In-cloud vertical motion retrieval

Florian Poydenot (florian.poydenot@mpimet.mpg.de), August 2026

This folder contains a retrieval of vertical wind inside precipitating shallow clouds, obtained from a radar-based estimate combined with lidar observations. The vertical motion is retrieved from the slowest-moving edge of the radar spectrum, then corrected for SNR and turbulence broadening, following Zhu et al. (2021). For a detailed description of the processing, see Poydenot et al. (2026). The resulting data has 30 m/2 s resolution, with uncertainty of 0.2 m/s above cloud base.

The processing code is available at https://zenodo.org/records/19221500.

Instruments used

Ka-band radars

• MBR2: August 2019-January 2023

• MBRS: January 2023-December 2025

Lidars

• WindLidar1: vertically-pointing mode between March 2016 and October 2019, then between July 2021 and September 2024, and again since September 2025

• WindLidar2: VAD scans between March 2019 and July 2021, vertically-pointing mode between September 2024 and September 2025

How to use

The vertical coordinate is the radar range. Do not use the retrieved winds in deep convective clouds as the retrieval is not suited for this regime (use the provided cloud masks to filter it out).

Data Availability

The processing is not integrated (yet) into our automated workflows, and as a result, the dataset does not grow by the day but has an end date determined by when the last round of manual processing was run.

import intake

cat = intake.open_catalog("https://tcodata.mpimet.mpg.de/catalog.yaml")
ds = cat.BCO.verticalwind.to_dask()
mbrs = cat.BCO.mbrs.to_dask().sel(range=slice(None, ds.range.values[-1]))
ds

/builds/tco/bco/docs/.venv/lib/python3.14/site-packages/intake_xarray/base.py:21: FutureWarning: The return type of `Dataset.dims` will be changed to return a set of dimension names in future, in order to be more consistent with `DataArray.dims`. To access a mapping from dimension names to lengths, please use `Dataset.sizes`.
  'dims': dict(self._ds.dims),

/builds/tco/bco/docs/.venv/lib/python3.14/site-packages/intake_xarray/base.py:21: FutureWarning: The return type of `Dataset.dims` will be changed to return a set of dimension names in future, in order to be more consistent with `DataArray.dims`. To access a mapping from dimension names to lengths, please use `Dataset.sizes`.
  'dims': dict(self._ds.dims),

<xarray.Dataset> Size: 239GB
Dimensions:               (time: 55424828, range: 187)
Coordinates:
  * time                  (time) datetime64[ns] 443MB 2019-08-01T00:00:14 ......
  * range                 (range) float32 748B 155.9 187.1 ... 5.955e+03
    alt                   float32 4B ...
    lat                   float32 4B ...
    lon                   float32 4B ...
Data variables:
    allprecip_mask        (time, range) bool 10GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    drizzle_mask          (time, range) bool 10GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    id_mask               (time, range) int64 83GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    longlived_rain_mask   (time, range) bool 10GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    nonprecip_mask        (time, range) bool 10GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    rain_mask             (time, range) bool 10GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    shortlived_rain_mask  (time, range) bool 10GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    unclassified_mask     (time, range) bool 10GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    w                     (time, range) float32 41GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
    w_radar               (time, range) float32 41GB dask.array<chunksize=(131072, 32), meta=np.ndarray>
Attributes: (12/18)
    Conventions:           CF-1.12
    _logical_cutoff_date:  2024-09-29T00:00:00Z
    author:                Florian Poydenot (florian.poydenot@uni-hamburg.de)
    bcoproc_version:       0.0.0.post2156.dev0+d4f3d45
    converted_by:          Florian Poydenot (florian.poydenot@uni-hamburg.de)
    created_with:          make_final_data_netcdf.py
    ...                    ...
    sensor_alt:            18.5m
    source:                Ka-Band radar and Doppler wind lidar observations
    source_file:           /work/um1487/OBS/SCORE/Level_1/B_Reflectivity/Ka-B...
    title:                 Composite lidar-radar retrieval of the cloud-scale...
    tool_versions:         {"Python": "3.11.2 (main, Apr 28 2025, 14:11:48) [...
    version:               git-v1.3-132-gfbcc69d

xarray.Dataset

• Dimensions:
  • time: 55424828
  • range: 187
• Coordinates: (5)
  • time
    (time)
    datetime64[ns]
    2019-08-01T00:00:14 ... 2024-09-...

    axis :

    T

    standard_name :

    time

    units_metadata :

    leap_seconds: utc

    array(['2019-08-01T00:00:14.000000000', '2019-08-01T00:00:16.000000000',
           '2019-08-01T00:00:18.000000000', ..., '2024-09-28T23:59:59.000000000',
           '2024-09-29T00:00:01.000000000', '2024-09-29T00:00:03.000000000'],
          shape=(55424828,), dtype='datetime64[ns]')

  • range
    (range)
    float32
    155.9 187.1 ... 5.924e+03 5.955e+03

    axis :

    Z

    long_name :

    range from antenna to the center of each range gate

    standard_name :

    height

    units :

    m

    array([ 155.9 ,  187.08,  218.25,  249.43,  280.61,  311.79,  342.97,  374.15,
            405.33,  436.51,  467.69,  498.87,  530.05,  561.23,  592.4 ,  623.58,
            654.76,  685.94,  717.12,  748.3 ,  779.48,  810.66,  841.84,  873.02,
            904.2 ,  935.38,  966.56,  997.73, 1028.91, 1060.09, 1091.27, 1122.45,
           1153.63, 1184.81, 1215.99, 1247.17, 1278.35, 1309.53, 1340.71, 1371.88,
           1403.06, 1434.24, 1465.42, 1496.6 , 1527.78, 1558.96, 1590.14, 1621.32,
           1652.5 , 1683.68, 1714.86, 1746.04, 1777.21, 1808.39, 1839.57, 1870.75,
           1901.93, 1933.11, 1964.29, 1995.47, 2026.65, 2057.83, 2089.01, 2120.19,
           2151.36, 2182.54, 2213.72, 2244.9 , 2276.08, 2307.26, 2338.44, 2369.62,
           2400.8 , 2431.98, 2463.16, 2494.34, 2525.52, 2556.69, 2587.87, 2619.05,
           2650.23, 2681.41, 2712.59, 2743.77, 2774.95, 2806.13, 2837.31, 2868.49,
           2899.67, 2930.84, 2962.02, 2993.2 , 3024.38, 3055.56, 3086.74, 3117.92,
           3149.1 , 3180.28, 3211.46, 3242.64, 3273.82, 3305.  , 3336.17, 3367.35,
           3398.53, 3429.71, 3460.89, 3492.07, 3523.25, 3554.43, 3585.61, 3616.79,
           3647.97, 3679.15, 3710.32, 3741.5 , 3772.68, 3803.86, 3835.04, 3866.22,
           3897.4 , 3928.58, 3959.76, 3990.94, 4022.12, 4053.3 , 4084.48, 4115.65,
           4146.83, 4178.01, 4209.19, 4240.37, 4271.55, 4302.73, 4333.91, 4365.09,
           4396.27, 4427.45, 4458.63, 4489.8 , 4520.98, 4552.16, 4583.34, 4614.52,
           4645.7 , 4676.88, 4708.06, 4739.24, 4770.42, 4801.6 , 4832.78, 4863.96,
           4895.13, 4926.31, 4957.49, 4988.67, 5019.85, 5051.03, 5082.21, 5113.39,
           5144.57, 5175.75, 5206.93, 5238.11, 5269.28, 5300.46, 5331.64, 5362.82,
           5394.  , 5425.18, 5456.36, 5487.54, 5518.72, 5549.9 , 5581.08, 5612.26,
           5643.44, 5674.61, 5705.79, 5736.97, 5768.15, 5799.33, 5830.51, 5861.69,
           5892.87, 5924.05, 5955.23], dtype=float32)

  • alt
    ()
    float32
    ...

    [1 values with dtype=float32]

  • lat
    ()
    float32
    ...

    [1 values with dtype=float32]

  • lon
    ()
    float32
    ...

    [1 values with dtype=float32]

• Data variables: (10)
  • allprecip_mask
    (time, range)
    bool
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    applied_thresholds :

    At least one pixel within the cloud with Z > -15 dBZ. Z < -50 dBZ pixels are removed.

    description :

    The mask is the rectangular box (cloud_start_time, cloud_end_time) x (cloud_base_height, cloud_top_height) that contains all the cloud; it can contain clear-air pixels.

    long_name :

    all precipitating cloud mask

    standard_name :

    cloud_binary_mask

    units :

    1

    Array Chunk Bytes 9.65 GiB 4.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type bool numpy.ndarray

  • drizzle_mask
    (time, range)
    bool
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    applied_thresholds :

    At least one pixel within the cloud with Z > -15 dBZ, no pixels above 0 dBZ at any point within the cloud. Z < -50 dBZ pixels are removed.

    description :

    The mask is the rectangular box (cloud_start_time, cloud_end_time) x (cloud_base_height, cloud_top_height) that contains all the cloud; it can contain clear-air pixels.

    long_name :

    drizzling cloud mask

    standard_name :

    cloud_binary_mask

    units :

    1

    Array Chunk Bytes 9.65 GiB 4.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type bool numpy.ndarray

  • id_mask
    (time, range)
    int64
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    applied_thresholds :

    Z < -50 dBZ pixels are removed.

    description :

    Each pixel is labelled with the unique identifier cloud_id of the cloud it belongs to.

    long_name :

    cloud id label mask

    standard_name :

    cloud_binary_mask

    units :

    1

    Array Chunk Bytes 77.22 GiB 32.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type int64 numpy.ndarray

  • longlived_rain_mask
    (time, range)
    bool
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    applied_thresholds :

    Cloud duration longer than 20 min. At least one pixel within the cloud with Z > -15 dBZ, cloud base height AGL below 250 m. Z < -50 dBZ pixels are removed.

    description :

    The mask is the rectangular box (cloud_start_time, cloud_end_time) x (cloud_base_height, cloud_top_height) that contains all the cloud; it can contain clear-air pixels.

    long_name :

    short raining cloud mask

    standard_name :

    cloud_binary_mask

    units :

    1

    Array Chunk Bytes 9.65 GiB 4.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type bool numpy.ndarray

  • nonprecip_mask
    (time, range)
    bool
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    applied_thresholds :

    Z < -15 dBZ at all points within the cloud. Z < -50 dBZ pixels are removed.

    description :

    The mask is the rectangular box (cloud_start_time, cloud_end_time) x (cloud_base_height, cloud_top_height) that contains all the cloud; it can contain clear-air pixels.

    long_name :

    non-precipitating cloud mask

    standard_name :

    cloud_binary_mask

    units :

    1

    Array Chunk Bytes 9.65 GiB 4.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type bool numpy.ndarray

  • rain_mask
    (time, range)
    bool
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    applied_thresholds :

    At least one pixel within the cloud with Z > -15 dBZ, cloud base height AGL below 250 m. Z < -50 dBZ pixels are removed.

    description :

    The mask is the rectangular box (cloud_start_time, cloud_end_time) x (cloud_base_height, cloud_top_height) that contains all the cloud; it can contain clear-air pixels.

    long_name :

    raining cloud mask

    standard_name :

    cloud_binary_mask

    units :

    1

    Array Chunk Bytes 9.65 GiB 4.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type bool numpy.ndarray

  • shortlived_rain_mask
    (time, range)
    bool
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    applied_thresholds :

    Cloud duration shorter than 20 min. At least one pixel within the cloud with Z > -15 dBZ, cloud base height AGL below 250 m. Z < -50 dBZ pixels are removed.

    description :

    The mask is the rectangular box (cloud_start_time, cloud_end_time) x (cloud_base_height, cloud_top_height) that contains all the cloud; it can contain clear-air pixels.

    long_name :

    short raining cloud mask

    standard_name :

    cloud_binary_mask

    units :

    1

    Array Chunk Bytes 9.65 GiB 4.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type bool numpy.ndarray

  • unclassified_mask
    (time, range)
    bool
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    applied_thresholds :

    Cloud top height AGL above 4500 m. Z < -50 dBZ pixels are removed.

    description :

    The mask is the rectangular box (cloud_start_time, cloud_end_time) x (cloud_base_height, cloud_top_height) that contains all the cloud; it can contain clear-air pixels.

    long_name :

    unclassified cloud mask

    standard_name :

    cloud_binary_mask

    units :

    1

    Array Chunk Bytes 9.65 GiB 4.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type bool numpy.ndarray

  • w
    (time, range)
    float32
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    description :

    Air velocity (precipitating and non-precipitating clouds, lidar composite). Positive is upwards.

    long_name :

    radar-lidar composite vertical air velocity

    standard_name :

    upward_air_velocity

    units :

    m s-1

    Array Chunk Bytes 38.61 GiB 16.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type float32 numpy.ndarray

  • w_radar
    (time, range)
    float32
    dask.array<chunksize=(131072, 32), meta=np.ndarray>

    description :

    Air velocity (precipitating and non-precipitating clouds, radar retrieval). Positive is upwards.

    long_name :

    retrieved radar vertical air velocity

    standard_name :

    upward_air_velocity

    units :

    m s-1

    Array Chunk Bytes 38.61 GiB 16.00 MiB Shape (55424828, 187) (131072, 32) Dask graph 2538 chunks in 2 graph layers Data type float32 numpy.ndarray

• Attributes: (18)

  Conventions :

  CF-1.12

  _logical_cutoff_date :

  2024-09-29T00:00:00Z

  author :

  Florian Poydenot (florian.poydenot@uni-hamburg.de)

  bcoproc_version :

  0.0.0.post2156.dev0+d4f3d45

  converted_by :

  Florian Poydenot (florian.poydenot@uni-hamburg.de)

  created_with :

  make_final_data_netcdf.py

  creation_date :

  Wed 25 Mar 2026 13:45:33

  institution :

  Max Planck Institute for Meteorology, Hamburg

  instrument :

  Radar: mbrs, Lidar: wbandlidar

  license :

  CC0-1.0

  location :

  The Barbados Cloud Observatory (BCO), Deebles Point, Barbados, West Indies

  platform :

  BCO

  sensor_alt :

  18.5m

  source :

  Ka-Band radar and Doppler wind lidar observations

  source_file :

  /work/um1487/OBS/SCORE/Level_1/B_Reflectivity/Ka-Band/MBRS/2s/202409/MMCR__MBRS__Spectral_Moments__2s__155m-18km__240924.nc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_1200.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_1300.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_1400.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_1500.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_1600.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_1700.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_1800.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_1900.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_2000.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_2100.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_2200.znc, /work/um1487/OBS/SCORE/Level_0/MBRS/mom_fixed/20240924/20240924_2300.znc, /work/um1487/m301163/lidar/lidar_interpolated/MBRS/WindLidar2/202409/WindLidar__Deebles_Point__WBAND__VerticalVelocity__STARE__1s__20240924_interpolated_mbrs.nc

  title :

  Composite lidar-radar retrieval of the cloud-scale vertical air motion at BCO

  tool_versions :

  {"Python": "3.11.2 (main, Apr 28 2025, 14:11:48) [GCC 12.2.0]", "attrs": "25.3.0", "pandas": "2.2.3", "bcoproc": "0.0.0.post2156.dev0+d4f3d45", "MarkupSafe": "3.0.3", "idna": "3.10", "pluggy": "1.5.0", "aiosignal": "1.3.2", "scipy": "1.17.1", "locket": "1.0.0", "pyyaml": "6.0.2", "bcologs": "0.0.0.post2156.dev0+d4f3d45", "certifi": "2025.4.26", "fsspec": "2025.3.2", "numpy": "2.2.5", "click": "8.1.8", "aiohttp": "3.11.18", "six": "1.17.0", "frozenlist": "1.6.0", "zipp": "3.21.0", "packaging": "25.0", "propcache": "0.3.1", "dask": "2025.4.1", "tzdata": "2025.2", "partd": "1.4.2", "cftime": "1.6.4.post1", "python-dateutil": "2.9.0.post0", "importlib_metadata": "8.7.0", "multidict": "6.4.3", "iniconfig": "2.1.0", "pytz": "2025.2", "datarecords": "0.0.0.post2156.dev0+d4f3d45", "yarl": "1.20.0", "flox": "0.11.2", "numpy_groupies": "0.11.3", "xarray": "2025.4.0", "netcdf4": "1.7.2", "jinja2": "3.1.6", "toolz": "1.0.0", "pendulum": "3.1.0", "cloudpickle": "3.1.1", "pytest": "8.3.5", "ruff": "0.13.0", "aiohappyeyeballs": "2.6.1"}

  version :

  git-v1.3-132-gfbcc69d

Sample Plot

rainy_orcestra = slice("2024-09-23T11:45", "2024-09-23T12:35")
ds.sel(time=rainy_orcestra).w.plot(x="time")

<matplotlib.collections.QuadMesh at 0x7f8d6d543380>

mbrs.sel(time=rainy_orcestra).Ze.plot(x="time")

<matplotlib.collections.QuadMesh at 0x7f8d6d4fe5d0>

References:

1. Zhu, Z., Kollias, P., Yang, F. & Luke, E. On the Estimation of In-Cloud Vertical Air Motion Using Radar Doppler Spectra. Geophysical Research Letters 48, e2020GL090682 (2021).

2. Poydenot, F., Robbins-Blanch, N., Zhu, Z. & Vogel, R. How rain shapes cloud-scale dynamics and mass flux in the trades. EGUsphere 1–25 (2026) doi:10.5194/egusphere-2026-1974.