Use cloud probability layer for each S2 scene (s2cloudless)

[dbuscombe-usgs]

GEE has provided a cloud probability layer for each S2 scene, based on s2cloudless.

CoastSat has implemented this to mask S3 clouds. So should we.

See, in SDS_preprocess.py:

• https://github.com/kvos/CoastSat/blob/d7c14e2de52f7a1d5256912eb0ac55aaf26bf86c/coastsat/SDS_preprocess.py#L218
• https://github.com/kvos/CoastSat/blob/d7c14e2de52f7a1d5256912eb0ac55aaf26bf86c/coastsat/SDS_preprocess.py#L250
• https://github.com/kvos/CoastSat/blob/d7c14e2de52f7a1d5256912eb0ac55aaf26bf86c/coastsat/SDS_preprocess.py#L350

and in SDS_download.py:

• https://github.com/kvos/CoastSat/blob/d7c14e2de52f7a1d5256912eb0ac55aaf26bf86c/coastsat/SDS_download.py#L98
• https://github.com/kvos/CoastSat/blob/d7c14e2de52f7a1d5256912eb0ac55aaf26bf86c/coastsat/SDS_download.py#L118
• https://github.com/kvos/CoastSat/blob/d7c14e2de52f7a1d5256912eb0ac55aaf26bf86c/coastsat/SDS_download.py#L191
• https://github.com/kvos/CoastSat/blob/d7c14e2de52f7a1d5256912eb0ac55aaf26bf86c/coastsat/SDS_download.py#L952

GEE article
s2cloudless benchmarking article

[2320sharon]

This issue sounds like it belongs in coastsat package. I'm going to move it there.

[2320sharon]

• This issue was developed on branch issue/164.
• This feature is now available in coastseg version 1.1.0.dev2 pip install coastseg==1.1.0.dev2
• coastsat_package version 0.1.21 contains S2 cloudless band as well as all the updated code and it has been updated to make it backwards compatible.

ChangeLog

CoastSeg

1. filter_metadata

• backwards compatible with old S2 imagery

2. edit_metadata

• backwards compatible with old S2 imagery

CoastSat_package

1. SDS_preprocess.read_bands

• modified this function to be able to read bands from the new and old S2 imagery. New imagery contains 5 raster bands and old S2 imagery contains 4 raster bands

2. SDS_download.get_metadata

• writes the new field im_dimensions to the metadata.txt
• optionally writes im_quality and im_dimensions to file if they exist

3. SDS_download.read_metadata_file

• reads new field for im_width and im_height which are used to create im_dimensions in metadata.txt
• converts the metadata values to the correct data types and can read metadata files regardless if they contain im_quality and im_dimensions or not.

4. SDS_download.retrieve_images

• now reads for more quality flags from S2
• no longer saves acc_georef 1 for PASSED image quality for S2 imagery
• saves the new cloudless band for S2 imagery