It should be useful to adapt watermask algorithm to improve watermask in coastal areas. The following ideas could be implemented/tested :
- use higher (more restrictive) Pekel threshold (ex: 90%) to avoid learn water on low tide areas
- correct NDWI threshold when selecting samples
- crop Pekel mask to pick up samples further from the coastline (ex : 10/20 m)
- provide two versions of the watermask : "inner water" (high confidence) and "coastal water" (lower confidence)
It should be useful to adapt watermask algorithm to improve watermask in coastal areas. The following ideas could be implemented/tested :