Setting recalc_transform=False when saving aligned raster on disk

[victoriafoing]

https://corteva.github.io/rioxarray/stable/examples/reproject_match.html

I want to align a raster to another reference raster on the disk. When I use reproject_match to align a raster in memory, it has the same transform of the reference raster (e.g. bounds, crs, resolution). However, when I save the aligned raster with rio.to_raster(), it has a slightly different transform than the reference raster. Do I need to set recalc_transform=False when saving the aligned raster to make sure it is aligned with the reference raster?

[snowman2]

Can you provide a minimal re-producible example?

[victoriafoing]

If I align a raster to a reference raster (or even align it to itself) using the following:

import rioxarray
from rasterio.enums import Resampling

reference_raster = rioxarray.open_rasterio(reference_raster_path)
array_coords = {
    "x": reference_raster.x,
    "y": reference_raster.y
}
reprojected_raster = reference_raster.rio.reproject_match(
    reference_raster, resampling=Resampling.nearest
)
aligned_raster = reprojected_raster.assign_coords(array_coords)

And then I save the raster with recalc_transform=False, its resolution matches that of the reference raster.

aligned_raster.rio.to_raster(aligned_raster_path, recalc_transform=False)
aligned_raster = rioxarray.open_rasterio(aligned_raster_path)
aligned_raster.rio.resolution () == reference_raster.rio.resolution()
# Returns True

But if I save the raster with recalc_transform=True, its resolution does not match that of the reference raster.

aligned_raster.rio.to_raster(aligned_raster_path)
aligned_raster = rioxarray.open_rasterio(aligned_raster_path)
aligned_raster.rio.resolution () == reference_raster.rio.resolution()
# Returns False

I want the saved aligned raster to have the same bounds, resolution, and coordinates as the reference raster. Do I always have to set recalc_transform=False?

[snowman2]

Side note: You don't need to do assign_coords as reproject_match does that for you already.

What happens if you write the transform from the original raster to the new raster using rio.write_transform? documentation