Transform ARGO-KE products into DGGS (Healpix)

Transform ARGO-KE products into DGGS (Healpix)#

Context#

Purpose#

The goal is to have the ARGO-KE into DGGS (healpix grid).

Description#

In this notebook, we will:

  • Open the timemean-velocity-product (Gray, Wortham, Kihle, TBD)

  • Transform the product into mean kinetic energy, MKE

  • Perform the regridding operation and visualize the regridded data with xdggs.

Contributions#

Notebook#

  • Ola Formo Kihle (author), Independent Consultant / UW Contractor, (Norway), @ofk123

  • Even Moa Myklebust (reviewer), Simula Research Laboratory (Norway), @evenmm

  • Anne Fouilloux (reviewer), Simula Research Laboratory (Norway), @annefou

Bibliography and other interesting resources#

Install packages#

  • If xdggs and/or xarray-healpy are not available in your Python environment, uncomment the cells below to install these packages.

#!pip install git+https://github.com/xarray-contrib/xdggs.git

#!pip install git+https://github.com/IAOCEA/xarray-healpy.git

Install Libraries#

import xarray as xr, numpy as np
import s3fs
import xdggs
#_ = xr.set_options(display_expand_data=False)
_ = xr.set_options(display_expand_data=False, display_expand_attrs=False, keep_attrs=True)

Open ARGO-KE data product#

  • The data sample is stored in a bucket on PANGEO-EOSC and is readable by anyone.

s3 = s3fs.S3FileSystem(
    anon=True,
    endpoint_url='https://pangeo-eosc-minioapi.vm.fedcloud.eu/'
)
s3_bucket= 'okihle-workathon/'
store_s3 = s3fs.S3Map(
    root=s3_bucket + 'argo-vel-data',
    s3=s3,
    check=False
)
ds = xr.open_zarr(store_s3, consolidated=True)

Visualise Mean KE#

np.log(1e4 * 0.5 * (ds.u**2 + ds.v**2)).compute().plot(
    figsize=(9, 5.5), cmap="viridis", vmin=-2, vmax=7
)

<matplotlib.collections.QuadMesh at 0x70922b2f2c00>
_images/4c1676e209ca067fb97905bffac6f04f59be0db27f07326ca65ca18442002326.png

Prepare Healpix grid for regridding#

nside = 64

def center_longitude(ds, center):
    if center == 0:
        centered = (ds.longitude + 180) % 360 - 180
    else:
        centered = (ds.longitude - 180) % 360 + 180
    return ds.assign_coords(longitude=centered)

from xarray_healpy import HealpyGridInfo, HealpyRegridder

# Define the target Healpix grid information
grid = HealpyGridInfo(level=int(np.log2(nside)))
target_grid = grid.target_grid(ds).pipe(center_longitude, 0).drop_attrs(deep=False).dggs.decode({"grid_name": "healpix", "nside": nside, "indexing_scheme": "nested"})

Compute the interpolation weights for regridding the diff data#

%%time
regridder = HealpyRegridder(
    ds[["longitude", "latitude"]].compute(),
    target_grid,
    method="bilinear",
    interpolation_kwargs={},
)
regridder

CPU times: user 4.24 s, sys: 185 ms, total: 4.42 s
Wall time: 4.43 s

HealpyRegridder(method='bilinear', interpolation_kwargs={})

Perform the regridding operation using the computed interpolation weights.#

%%time
regridded = regridder.regrid_ds(ds).pipe(xdggs.decode)

CPU times: user 1.34 s, sys: 568 ms, total: 1.9 s
Wall time: 1.92 s

Visualise regridded data with xdggs#

da = np.log(1e4 * 0.5 * (regridded.u**2 + regridded.v**2)).compute().squeeze()
da.dggs.explore()