high SWIR water reflectance values

[dimitris_p]

Hi Quinten,

I'm running ACOLITE v.20221114 for a series of S-2 images to get RC and aerosol corrected reflectances.

I've noticed high SWIR aerosol corrected values for water pixels (both dark and glint-affected pixels).

I'm attaching snips with the spectral profiles and the settings files. The AOI is in Lesvos Island (Northern Aegean, Mediterranean). The plots are from open water conditions and more turbid waters inside the gulf of Gera. The acquisitions I'm interested in are mid-summer, but I've noticed the same issue with acquisitions with lower sun angles (attached).

It's the upwards slope from the Red Edge to NIR to SWIR bands that I'm curious about, and more specifically 865 to 1610 and 1610 to 2200 nm.
I'm not running glint correction, so I'm guessing it'd be normal for the glint affected pixels? But the slope is also present with the dark water pixel reflectances, and is mostly prevalent in the aerosol corrected reflectances. It's the same with full-scene and cropped outputs, and tiled vs. fixed aot estimation. I'm guessing I might be doing something wrong with the other settings? There is a difference with lower sun angles, but I'm still getting the upwards slope from 1610 to 2200 nm.

Any insight would be much appreciated and thank you very much for all your work!

Dimitris

PS: Here's a WeTranfer link for the rest of the PNG and settings files I couldn't attach: https://we.tl/t-r7Edk2SxLv

[quinten]

Hi Dimitris

Thanks for your question! I think both the dark and bright areas are affected by sun glint. In the 2022-06-16 image (tile 35SMD, subset to limit=39.03,26.51,39.05,26.54) all water pixels have TOA SWIR1 reflectance greater than the default L2W masking threshold (0.0215). The glint dominates the NIR/SWIR signal while it has a lower impact on the VIS bands, and is hence less obvious on the RGB composites. The spectral shape of the glint spectrum depends on atmospheric conditions and the band specific viewing geometry.

I have tried the glint correction on this image but it gives negatives from the red bands onward since the water signal at longer wavelengths is so low in this area. Depending on your application it may be worth using a processor which models the water reflectance shape (such as POLYMER or C2RCC).

I hope this helps!

Quinten

[dimitris_p]

Hi Quinten,

Thanks a lot for your reply!

So if I'm understanding correctly, the water spectral shape is such due to glint. In this case shouldn't the shape be similar in the SWIR part for both AC and Rayleigh reflectance? Also, in the full-scene snip I attached, the area where pin 1 is placed actually has a lower SWIR1 than the threshold, but the upwards slope still exists (although with negative values), is this also due to glint?

A follow-up question: I'm running the chl-a algorithms for a specific site inside the limited region and I'm getting abnormally high values compared to in situ measurements, could this also be due to high-glint conditions?

Thank you again and please excuse my persistence!

Dimitris

[quinten]

Hi Dimitris

The aerosol reflectance and transmittance will play a role in the shape of the spectrum. Rayleigh reflectance is very low SWIR and the Rayleigh transmittance is practically 1. For example, you can see this when comparing the rhot_* (diamonds) and rhorc_* (triangles) spectra, in the SWIR they practically align:

Dimitris_rhorc.png (636.62 KiB) Viewed 112985 times

Which chlorophyll retrieval algorithms are you using?

Quinten

[dimitris_p]

Hi Quinten,

I've tried all besides oc2 and oc3 since I'm working with turbid waters, mostly Mishra and Bramich though. Both use red and red-edge bands, so it makes sense that they're overestimating?

I've only run them on summer-autumn data (so a lot of glint due to high sun angles) since I didn't have any other in situ data, but I'll just run them again with winter acquisitions to see if there's any significant difference.

I haven't change any of the default values though, so maybe it's something I'm not setting up properly?

Thank you!

Dimitris

[quinten]

Hi Dimitris

I don't really support the individual parameter retrieval algorithms in ACOLITE. I mainly just implement them from literature.

For the waters in this image I would however assume the red edge algorithms not to perform well due to the low turbidity and low productivity. Generally a high backscattering signal and high chlorophyll a concentration (>10 mg/m3) is needed for those algorithms to work. Which turbidities and chlorophyll a concentration do you encounter in these waters?

Quinten

[dimitris_p]

Hi Quinten,

Chl-a concentrations are generally in the 0.2 to 2 mg/m3 range, so yeah much lower than 10 mg. In terms of turbidity I don't have specific measurements, but as far as I know Secchi disk depths are at about 5-8 meters.

Thanks again!

Dimitris

[quinten]

Hi Dimitris

A quick google search tells me a Secchi depth of 5-8 metres would be a turbidity of <1-2 FNU, together with the low chlorophyll a concentration the red edge products are probably not suited for your region! I guess turbid waters is quite a relative term.

Quinten

[dimitris_p]

Hi Quinten,

Yes, I guess that's true . Looks like I need to look for a different approach.

Thanks a lot for your help!

Dimitris