In the final step of a multi-temporal dual-pol GRD Sentinel-1 processing graph, we apply terrain correction to a speckle-filtered stack. To retain water bodies, we uncheck the “Mask out areas without elevation” option (in GPT this corresponds to setting to false).
This setting appears to be respected only for the master image pair. For all other images in the stack, water bodies are still masked out (set to no-data).
This same behavior is observed in SNAP versions 9 through 12, both using the GUI tool and GPT, and on Windows and Linux systems.
Has anyone encountered this issue? Is there a known workaround, other than subsetting the stack by bands (dates) and applying terrain correction pair-by-pair?
Could you provide more details (e.g. product names, processing steps, parameters used and graph etc.) regarding your processing so that we can reproduce the issue on our side? Thank you.
The processing is divided into two graphs, one single date and the other one multi-temporal. The only parameters modified from the defaults are the DEM used for coregistration (COP_GLO30), the speckle filter (Refined Lee) and the water mask flag.
I think the problem was caused by that you had the “Mask out areas with no elevation” option selected in DEM-Assisted-Coregistration (see image below). Uncheck this box, the result should be fine.
Dear @jun_lu , as you can see from the multi-temporal_graph.xml file I shared, the nodataValueAtSea parameter, which corresponds to the flag you are pointing out in the GUI, is set to false. Unchecking the flag will not apply masking only to the master in the stack, as you can see below.
Running the example graphs that I provided on those two frames will yield an unmasked master:
Unchecking the option in the DEM-assisted Coregistration fixed the issue @jun_lu , thank you. We did not notice that option.
I have a question though, is the output master not masked for no elevation by default after coregistration? The help seems to be pretty outdated: Online Help – STEP