How should I create the layer stack for multitemporal filtering

I work with a time series of preprocessed Sentinel-1 images, my outputs are in tif. I’ve created a layer stack with gdalbuildvrt then gdal_translate. Unfortunately, it seems that I’ve lost some metadata information because I have an operator exception telling me that "band ‘band1’ has no unit. Can I update my metadata somehow to provide SNAP with the required information ?

You should stick to the BEAM DIMAP format during the processing and stack the images in SNAP. Basically, you need to coregistrater all rasters, or alternatively, terrain correct and stack, before multi-temporal filtering.

1 Like

Thank you for the tips. My preprocessing does well include terrain correction, but the output was in tif and I would like not to run it again because I have many images.

I was able to launch the multitemporal filter with a dummy value for unit, manually defined after selecting all band in the metada of my file (opening band “folder”, selecting all bands, right click > properties, editing the “unit”. But isn’t it possible to do it with a command line or within a graph, because I need to launch several MTF and I have more than 300 images?

I’m afraid only the BEAM DIMAP format stores all metadata correctly. You can easily repeat the preprocessing by creating a xml file in the graph builder which is then applied to all original images in batch mode.

An example is given here:

OK, thank you. I’ve already got a graph but it takes about 2hours per image (gamma0 with terrain corrections), so I wanted to avoid running it again simply because one metadata value needs to be set. For sure I will update it for the next time.