elevation pressure dependency derived from the DEM:

Code: Select all

dem_pressure=True

I am not necessarily a fan of using OLCI 1020 nm for glint correction, especially not in such a narrow water body with mixed pixels and adjacency effects.

I hope this helps!

Quinten

[quinten]

Hi Patricia

For older versions of ACOLITE the l2w_mask_smooth option was always on, which adds a buffer around the masks and explains why your narrow lake is masked even by disabling negatives masking. The new default for OLCI is l2w_mask_smooth=True, and you could try specifying this even in version 20221114.0 to allow for more pixels to come through. Maybe the masking threshold needs to be slightly increased as well for this scene.

I hope this helps!

Quinten

[patilla7]

Hi Quinten,

Thank you very much for your reply. In fact, I'm still working with v20221114.0, so the l2w_mask_smooth is not specifically declared for this version. However, I followed your advice and set both l2w_mask_smooth and l2w_mask_negative_rhow to false, which resolved the issue, now I can see the negative Rrs values. Regarding using 1020 band for glint correction, I tried processing all my data, most of them coming from narrow lakes like the one I have shown, with and without glint correction using 1020 band, and results were significantly better applying the glint correction.

Thans for your help!
Patricia

[quinten]

Hi Patricia

I'm glad your issue is resolved! It would be useful to compare the negatives that are introduced by the glint correction, as in this lake the 1020 may be influenced by mixed pixels and adjacency.

For most settings, the values from the general defaults.txt will be used, replaced only with sensor specific values when they are set. The l2w_mask_smooth was set in the defaults for version 20221114.0 but indeed not specifically turned off for S3A/B OLCI. This is why I said it was always on in older versions in the previous post: set to True in the defaults, and not set to False in the sensor specific defaults. User settings take highest priority, so if you set it to False explicitly it should work.

I hope this helps!

Quinten