Fix z_l->lev rename for OM4 snapshot sources

[alxmrs]

The 5-daily snapshot source (1958-2022.OM4p25_5daily_snapshots) exposes the depth-center coordinate as z_l but has no z_i. The old rename gated the z_l->lev rename on z_i being present, so z_l was left intact and downstream rechunking on "lev" failed with a ValueError (issue #450).

Extract normalize_vertical_coords() to rename z_l->lev and z_i->ilev independently, and key the interface-info branch on ilev rather than z_i. Behavior is unchanged for the averaged sources (which carry z_i).

🤖

[alxmrs]

I ran this script on with the small data setting and wrote the results to my machine. Here is the dataset it produced:

>>> ds
<xarray.Dataset> Size: 211MB
Dimensions:    (time: 10, y: 180, x: 360)
Coordinates:
  * time       (time) object 80B 1958-01-06 00:00:00 ... 1958-02-20 00:00:00
  * x          (x) float64 3kB 0.5 1.5 2.5 3.5 4.5 ... 356.5 357.5 358.5 359.5
  * y          (y) float64 1kB -89.24 -88.25 -87.25 -86.26 ... 87.25 88.25 89.24
Data variables: (12/100)
    hfds       (time, y, x) float32 3MB dask.array<chunksize=(1, 180, 360), meta=np.ndarray>
    mask_0     (y, x) bool 65kB dask.array<chunksize=(180, 360), meta=np.ndarray>
    mask_1     (y, x) bool 65kB dask.array<chunksize=(180, 360), meta=np.ndarray>
    mask_10    (y, x) bool 65kB dask.array<chunksize=(180, 360), meta=np.ndarray>
    mask_11    (y, x) bool 65kB dask.array<chunksize=(180, 360), meta=np.ndarray>
    mask_12    (y, x) bool 65kB dask.array<chunksize=(180, 360), meta=np.ndarray>
    ...         ...
    vo_6       (time, y, x) float32 3MB dask.array<chunksize=(1, 180, 360), meta=np.ndarray>
    vo_7       (time, y, x) float32 3MB dask.array<chunksize=(1, 180, 360), meta=np.ndarray>
    vo_8       (time, y, x) float32 3MB dask.array<chunksize=(1, 180, 360), meta=np.ndarray>
    vo_9       (time, y, x) float32 3MB dask.array<chunksize=(1, 180, 360), meta=np.ndarray>
    wfo        (time, y, x) float32 3MB dask.array<chunksize=(1, 180, 360), meta=np.ndarray>
    zos        (time, y, x) float32 3MB dask.array<chunksize=(1, 180, 360), meta=np.ndarray>
Attributes: (12/13)
    hfds:                              {'cell_measures': 'area: areacello', '...
    m2lines/cli_args:                  /Users/alxmrs/git/Ocean_Emulator/data/...
    m2lines/date_created:              2026-06-04T14:11:01.542698
    m2lines/ocean_emulators_git_hash:  https://github.com/Open-Athena/Ocean_E...
    regrid_method:                     conservative
    so:                                {'cell_measures': 'area: areacello', '...
    ...                                ...
    tauvo:                             {'cell_methods': 'yq:point xh:mean tim...
    thetao:                            {'cell_measures': 'area: areacello', '...
    uo:                                {'cell_methods': 'z_l:mean yh:mean xq:...
    vo:                                {'cell_methods': 'z_l:mean yq:point xh...
    wfo:                               {'cell_measures': 'area: areacello', '...
    zos:                               {'cell_measures': 'area: areacello', '...
>>> ds.time
<xarray.DataArray 'time' (time: 10)> Size: 80B
array([cftime.DatetimeJulian(1958, 1, 6, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 1, 11, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 1, 16, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 1, 21, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 1, 26, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 1, 31, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 2, 5, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 2, 10, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 2, 15, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeJulian(1958, 2, 20, 0, 0, 0, 0, has_year_zero=False)],
      dtype=object)
Coordinates:
  * time     (time) object 80B 1958-01-06 00:00:00 ... 1958-02-20 00:00:00

Here is the command I used:

>>> ds.attrs["m2lines/cli_args"]
'/Users/alxmrs/git/Ocean_Emulator/data/ocean_preprocessing/__main__.py om4 s3://emulators/wg2437/1958-2022.OM4p25_5daily_snapshots s3://m2lines-pubs/FOMO/raw/ocean_static_no_mask_table.zarr s3://m2lines-pubs/FOMO/raw/grids/ocean_hgrid.zarr s3://m2lines-pubs/FOMO/raw/grids/gaussian_grid_180_by_360.zarr --output_path=/tmp/om4_onedeg_snap_test.zarr --skip_spatial_filtering --skip_validation --small_run --cluster=local --n_workers=4'

[alxmrs]

@codex may I have your review?

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Keep it up!

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".