Why are TOA images used instead of L2 images (Surface reflectance)?

[fmartinp]

Congratulations for the toolbox.

I would like to ask you why you decided to use Top of Atmosphere images instead of the atmosphere corrected images (L2 products - surface reflectance).

Thanks in advance.

[kvos]

hi @fmartinp , are you referring to Sentinel-2?
For Landsat we are using the TOA collection as it has a better temporal depth than SR. With Sentinel-2 I also went for TOA to be compatible with the Landsat shoreline outputs. However, I agree that we could put more effort in evaluating the advantages/disadvantages of using L2 for Sentinel instead of L1 as there are a lot of issues with cloud masking on the TOA products.

[fmartinp]

Hi @kvos,

Actually, I was referring for Landsat and Sentinel-2. I see your point about the temporal depth. Even Sentinel-2 L2 products start in 2017 instead of 2015 as for TOA products.

From your answer, I understand that you took the decision of using TOA products in order to maximise the time span. Is that correct?

Best regards.

[kvos]

yes that's correct!

[fmartinp]

Thanks for the clarification!

I believe that it would be interesting to compare the coastline results between TOA and Surface Reflectance (SR) in order to see if there is a significant benefit in using SR.

Probably, I will try it for a small area, although my results will not be very relevant since I am not familiar with the classification algorithms. So, I will use the trained neural network as provided in your repository. Since the NN is trained with the TOA images, perhaps does not work as good as it could be if trained with SR.

I will post some results if I get the time for that test.

Regards.

[kvos]

thanks @fmartinp for looking into this, should be interesting.
To test the accuracy of the shorelines you can use the SDS_Benchmark repository (https://github.com/SatelliteShorelines/SDS_Benchmark)!