Offset between datasets

[tnair-c2s]

Hi I'm wondering why the datasets created are slightly offset geographically from one another.

Here's an example with ldasout and rtout

I understand that the resolutions are different (1km vs 250m), but don't see any documentation about these offsets. Was there a specific motivation for this, or special handling you're doing when trying to read the exact same bounds of data?

[jmccreight]

Hmm. I dont know what's going on here. I didnt perform any transformations on the data that I recall, so maybe this is part of the source data or a known thing with how that tool is interpreting the metadata?
@kmsampson, any ideas?

[tnair-c2s]

Had a look a the source data, and I missed it before that these are indeed offset in exactly the same way. So not an artifact of the zarr-creation, but just leads to a bit of inefficient reads when reading the same geographic region across the two data files since we now have to read from extra chunks.

[kmsampson]

The extents of the NWM v2.1 grids are in fact identical. Here are the GDAL GeoTransform strings for each of the grids (1000m, 250m), as specified in the NWM input file metadata:

1000m LSM grid:
GeoTransform = "-2303999.17655 1000.0 0 1919999.66329 0 -1000.0

250m routing grid:
GeoTransform = "-2303999.17655 250.0 0 1919999.66329 0 -250.0 "

As you can see, the upper left corners have identical projected coordinates and only the grid spacing differs. I was able to replicate the offset you noted using the same syntax that you provided. My guess is this has something to do with the odc geobox method and its interpretation of the CF-metadata in the NWM files. I have noticed that different libraries yield slightly different grid extents. You may wish to manually override the extent information you are getting in your process.