Glint in island wake (original title: high aerosols variablility)

[jcs1973]

Hello

Thank you very much for providing acolite software and their support. We are processing Landsat 8 and Sentinel 2 data using acolite 20180908.0 version.When the quality of the input data were good the results were great. However, our area of interest is highly complicated from atmospheric aerosols variability.

We are sending an example of the output from acolite (reduced resolution). It corresponds to a subset of Landsat 8 rgb image (rhos_655, rhos_561, rhos_483) linearly scaled from 0 to 0,1. Also we also sent the corresponding acolite_run_20180920_145703_settings.txt file. The dark spectrum fitting (DSF) and glint correction were chosen. Would it be posible to improve the results over island wake downward throughout settings file modification.

When there were not "dark ocean" areas over the most image, the results seems were not suitable. What are about aproximately, the maximum values of AOD with those who acolite (DSF) is able to get a successful atmospheric correction?

Thank you very much

Josep

[quinten]

Hi Josep

Thanks for your kind words and the opportunity to look at your beautiful study area.

I think the issue here is not so much the variability of aerosols, but the strong sun glint you see in the wake of the island. By default, the sun glint correction is only performed for reflectances at 1.6 µm < 0.05. For most of the glint in your image this threshold is exceeded, and needs to be raised (by using glint_mask_rhos_threshold). You can also try a near-infrared band for the glint correction, as there is probably no significant water signal in the NIR. It is likely though that the results for the pixels inside this strong glint will not be very accurate.

Quinten

RGB composite after default processing (no glint correction)

L8_OLI_2018_07_16_11_41_12_208040_rgb_rhos_default.png (329.98 KiB) Viewed 10332 times

RGB composite after applying glint correction (glint_correction=True, glint_mask_rhos_threshold=0.12)

L8_OLI_2018_07_16_11_41_12_208040_rgb_rhos_gc.png (309.95 KiB) Viewed 10332 times

RGB composite after applying glint correction using the NIR band (glint_correction=True, glint_mask_rhos_threshold=0.12, glint_force_band=865)

L8_OLI_2018_07_16_11_41_12_208040_rgb_rhos_gc_nir.png (297.84 KiB) Viewed 10332 times

[quinten]

Some transects along 28° North for the images in the previous post. You can clearly see the edges of the island wake at around pixel 800. With the default processing the surface reflectance is way too high inside the wake. The glint correction does improve significantly, but there may be some residual glint signal in the wake (there is still a clear jump around pixel 800). It seems the NIR based glint correction gives less noisy results compared to the SWIR based one, and has a bit lower surface reflectances overall.

Top-of-atmosphere reflectance

transect_rhot.png (82.23 KiB) Viewed 10332 times

Surface reflectance from the default processing

transect_rhos_1_default.png (69.7 KiB) Viewed 10332 times

Surface reflectance after SWIR glint correction

transect_rhos_2_swirgc.png (99.42 KiB) Viewed 10332 times

Surface reflectance after NIR glint correction

transect_rhos_3_nirgc.png (90.93 KiB) Viewed 10332 times

[jcs1973]

Hello Quinten

Thank you very much for your prompt, complete and useful answer. The changes that you has pointed out, has improved a lot the results.

You were right about the fact that glint over wake was the major issue. Although islands downward wakes also produces different patterns of aerosol accumulation, calmer waters over there lead to different glint (higher) patterns too.

Josep

[quinten]

Hi Josep

I understand the islands may cause changes in the aerosol composition and concentration. In this case, it is highly unlikely to detect that though, due to the extreme glint reflectance. Let me know if you find any scenes where the different aerosols may be detected - I'd be very interested in those!

Quinten

[jcs1973]

Hello

I will look for images where aerosol changes inside outside from islands wakes could be identified. Although the illumination patterns will not be the same, I will use to MODIS and VIIRS data to check and compare.

Josep