Include DEM coregistration within DEM generation workflow

[loreabad6]

After successful extraction of elevation values from the InSAR process, a vertical coregistration shoul be performed. To do so, we include the demcoreg module within our workflow. This are some steps and points to consider:

• [x] The demcoreg module relies on a separate conda environment: find out how to call it properly within the workflow
• [x] The parameters ref_fn and src_fn should be correctly defined: reference DEM would be our external DEM data, while the source would be the recently generated S1-based DEM; document this properly
• [x] Make sure that both the reference and the source DEM are compatible: using different projections and spatial resolutions for both files seems to generate horizontal decoregistration and other problems; document this properly
• [x] When not using a mask layer, demcoreg selects surfaces for coregistration that might not always be stable: make sure to document the need of a mask layer, in our case, that removes the landslide area and when relevant the water and coast areas. See figure in comments 🖼️

[loreabad6]

Example of a coregistration test with the Reference DEM from official Norway repos resampled to 30 meters. Our resulting DEM was warped to EPSG:25835 to agree with the reference.

Things to note:

• Landslide area is used as surface for co-registration
• Coastal areas seem to be problematic

This was the code used (note-to-self: temporal directory names, might change locally):

dem_align.py home/data/ref_dem/NDH_Alta_2pkt_2018_DTM_resample30m.tif home/data/ani_test/20200605_202008
10_elevation_EPSG25835.tif -mode nuth -mask_list none -outdir home/data/ani_test/test4 -res max -max_offset 500

[loreabad6]

Unstable areas were removed before co-registration. New issue is a horizontal shift on the resulting "aligned" DEM. See the docs repo here for updates.

[loreabad6]

Switching to an approach with xdem however, horizontal shifts still happen with Nuth-Kääb