DSF correction

[emanuele.ciancia]

Hi Quinten,
I'd like to better understand the DSF method for the atmospheriuc correction, in order to asses its performance. In the attached files I'm showing you an RGB image (by rhos) of the Pertusillo lake (Basilicata region, southern Italy), the corresponding dark spectrum plot and the derived SPM map (by implementing Nechad 2016 at 665 nm). As you can see there are very high SPM values, anomalous for the season, considering that in August (image has been acquirted on 26/08/2017) the river outflow usually is minimal as well as the rainfall.
So, do you think tehre any issues due to the DSF algorithm? Any adjacency effects could have affect the choice of the most suitable band for the DSF ?
Thank you very much Quinten for all the efforts you do within ACOLITE.

Best Regards,
Emanuele

[quinten]

Hi Emanuele

I had a look at this image and think the retrieved SPM pattern is quite realistic. The selected bands and retrieved path reflectance from the DSF are not unreasonable. (Bands 12 and 1 are used, giving a SWIR-Blue fit to retrieve the path reflectance.) I believe the DSF has here largely avoided the impact of (especially NIR) adjacency in the aerosol correction. The red band has usually a relatively low impact of adjacency effects. Here the red band signal does not follow the typical pattern of adjacency effects (decay of the signal with distance to land), and is higher suspiciously close to the river inputs.

S2A_MSI_2017_08_26_09_50_31_T33TWE_L2R_RGB.jpg (207.12 KiB) Viewed 6074 times

S2A_MSI_2017_08_26_09_50_31_T33TWE_L2R_SpectrumView.png (25.82 KiB) Viewed 6074 times

Adjacency effects are present in the water spectra, but seem to manifest themselves mainly in the longer (>700 nm) wavelengths. You can see this clearly in the water reservoir in the southern part of your ROI, where the NIR/SWIR bands are dominated by the adjacency effects, due to the shape and size of the reservoir. Comparing the west and east sides of the lake, the NIR/SWIR shape is fairly consistent, and of a lower magnitude compared to the reservoir, especially in the SWIR. The NIR is still impacted with adjacency to a large degree! Note that the west side NIR signal seems to be higher than the east side due to the higher SPM concentration.

Could the river have higher SPM concentration due to the decrease in discharge? Are there possible influences of lake level?

If you need further information on the DSF, perhaps this paper helps.

Quinten

[emanuele.ciancia]

Hi Quinten,
first of all thank you very much for the explanation. Furthermore I have read your last work on RSE and thus I have understood the advantages of DSF algorithm within possible adjacent effects.
Concerning yotu question:
- From August to October the lake reaches theminimum net volume due both to low level of rainfall as well as human-induced exploitation (hydroelectric, drinking and irrigation causes). I suppose that water level fluctuations in shallow waters (the western part of the lake) can expose sediments to air, thus determining wind-induced resuspension phenomena. Anyway I think this is the reason why I retrieve high SPM values especially in the western region of the lake (I don't think the peak in reflectance in the green band is tributable to the bottom reflectance).

Thank you, again

Regards,
Emanuele