Notebook

NOAA UFS Replay on Google Cloud Storage (GCS)¶

We have put the atmosphere and ocean model components from the UFS Replay on GCS, which includes a model run natively at 1/4 degree for ~30 years and at 1 degree for ~5.5 years. Additionally, for both components we have datasets that are spatially subsampled from 1/4 degree to approximately 1 degree by selecting every 4th point in latitude and longitude.

The atmosphere dataset is at 3 hour frequency, and ocean dataset is at 6 hour frequency. The time stamps selected correspond to the "analysis" portion of the Replay, coinciding with the IAU. While the atmosphere fields are snapshots, the ocean fields are a 6 hour temporal average over the IAU period.

The ocean fields were conservatively regridded from the native Tripolar grid to the same Gaussian grid as the atmosphere model output. The atmosphere fields were regridded from the cubed sphere FV3 grid to Gaussian grid.

Note that there are many land surface and some ice fields in the "atmosphere" (fv3) dataset.

Locations on GCS¶

One can browse the datasets here: https://console.cloud.google.com/storage/browser/noaa-ufs-gefsv13replay

Atmosphere¶

1/4 degree: gs://noaa-ufs-gefsv13replay/ufs-hr1/0.25-degree/03h-freq/zarr/fv3.zarr
1/4 degree subsampled: gs://noaa-ufs-gefsv13replay/ufs-hr1/0.25-degree-subsampled/03h-freq/zarr/fv3.zarr
1 degree: gs://noaa-ufs-gefsv13replay/ufs-hr1/1.00-degree/03h-freq/zarr/fv3.zarr

Ocean¶

1/4 degree: gs://noaa-ufs-gefsv13replay/ufs-hr1/0.25-degree/06h-freq/zarr/mom6.zarr
1/4 degree subsampled: gs://noaa-ufs-gefsv13replay/ufs-hr1/0.25-degree-subsampled/06h-freq/zarr/mom6.zarr
1 degree: gs://noaa-ufs-gefsv13replay/ufs-hr1/1.00-degree/06h-freq/zarr/mom6.zarr

⚠️ Warning about the Atmosphere Vertical Coordinate ⚠️

The atmosphere data is on the native model vertical coordinate. Despite the fact that the coordinate is named pfull and has units hPa, the vertical coordinate is not pressure levels. The coordinate refers to a "reference" pressure level, if the surface pressure at this particular point in time and horizontal space were 1000 hPa.

The python package ufs2arco can be used to interpolate from the native FV3 vertical coordinate to pressure levels. Please see this documentation page for example usage.

1/4 Degree Data¶

30 years (1994 - 2023) of 1/4 degree at 3 hour frequency for atmosphere fields and 6 hour frequency for ocean fields

In [1]:

import xarray as xr

In [2]:

atm = xr.open_zarr(
    "gs://noaa-ufs-gefsv13replay/ufs-hr1/0.25-degree/03h-freq/zarr/fv3.zarr",
    storage_options={"token": "anon"},
)

In [3]:

atm

Out[3]:

<xarray.Dataset> Size: 901TB
Dimensions:        (time: 87020, grid_yt: 768, grid_xt: 1536, pfull: 127)
Coordinates:
    cftime         (time) object 696kB dask.array<chunksize=(21755,), meta=np.ndarray>
    ftime          (time) timedelta64[ns] 696kB dask.array<chunksize=(21755,), meta=np.ndarray>
  * grid_xt        (grid_xt) float64 12kB 0.0 0.2344 0.4688 ... 359.5 359.8
  * grid_yt        (grid_yt) float64 6kB 89.82 89.59 89.35 ... -89.59 -89.82
  * pfull          (pfull) float32 508B 0.01278 0.02033 0.03177 ... 996.2 998.8
  * time           (time) datetime64[ns] 696kB 1993-12-31T18:00:00 ... 2023-1...
Data variables: (12/179)
    acond          (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    albdo_ave      (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    alnsf          (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    alnwf          (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    alvsf          (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    alvwf          (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    ...             ...
    xtts           (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    xu             (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    xv             (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    xz             (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    xzts           (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    zc             (time, grid_yt, grid_xt) float32 411GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
Attributes:
    abstract:         The NOAA Unified Forecast System (UFS) / Global Ensembl...
    acknowledgement:  To provide appropriate attribution under this license, ...
    description:      This zarr store is a subset of the original GEFSv13 UFS...
    license:          The Global Ensemble Forecast System version 13 (GEFSv13...

In [4]:

ocn = xr.open_zarr(
    "gs://noaa-ufs-gefsv13replay/ufs-hr1/0.25-degree/06h-freq/zarr/mom6.zarr",
    storage_options={"token": "anon"},
)

In [5]:

ocn

Out[5]:

<xarray.Dataset> Size: 80TB
Dimensions:    (time: 43826, lat: 768, lon: 1536, zl: 75, z_l: 75)
Coordinates:
    cftime     (time) object 351kB dask.array<chunksize=(21913,), meta=np.ndarray>
    ftime      (time) timedelta64[ns] 351kB dask.array<chunksize=(21913,), meta=np.ndarray>
  * lat        (lat) float64 6kB 89.82 89.59 89.35 ... -89.35 -89.59 -89.82
  * lon        (lon) float64 12kB 0.0 0.2344 0.4688 0.7031 ... 359.3 359.5 359.8
  * time       (time) datetime64[ns] 351kB 1993-12-31T18:00:00 ... 2023-12-31
  * z_l        (z_l) float64 600B 0.5154 1.571 2.687 ... 5.698e+03 5.902e+03
  * zl         (zl) float64 600B 1.0 3.0 5.0 ... 3.338e+03 3.591e+03 5.111e+03
Data variables: (12/19)
    Heat_PmE   (time, lat, lon) float32 207GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    LW         (time, lat, lon) float32 207GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    LwLatSens  (time, lat, lon) float32 207GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    SSH        (time, lat, lon) float32 207GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    SW         (time, lat, lon) float32 207GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    evap       (time, lat, lon) float32 207GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    ...         ...
    so         (time, z_l, lat, lon) float32 16TB dask.array<chunksize=(1, 1, 768, 1536), meta=np.ndarray>
    taux       (time, lat, lon) float32 207GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    tauy       (time, lat, lon) float32 207GB dask.array<chunksize=(1, 768, 1536), meta=np.ndarray>
    temp       (time, z_l, lat, lon) float32 16TB dask.array<chunksize=(1, 1, 768, 1536), meta=np.ndarray>
    uo         (time, z_l, lat, lon) float32 16TB dask.array<chunksize=(1, 1, 768, 1536), meta=np.ndarray>
    vo         (time, z_l, lat, lon) float32 16TB dask.array<chunksize=(1, 1, 768, 1536), meta=np.ndarray>
Attributes:
    abstract:         The NOAA Unified Forecast System (UFS) / Global Ensembl...
    acknowledgement:  To provide appropriate attribution under this license, ...
    description:      This zarr store is a subset of the original GEFSv13 UFS...
    license:          The Global Ensemble Forecast System version 13 (GEFSv13...

Data and chunksizes are as follows

3D Atmosphere fields

In [6]:

print("chunks = ", atm.tmp.encoding["preferred_chunks"])
atm.tmp.data

chunks =  {'time': 1, 'pfull': 1, 'grid_yt': 768, 'grid_xt': 1536}

Out[6]:

	Array	Chunk
Bytes	47.43 TiB	4.50 MiB
Shape	(87020, 127, 768, 1536)	(1, 1, 768, 1536)
Dask graph	11051540 chunks in 2 graph layers
Data type	float32 numpy.ndarray

3D Ocean fields

In [7]:

print("chunks = ", ocn.temp.encoding["preferred_chunks"])
ocn.temp.data

chunks =  {'time': 1, 'z_l': 1, 'lat': 768, 'lon': 1536}

Out[7]:

	Array	Chunk
Bytes	14.11 TiB	4.50 MiB
Shape	(43826, 75, 768, 1536)	(1, 1, 768, 1536)
Dask graph	3286950 chunks in 2 graph layers
Data type	float32 numpy.ndarray

2D Atmosphere fields

In [8]:

print("chunks = ", atm.tmp2m.encoding["preferred_chunks"])
atm.tmp2m.data

chunks =  {'time': 1, 'grid_yt': 768, 'grid_xt': 1536}

Out[8]:

	Array	Chunk
Bytes	382.41 GiB	4.50 MiB
Shape	(87020, 768, 1536)	(1, 768, 1536)
Dask graph	87020 chunks in 2 graph layers
Data type	float32 numpy.ndarray

2D Ocean fields

In [9]:

print("chunks = ", ocn.SSH.encoding["preferred_chunks"])
ocn.SSH.data

chunks =  {'time': 1, 'lat': 768, 'lon': 1536}

Out[9]:

	Array	Chunk
Bytes	192.59 GiB	4.50 MiB
Shape	(43826, 768, 1536)	(1, 768, 1536)
Dask graph	43826 chunks in 2 graph layers
Data type	float32 numpy.ndarray

Some example plots...

In [10]:

%%time
atm.tmp2m.isel(time=slice(8)).plot(col="time", col_wrap=4)

CPU times: user 1.84 s, sys: 296 ms, total: 2.13 s
Wall time: 2.86 s

Out[10]:

<xarray.plot.facetgrid.FacetGrid at 0x107c80390>

In [11]:

%%time
ocn.SSH.isel(time=slice(8)).plot(col="time", col_wrap=4)

CPU times: user 1.26 s, sys: 142 ms, total: 1.4 s
Wall time: 1.44 s

Out[11]:

<xarray.plot.facetgrid.FacetGrid at 0x1788ef650>

Subsampled 1/4 Degree Dataset¶

30 years (1994 - 2023) of 1/4 degree data, subsampled from the 1/4 degree data to ~1 degree resolution by selecting every 4th point in latitude/longitude (dimensions grid_yt/grid_xt in the atmosphere dataset)

In [12]:

atm = xr.open_zarr(
    "gs://noaa-ufs-gefsv13replay/ufs-hr1/0.25-degree-subsampled/03h-freq/zarr/fv3.zarr",
    storage_options={"token": "anon"},
)

In [13]:

atm

Out[13]:

<xarray.Dataset> Size: 60TB
Dimensions:        (time: 87020, grid_yt: 192, grid_xt: 384, pfull: 127)
Coordinates:
    cftime         (time) object 696kB dask.array<chunksize=(21755,), meta=np.ndarray>
    ftime          (time) timedelta64[ns] 696kB dask.array<chunksize=(21755,), meta=np.ndarray>
  * grid_xt        (grid_xt) float64 3kB 0.0 0.9375 1.875 ... 357.2 358.1 359.1
  * grid_yt        (grid_yt) float64 2kB 89.82 88.89 87.95 ... -88.18 -89.12
  * pfull          (pfull) float32 508B 0.01278 0.02033 0.03177 ... 996.2 998.8
  * time           (time) datetime64[ns] 696kB 1993-12-31T18:00:00 ... 2023-1...
Data variables: (12/182)
    acond          (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    albdo_ave      (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    alnsf          (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    alnwf          (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    alvsf          (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    alvwf          (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    ...             ...
    xtts           (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    xu             (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    xv             (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    xz             (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    xzts           (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    zc             (time, grid_yt, grid_xt) float32 26GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
Attributes:
    abstract:         The NOAA Unified Forecast System (UFS) / Global Ensembl...
    acknowledgement:  To provide appropriate attribution under this license, ...
    description:      This zarr store is a subset of the original GEFSv13 UFS...
    license:          The Global Ensemble Forecast System version 13 (GEFSv13...

In [14]:

ocn = xr.open_zarr(
    "gs://noaa-ufs-gefsv13replay/ufs-hr1/0.25-degree-subsampled/06h-freq/zarr/mom6.zarr",
    storage_options={"token": "anon"},
)

In [15]:

ocn

Out[15]:

<xarray.Dataset> Size: 5TB
Dimensions:    (time: 43826, lat: 192, lon: 384, zl: 75, z_l: 75)
Coordinates:
    cftime     (time) object 351kB dask.array<chunksize=(21913,), meta=np.ndarray>
    ftime      (time) timedelta64[ns] 351kB dask.array<chunksize=(21913,), meta=np.ndarray>
  * lat        (lat) float64 2kB 89.82 88.89 87.95 ... -87.25 -88.18 -89.12
  * lon        (lon) float64 3kB 0.0 0.9375 1.875 2.812 ... 357.2 358.1 359.1
  * time       (time) datetime64[ns] 351kB 1993-12-31T18:00:00 ... 2023-12-31
  * z_l        (z_l) float64 600B 0.5154 1.571 2.687 ... 5.698e+03 5.902e+03
  * zl         (zl) float64 600B 1.0 3.0 5.0 ... 3.338e+03 3.591e+03 5.111e+03
Data variables: (12/19)
    Heat_PmE   (time, lat, lon) float32 13GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    LW         (time, lat, lon) float32 13GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    LwLatSens  (time, lat, lon) float32 13GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    SSH        (time, lat, lon) float32 13GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    SW         (time, lat, lon) float32 13GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    evap       (time, lat, lon) float32 13GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    ...         ...
    so         (time, z_l, lat, lon) float32 969GB dask.array<chunksize=(1, 75, 192, 384), meta=np.ndarray>
    taux       (time, lat, lon) float32 13GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    tauy       (time, lat, lon) float32 13GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    temp       (time, z_l, lat, lon) float32 969GB dask.array<chunksize=(1, 75, 192, 384), meta=np.ndarray>
    uo         (time, z_l, lat, lon) float32 969GB dask.array<chunksize=(1, 75, 192, 384), meta=np.ndarray>
    vo         (time, z_l, lat, lon) float32 969GB dask.array<chunksize=(1, 75, 192, 384), meta=np.ndarray>
Attributes:
    abstract:         The NOAA Unified Forecast System (UFS) / Global Ensembl...
    acknowledgement:  To provide appropriate attribution under this license, ...
    description:      This zarr store is a subset of the original GEFSv13 UFS...
    license:          The Global Ensemble Forecast System version 13 (GEFSv13...

Data and chunksizes are as follows

3D Atmosphere fields

In [16]:

print("chunks = ", atm.tmp.encoding["preferred_chunks"])
atm.tmp.data

chunks =  {'time': 1, 'pfull': 127, 'grid_yt': 192, 'grid_xt': 384}

Out[16]:

	Array	Chunk
Bytes	2.96 TiB	35.72 MiB
Shape	(87020, 127, 192, 384)	(1, 127, 192, 384)
Dask graph	87020 chunks in 2 graph layers
Data type	float32 numpy.ndarray

3D Ocean fields

In [17]:

print("chunks = ", ocn.temp.encoding["preferred_chunks"])
ocn.temp.data

chunks =  {'time': 1, 'z_l': 75, 'lat': 192, 'lon': 384}

Out[17]:

	Array	Chunk
Bytes	902.79 GiB	21.09 MiB
Shape	(43826, 75, 192, 384)	(1, 75, 192, 384)
Dask graph	43826 chunks in 2 graph layers
Data type	float32 numpy.ndarray

2D Atmosphere fields

In [18]:

print("chunks = ", atm.tmp2m.encoding["preferred_chunks"])
atm.tmp2m.data

chunks =  {'time': 1, 'grid_yt': 192, 'grid_xt': 384}

Out[18]:

	Array	Chunk
Bytes	23.90 GiB	288.00 kiB
Shape	(87020, 192, 384)	(1, 192, 384)
Dask graph	87020 chunks in 2 graph layers
Data type	float32 numpy.ndarray

2D Ocean fields

In [19]:

print("chunks = ", ocn.SSH.encoding["preferred_chunks"])
ocn.SSH.data

chunks =  {'time': 1, 'lat': 192, 'lon': 384}

Out[19]:

	Array	Chunk
Bytes	12.04 GiB	288.00 kiB
Shape	(43826, 192, 384)	(1, 192, 384)
Dask graph	43826 chunks in 2 graph layers
Data type	float32 numpy.ndarray

Some example plots...

In [20]:

%%time
atm.tmp2m.isel(time=slice(8)).plot(col="time", col_wrap=4)

CPU times: user 471 ms, sys: 41.4 ms, total: 512 ms
Wall time: 660 ms

Out[20]:

<xarray.plot.facetgrid.FacetGrid at 0x178e63d10>

In [21]:

%%time
ocn.SSH.isel(time=slice(8)).plot(col="time", col_wrap=4)

CPU times: user 370 ms, sys: 17.1 ms, total: 387 ms
Wall time: 470 ms

Out[21]:

<xarray.plot.facetgrid.FacetGrid at 0x199855ad0>

1 Degree Data¶

5.5 years (1994 - 1999) of 1 degree at 3 hour frequency for atmosphere fields and 6 hour frequency for ocean fields

In [22]:

atm = xr.open_zarr(
    "gs://noaa-ufs-gefsv13replay/ufs-hr1/1.00-degree/03h-freq/zarr/fv3.zarr",
    storage_options={"token": "anon"},
)

In [23]:

atm

Out[23]:

<xarray.Dataset> Size: 11TB
Dimensions:        (time: 15916, grid_yt: 192, grid_xt: 384, pfull: 127)
Coordinates:
    cftime         (time) object 127kB dask.array<chunksize=(15916,), meta=np.ndarray>
    ftime          (time) timedelta64[ns] 127kB dask.array<chunksize=(15916,), meta=np.ndarray>
  * grid_xt        (grid_xt) float64 3kB 0.0 0.9375 1.875 ... 357.2 358.1 359.1
  * grid_yt        (grid_yt) float64 2kB 89.28 88.36 87.42 ... -88.36 -89.28
  * pfull          (pfull) float32 508B 0.01278 0.02033 0.03177 ... 996.2 998.8
  * time           (time) datetime64[ns] 127kB 1993-12-31T18:00:00 ... 1999-0...
Data variables: (12/182)
    acond          (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    albdo_ave      (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    alnsf          (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    alnwf          (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    alvsf          (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    alvwf          (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    ...             ...
    xtts           (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    xu             (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    xv             (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    xz             (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    xzts           (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    zc             (time, grid_yt, grid_xt) float32 5GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
Attributes:
    abstract:         The NOAA Unified Forecast System (UFS) / Global Ensembl...
    acknowledgement:  To provide appropriate attribution under this license, ...
    description:      This zarr store is a subset of the original GEFSv13 UFS...
    license:          The Global Ensemble Forecast System version 13 (GEFSv13...

In [24]:

ocn = xr.open_zarr(
    "gs://noaa-ufs-gefsv13replay/ufs-hr1/1.00-degree/06h-freq/zarr/mom6.zarr",
    storage_options={"token": "anon"},
)

In [25]:

ocn

Out[25]:

<xarray.Dataset> Size: 913GB
Dimensions:       (time: 7961, lat: 192, lon: 384, zl: 75, z_l: 75)
Coordinates:
    cftime        (time) object 64kB dask.array<chunksize=(7961,), meta=np.ndarray>
    ftime         (time) timedelta64[ns] 64kB dask.array<chunksize=(7961,), meta=np.ndarray>
  * lat           (lat) float64 2kB 89.28 88.36 87.42 ... -87.42 -88.36 -89.28
  * lon           (lon) float64 3kB 0.0 0.9375 1.875 2.812 ... 357.2 358.1 359.1
  * time          (time) datetime64[ns] 64kB 1993-12-31T18:00:00 ... 1999-06-...
  * z_l           (z_l) float64 600B 0.5154 1.571 2.687 ... 5.698e+03 5.902e+03
  * zl            (zl) float64 600B 1.0 3.0 5.0 ... 3.591e+03 5.111e+03
Data variables: (12/20)
    Heat_PmE      (time, lat, lon) float32 2GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    LW            (time, lat, lon) float32 2GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    LwLatSens     (time, lat, lon) float32 2GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    SSH           (time, lat, lon) float32 2GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    SW            (time, lat, lon) float32 2GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    evap          (time, lat, lon) float32 2GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    ...            ...
    so            (time, z_l, lat, lon) float32 176GB dask.array<chunksize=(1, 75, 192, 384), meta=np.ndarray>
    taux          (time, lat, lon) float32 2GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    tauy          (time, lat, lon) float32 2GB dask.array<chunksize=(1, 192, 384), meta=np.ndarray>
    temp          (time, z_l, lat, lon) float32 176GB dask.array<chunksize=(1, 75, 192, 384), meta=np.ndarray>
    uo            (time, z_l, lat, lon) float32 176GB dask.array<chunksize=(1, 75, 192, 384), meta=np.ndarray>
    vo            (time, z_l, lat, lon) float32 176GB dask.array<chunksize=(1, 75, 192, 384), meta=np.ndarray>
Attributes:
    abstract:         The NOAA Unified Forecast System (UFS) / Global Ensembl...
    acknowledgement:  To provide appropriate attribution under this license, ...
    description:      This zarr store is a subset of the original GEFSv13 UFS...
    license:          The Global Ensemble Forecast System version 13 (GEFSv13...

Data and chunksizes are as follows

3D Atmosphere fields

In [26]:

print("chunks = ", atm.tmp.encoding["preferred_chunks"])
atm.tmp.data

chunks =  {'time': 1, 'pfull': 127, 'grid_yt': 192, 'grid_xt': 384}

Out[26]:

	Array	Chunk
Bytes	555.18 GiB	35.72 MiB
Shape	(15916, 127, 192, 384)	(1, 127, 192, 384)
Dask graph	15916 chunks in 2 graph layers
Data type	float32 numpy.ndarray

3D Ocean fields

In [27]:

print("chunks = ", ocn.temp.encoding["preferred_chunks"])
ocn.temp.data

chunks =  {'time': 1, 'z_l': 75, 'lat': 192, 'lon': 384}

Out[27]:

	Array	Chunk
Bytes	163.99 GiB	21.09 MiB
Shape	(7961, 75, 192, 384)	(1, 75, 192, 384)
Dask graph	7961 chunks in 2 graph layers
Data type	float32 numpy.ndarray

2D Atmosphere fields

In [28]:

print("chunks = ", atm.tmp2m.encoding["preferred_chunks"])
atm.tmp2m.data

chunks =  {'time': 1, 'grid_yt': 192, 'grid_xt': 384}

Out[28]:

	Array	Chunk
Bytes	4.37 GiB	288.00 kiB
Shape	(15916, 192, 384)	(1, 192, 384)
Dask graph	15916 chunks in 2 graph layers
Data type	float32 numpy.ndarray

2D Ocean fields

In [29]:

print("chunks = ", ocn.SSH.encoding["preferred_chunks"])
ocn.SSH.data

chunks =  {'time': 1, 'lat': 192, 'lon': 384}

Out[29]:

	Array	Chunk
Bytes	2.19 GiB	288.00 kiB
Shape	(7961, 192, 384)	(1, 192, 384)
Dask graph	7961 chunks in 2 graph layers
Data type	float32 numpy.ndarray

Some example plots...

In [30]:

%%time
atm.tmp2m.isel(time=slice(8)).plot(col="time", col_wrap=4)

CPU times: user 494 ms, sys: 44.1 ms, total: 538 ms
Wall time: 768 ms

Out[30]:

<xarray.plot.facetgrid.FacetGrid at 0x178b05550>

In [31]:

%%time
ocn.SSH.isel(time=slice(8)).plot(col="time", col_wrap=4)

CPU times: user 341 ms, sys: 20.6 ms, total: 362 ms
Wall time: 409 ms

Out[31]:

<xarray.plot.facetgrid.FacetGrid at 0x19f0bebd0>

In [ ]: