Hi,
I want to classify Sentinel-3 OLCI WFR products using the Forel-Ule classification. I’m particularly interested in characterizing turbid plumes in a large river estuary.
The processing runs fine in SNAP GUI and when using a graph from the SNAP graph builder.
Yet, it looks that the classification output is cropped to a max value of 16 instead of 21 in the FU scale. Many pixels within the estuary are masked as 0 (see print screen below).
Any idea how we can fix this ?
The valid pixel expression is set to default (WQSF_lsb.WATER and not WQSF_lsb.CLOUD).
Many thanks for your help
I worked around the issue by replacing all pixels with a value of 0 with a default value of 17.
A few outlier pixels appeared outside the estuary, but I was able to remove them through some post-processing steps.
It’s not a perfect solution, but it did the trick for my use case.
Best regards,
Olivier
I guess what happens is that those pixels are invalid after the atmospheric correction process (you are using L2 data, right?). Probably the range of turbidity is too high and they are masked.
Dear Ana,
Yes I’m using S3 OLCI WFR products as input.
Actually those pixels are not masked in the WFR products, they are set to 0 during the FU classification. The rest of the invalid pixels have a NaN value.
I think this is due to a low threshold applied on the Hue angle during the FU classification.
I could lower that threshold myself and recalculate the hue angle from intermediate results. I didn’t try that yet.
