geoscience-aman
commented
9 months ago Thanks for the feedback @jeremyh and @uchchwhash!
- "Hand-verification" here: https://nbviewer.org/gist/geoscience-aman/9576576892651d45060a736a5136159a
- Have removed passing xml elements and implemented other feedback.
- Level 1 zip path fallback can be a separate PR.
- May need your help for adding the xml logic test case @jeremyh.
Background
Sentinel-2 Level 1C data has a quality issue - “4.5 Striping due to lost source packets (no. 10)”, which is highlighted in ESA's Data Quality Report:
https://sentinels.copernicus.eu/documents/247904/685211/Sentinel-2_L1C_Data_Quality_Report.pdf/6ad66f15-48ca-4e65-b304-59ef00b7f0e0?version=5.0&t=1625665123197
For example, the black squares in the scene below are missing packets. They appear as low values, generally below a
10,000 Xscaled surface reflectance of ~50. Importantly, these low values fall through the contiguity logic, which asks if thedata > 0across all bands for a given pixel, but are not usable for any practical purpose, as far as I'm aware.This PR
This PR adds the functionality to check whether a scene has missing packets and if so, to mark the affected pixels as
nodata (-999)using ESA's missing packet mask. Based on this fix, the contiguity layer should automatically be updated to mark the missing packet pixels as non-contiguous.Additionally, the contiguity layer creation has been updated to include all Sentinel-2 bands, as originally intended.
Testing For the following scenes,
wagl.package()with default parameters has successfully run to completion. The band images and contiguity layer has been visually inspected to ensure the contiguity logic is being followed as expected..gmlvector mask). https://nbviewer.org/gist/geoscience-aman/9576576892651d45060a736a5136159a.gmlto.jp2. (https://sentinel.esa.int/web/sentinel/technical-guides/sentinel-2-msi/level-1c/masks)