Add DUACS daily->5-day-average preprocessing script

[alxmrs]

What

New ocean_preprocessing.make_duacs_5day utility that coarsens the daily (P1D) global gridded DUACS L4 altimetry product to non-overlapping consecutive 5-day means, mirroring the OM4 5-daily convention so the observations can sit alongside the emulator inputs.

How

• Binning: coarsen(time=5, boundary="trim").mean() — drops the trailing 1-day remainder → 366 → 73 steps. Each window is labelled by its midpoint timestamp.
• Variables: core ocean fields adt, sla, ugos, ugosa, vgos, vgosa + flag_ice re-derived as a 0–1 ice-presence fraction (metadata relabelled). err_* and tpa_correction dropped.
• Output: uncompressed float32, one chunk per timestep, written to s3://emulators/am16581/data/2026-06/...P5D...zarr (P1D→P5D name swap).
• Compute: reuses the OM4 pipeline's init_cluster, blosc single-thread guards, and retry-on-write. Coiled by default; OCEAN_DUACS_CLUSTER=local to run locally. Native time chunk (50) is a clean multiple of the window → no cross-chunk shuffle.
• Native DUACS naming/grid preserved (latitude/longitude, −180…180, 0.125°) — not conformed to the emulator x/y/0–360 schema.

Validation

--dry_run against the live store: structure correct, mid-grid sample yields physically sensible values (adt ≈ 0.43 m, sla ≈ 0.045 m, velocities m/s). Dry-run output (final dataset repr + full ds.time) to be pasted in a comment below for review.

Tests

tests/test_make_duacs_5day.py — block-mean correctness, ice-flag→fraction, missing-var guard.

Run:

cd data/
export FSSPEC_S3_ENDPOINT_URL=https://nyu1.osn.mghpcc.org/
export AWS_ACCESS_KEY_ID=...  AWS_SECRET_ACCESS_KEY=...
OCEAN_DUACS_CLUSTER=local python -m ocean_preprocessing.make_duacs_5day --dry_run

🤖 Generated with Claude Code

[alxmrs]

Dry-run output

OCEAN_DUACS_CLUSTER=local python -m ocean_preprocessing.make_duacs_5day --dry_run against the live source store (nothing written). Reduction validated against a mid-grid sample.

Final dataset (out)

<xarray.Dataset> Size: 8GB
Dimensions:    (time: 73, latitude: 1440, longitude: 2880)
Coordinates:
  * time       (time) datetime64[ns] 584B 2022-01-03 2022-01-08 ... 2022-12-29
  * latitude   (latitude) float32 6kB -89.94 -89.81 -89.69 ... 89.69 89.81 89.94
  * longitude  (longitude) float32 12kB -179.9 -179.8 -179.7 ... 179.8 179.9
Data variables:
    adt        (time, latitude, longitude) float32 1GB dask.array<chunksize=(1, 1440, 2880), meta=np.ndarray>
    sla        (time, latitude, longitude) float32 1GB dask.array<chunksize=(1, 1440, 2880), meta=np.ndarray>
    ugos       (time, latitude, longitude) float32 1GB dask.array<chunksize=(1, 1440, 2880), meta=np.ndarray>
    ugosa      (time, latitude, longitude) float32 1GB dask.array<chunksize=(1, 1440, 2880), meta=np.ndarray>
    vgos       (time, latitude, longitude) float32 1GB dask.array<chunksize=(1, 1440, 2880), meta=np.ndarray>
    vgosa      (time, latitude, longitude) float32 1GB dask.array<chunksize=(1, 1440, 2880), meta=np.ndarray>
    flag_ice   (time, latitude, longitude) float32 1GB dask.array<chunksize=(1, 1440, 2880), meta=np.ndarray>
Attributes: (12/48)
    Conventions:                       CF-1.6
    Metadata_Conventions:              Unidata Dataset Discovery v1.0
    cdm_data_type:                     Grid
    comment:                           Sea Surface Height measured by Altimet...
    contact:                           servicedesk.cmems@mercator-ocean.eu
    copernicusmarine_version:          2.3.0
    ...                                ...
    title:                             DT merged all satellites Global Ocean ...
    m2lines/source_dataset:            s3://emulators/jr7309/data/duacs/cmems...
    m2lines/temporal_coarsening:       5-day consecutive simple mean (coarsen...
    m2lines/ocean_emulators_git_hash:  https://github.com/Open-Athena/Ocean_E...
    m2lines/date_created:              2026-06-05T11:18:07.434449
    m2lines/cli_args:                  /Users/alxmrs/git/Ocean_Emulator/.clau...

ds.time (73 window-center labels, 5 days apart)

<xarray.DataArray 'time' (time: 73)> Size: 584B
array(['2022-01-03T00:00:00.000000000', '2022-01-08T00:00:00.000000000',
       '2022-01-13T00:00:00.000000000', '2022-01-18T00:00:00.000000000',
       '2022-01-23T00:00:00.000000000', '2022-01-28T00:00:00.000000000',
       '2022-02-02T00:00:00.000000000', '2022-02-07T00:00:00.000000000',
       '2022-02-12T00:00:00.000000000', '2022-02-17T00:00:00.000000000',
       '2022-02-22T00:00:00.000000000', '2022-02-27T00:00:00.000000000',
       '2022-03-04T00:00:00.000000000', '2022-03-09T00:00:00.000000000',
       '2022-03-14T00:00:00.000000000', '2022-03-19T00:00:00.000000000',
       '2022-03-24T00:00:00.000000000', '2022-03-29T00:00:00.000000000',
       '2022-04-03T00:00:00.000000000', '2022-04-08T00:00:00.000000000',
       '2022-04-13T00:00:00.000000000', '2022-04-18T00:00:00.000000000',
       '2022-04-23T00:00:00.000000000', '2022-04-28T00:00:00.000000000',
       '2022-05-03T00:00:00.000000000', '2022-05-08T00:00:00.000000000',
       '2022-05-13T00:00:00.000000000', '2022-05-18T00:00:00.000000000',
       '2022-05-23T00:00:00.000000000', '2022-05-28T00:00:00.000000000',
       '2022-06-02T00:00:00.000000000', '2022-06-07T00:00:00.000000000',
       '2022-06-12T00:00:00.000000000', '2022-06-17T00:00:00.000000000',
       '2022-06-22T00:00:00.000000000', '2022-06-27T00:00:00.000000000',
       '2022-07-02T00:00:00.000000000', '2022-07-07T00:00:00.000000000',
       '2022-07-12T00:00:00.000000000', '2022-07-17T00:00:00.000000000',
       '2022-07-22T00:00:00.000000000', '2022-07-27T00:00:00.000000000',
       '2022-08-01T00:00:00.000000000', '2022-08-06T00:00:00.000000000',
       '2022-08-11T00:00:00.000000000', '2022-08-16T00:00:00.000000000',
       '2022-08-21T00:00:00.000000000', '2022-08-26T00:00:00.000000000',
       '2022-08-31T00:00:00.000000000', '2022-09-05T00:00:00.000000000',
       '2022-09-10T00:00:00.000000000', '2022-09-15T00:00:00.000000000',
       '2022-09-20T00:00:00.000000000', '2022-09-25T00:00:00.000000000',
       '2022-09-30T00:00:00.000000000', '2022-10-05T00:00:00.000000000',
       '2022-10-10T00:00:00.000000000', '2022-10-15T00:00:00.000000000',
       '2022-10-20T00:00:00.000000000', '2022-10-25T00:00:00.000000000',
       '2022-10-30T00:00:00.000000000', '2022-11-04T00:00:00.000000000',
       '2022-11-09T00:00:00.000000000', '2022-11-14T00:00:00.000000000',
       '2022-11-19T00:00:00.000000000', '2022-11-24T00:00:00.000000000',
       '2022-11-29T00:00:00.000000000', '2022-12-04T00:00:00.000000000',
       '2022-12-09T00:00:00.000000000', '2022-12-14T00:00:00.000000000',
       '2022-12-19T00:00:00.000000000', '2022-12-24T00:00:00.000000000',
       '2022-12-29T00:00:00.000000000'], dtype='datetime64[ns]')
Coordinates:
  * time     (time) datetime64[ns] 584B 2022-01-03 2022-01-08 ... 2022-12-29
Attributes:
    axis:           T
    long_name:      Time
    standard_name:  time
    unit_long:      Days Since 1950-01-01

Sample values (mid-grid 64×64×2 corner)

  adt: min=0.3831 max=0.4754 mean=0.4337
  sla: min=0.02266 max=0.1084 mean=0.0455
  ugos: min=-0.4018 max=0.2412 mean=-0.02374
  ugosa: min=-0.3034 max=0.1707 mean=-0.01237
  vgos: min=-0.2278 max=0.1878 mean=-0.04271
  vgosa: min=-0.1528 max=0.06514 mean=-0.04667

flag_ice is 0 in that equatorial corner (no ice), as expected. Absolute fields (adt/ugos/vgos) are legitimately NaN near the poles — no MDT reference under the ice — while the anomaly fields retain values there.