can xradar read Cfradial2 netcdf files?

[aladinor]

Hi all. I just got a file in "cfradial2" format. It contains groups instead a single netcdf. Do we need to create a new backend for this kind of files or we can addapt cfradial1 backend. Attached, you will find a copy of the file

import xradar as xd

path = "radar-cfradial2.nc"
radar = xd.io.open_cfradial1_datatree(path)

returns this error

Traceback (most recent call last):
  File "/media/alfonso/drive/Alfonso/python/demos/era5-ingestion/truweather.py", line 23, in <module>
    radar = xd.io.open_cfradial1_datatree(path)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/alfonso/mambaforge/envs/raw2zarr-dev/lib/python3.12/site-packages/xradar/io/backends/cfradial1.py", line 355, in open_cfradial1_datatree
    "/": _get_required_root_dataset(ds, optional=optional),
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/alfonso/mambaforge/envs/raw2zarr-dev/lib/python3.12/site-packages/xradar/io/backends/cfradial1.py", line 82, in _get_required_root_dataset
    root = root.rename_vars(
           ^^^^^^^^^^^^^^^^^
  File "/home/alfonso/mambaforge/envs/raw2zarr-dev/lib/python3.12/site-packages/xarray/core/dataset.py", line 4234, in rename_vars
    raise ValueError(
ValueError: cannot rename 'sweep_number' because it is not a variable or coordinate in this dataset

this is because is a Netcdf file with nested groups. we can check it out using xarray.open_groups or xarray.open_datatree

groups = xr.open_groups(path)

print(groups)

{'/': <xarray.Dataset> Size: 40B
Dimensions:           (nsweep: 1)
Dimensions without coordinates: nsweep
Data variables:
    sweep_group_name  (nsweep) <U10 40B ...
Attributes:
    Conventions:  CF-1.7, Cf/Radial
    version:      2.0
    title:        Radar CfRadial2 Dataset
    history:      Updated CfRadial2 on 2025-05-01T00:46:09.040110Z
    RadarName:    denton.tx
    Latitude:     33.25355
    Longitude:    -97.15204
    Height:       224.0, 

'/sweep_0000': <xarray.Dataset> Size: 27MB
Dimensions:                       (ntime: 723, nrange: 1003)
Dimensions without coordinates: ntime, nrange
Data variables: (12/19)
    time_us                       (ntime) datetime64[ns] 6kB ...
    range_m                       (nrange) float64 8kB ...
    azimuth_deg                   (ntime) float64 6kB ...
    elevation_deg                 (ntime) float64 6kB ...
    range_gate_width_mm           (ntime) float64 6kB ...
    start_range_mm                (ntime) float64 6kB ...
    ...                            ...
    specific_differential_phase   (ntime, nrange) float32 3MB ...
    radial_velocity_mps           (ntime, nrange) float32 3MB ...
    signal_to_noise_ratio_db      (ntime, nrange) float32 3MB ...
    spectral_width_mps            (ntime, nrange) float32 3MB ...
    filtering                     (ntime, nrange) int8 725kB ...
    rain_rate_mmhr                (ntime, nrange) float32 3MB ...
Attributes:
    NumGates:          1003
    ScanType:          2
    AntennaGain:       33.0
    AntennaBeamwidth:  2.7
    PosVelocity:       15.0}

Let me know your thoughts.

File can be found in this link

[kmuehlbauer]

In general, cfradial2 should be read by xarray.open_datatree. No additional backend needed, IMHO.

[mgrover1]

I would be in favor of having a

open_cfradial2_datatree(path)

function under the io module, which calls xarray.open_datatree for consistency?

[syedhamidali]

I agree with @mgrover1, I can quickly create one, since it is pretty straight forward.

[kmuehlbauer]

https://github.com/openradar/xradar/issues/93

Seems like still a good idea to do it anyway. At least as compatibility layer and for consistency as @mgrover1 suggested.

@syedhamidali please go ahead, if you have the bandwidth. :+1:

[aladinor]

I will paste here my comments under #288.

Hi everyone,

Thanks for putting all of this together. I was looking at the cfradial2 data, and I can see it has a hierarchical structure with a root dataset as well as the sweep nodes. However, is this file fully FM301 compliant? I notice something like the root contains some attributes, but does it contain all the required attributes under the FM301 data model?

print(dtree['/'].ds)

<xarray.Dataset> Size: 40B
Dimensions:           (nsweep: 1)
Dimensions without coordinates: nsweep
Data variables:
    sweep_group_name  (nsweep) <U10 40B ...
Attributes:
    Conventions:  CF-1.7, Cf/Radial
    version:      2.0
    title:        Radar CfRadial2 Dataset
    history:      Updated CfRadial2 on 2025-05-01T00:46:09.040110Z
    RadarName:    denton.tx
    Latitude:     33.25355
    Longitude:    -97.15204
    Height:       224.0

On the other hand, I realized the dataset contains sweeps numbered as sweep_xxxx, is this the "correct" way to number the sweep under the WMO data model? Finally, I also realized this dataset contains coordinates and dimensions that I am not sure are cfradrial-FM301 compliant.

print(dtree['/sweep_0000']. ds 

<xarray.Dataset> Size: 27MB
Dimensions:                       (ntime: 723, nrange: 1003)
Dimensions without coordinates: ntime, nrange
Data variables: (12/19)
    time_us                       (ntime) datetime64[ns] 6kB ...
    range_m                       (nrange) float64 8kB ...
    azimuth_deg                   (ntime) float64 6kB ...
    elevation_deg                 (ntime) float64 6kB ...
    range_gate_width_mm           (ntime) float64 6kB ...
    start_range_mm                (ntime) float64 6kB ...
    ...                            ...
    specific_differential_phase   (ntime, nrange) float32 3MB ...
    radial_velocity_mps           (ntime, nrange) float32 3MB ...
    signal_to_noise_ratio_db      (ntime, nrange) float32 3MB ...
    spectral_width_mps            (ntime, nrange) float32 3MB ...
    filtering                     (ntime, nrange) int8 725kB ...
    rain_rate_mmhr                (ntime, nrange) float32 3MB ...
Attributes:
    NumGates:          1003
    ScanType:          2
    AntennaGain:       33.0
    AntennaBeamwidth:  2.7
    PosVelocity:       15.0

Please let me know your thoughts