ValueError: 'dem_to_be_aligned' had only NaNs when running fetch_and_coregister.py

[Jack-Hayes]

Running scripts/fetch_and_coregister.py can raise ValueError: 'dem_to_be_aligned' had only NaNs. This occurs when the fetch step produces one or more DEM COGs with no valid pixels (all masked/NaN) so the coregistration wrapper fails--you can't coregister a NaN-filled raster.

Reproduction

run from repo root, can edit local paths to your desire

python scripts/fetch_and_coregister.py \
  --bounds 161.18581265 56.60286466 161.19244308 56.60878054 \
  --fetch_output_dir /tmp/fetch_arcticdem/fetch \
  --coreg_output_dir /tmp/fetch_arcticdem/coregistered \
  --date_range 2015-10-01 2025-10-17

ValueError: 'dem_to_be_aligned' had only NaNs

The fetch step returned a DEM that are fully masked (all NaN). Need to double-check the valid-pixel masking and filtering logic. A quick fix is to just run the fetch tool and then move/delete the invalid DEM before you separately run the coregister tool

[Jack-Hayes]

Note that part of this issue is because the argument --min_valid_fraction is 0.0 in the fetch_and_coregister tool (and all tools):

https://github.com/Jack-Hayes/fetch-arcticdem/blob/main/scripts/fetch_and_coregister.py#L44

Though changing it to a positive value (i.e. 0.7) still returns the same error despite filtering out some of the problematic DEMs