Hi, I am trying to use the Forest Cover Change operator, but I don’t seem to get any sensible results. I have two S2 L2 images over Paraquay that are resampled to 20m and include all the required bands. However, none of the available Land Cover masks works over Paraquay as they contain a lot of misclassification and the latest one corresponds to 2015.
I assume the that the LandCover data are used as a training dataset, however this introduces a lot of error and omission in case that the LandCover does not correspond exactly to the date of the first input image. This is impossible for fast changing areas and it is not possible to add external forest mask.
Would it be possible to add an option to input an external forest mask? Or am I perhaps using the tool wrong? Any ideas?
Thank you.
I am not a land expert, but I think it is really not possible to say what is wrong with your process with the info you gave. There are many variables to be chosen within this processor. I guess you have to understand each of the possible masks to be used, probably many of them do not cover Paraguay, so that is why they misclassify. Since the reference map from LC CCI is 2015, and there are not S2 images previous to that date, it should be valid, right? I do not think is possible to include an external forest mask so far.
I am also interested in looking at forest change - specifically a loss of forest. I ran the forest cover change operator with the default settings and have produced a binary raster map. However I am not sure how to interpret it. Is there more information on what this does other than in the help? Perhaps its not the most straight forward tool for a beginner, as it needs more understanding and tuning of the parameters?
I would like to know more about what applications this tool can be used for, to see if it could be useful for me.
