Cloud, aerosol and precipitation classification from ceilometer and lidar backscatter and model temperature. Built on ceilopyter for reading and harmonizing the instrument data.
Not yet on PyPI, so install from source:
git clone https://github.com/actris-cloudnet/ceiloclass.git
cd ceiloclass
python3 -m venv venv
source venv/bin/activate
pip install .Fetch and classify a day of raw ceilometer data for a site (the instrument is discovered automatically; if a site has several, you are prompted to pick one):
ceiloclass -s munich -d 2025-05-25 --showAdd --harmonized to use the Cloudnet harmonized lidar product instead of raw
data, -i to narrow to a particular instrument, or pass local files directly:
ceiloclass --harmonized -s munich -d 2025-05-25 --show
ceiloclass -i cl61 ceilo.nc -m model.nc --plot out.png--harmonized covers every harmonized backscatter instrument at the site —
ceilometers, PollyXT, and the experimental doppler-lidars — prompting when
several are available. Use -i to pick one directly, e.g. a HALO doppler lidar:
ceiloclass --harmonized -i halo -s bucharest -d 2026-05-19 --showClassify the CL61 lidar product at Kenttärova, averaging into 30 s bins, using the HARMONIE model and showing the plot up to 6 km:
ceiloclass -s kenttarova -d 2023-09-04 -a 30 --harmonized -i cl61 -m harmonie-fmi-6-11 --max-y 6 --show --no-histogram
| Argument | Description |
|---|---|
files |
Ceilometer data file(s). If omitted, files are fetched using --site/--date. |
-i, --instrument |
Instrument. For local raw files, the reader: cl31, cl51, cl61, chm15k, cs135, ct25k, ld40. When fetching a harmonized product it is optional and filters by instrument-id substring (e.g. halo, pollyxt, cl61); you are prompted if several remain. |
--harmonized |
Use a Cloudnet harmonized backscatter product (ceilometers, PollyXT, doppler-lidars) instead of raw data. |
-m, --model |
Cloudnet model netCDF file, or a model id to fetch (e.g. ecmwf, harmonie-fmi-6-11) when using --site/--date. |
-s, --site |
Cloudnet site id (to fetch raw files and/or model), e.g. munich. |
-d, --date |
Date YYYY-MM-DD (to fetch raw files and/or model). |
--download-dir |
Directory for fetched files (default: the package data/ directory). |
--calibration-factor |
Override the backscatter calibration factor. |
-a, --average |
Average into time bins of this width (seconds) before classifying (faster). |
--plot |
Write a classification plot to this PNG file. |
--show |
Show the plot in a window. |
--max-y |
Upper limit of the range axis in plots (km). |
--no-histogram |
Omit the diagnostic backscatter histogram panel from plots. |
Each time–range pixel is classified from attenuated backscatter and model temperature alone. The method follows CloudnetPy's target classification, restricted to the parts that work without radar:
- Liquid layers are detected as sharp backscatter peaks. A layer is supercooled liquid where the model wet-bulb temperature is below 0 °C, and a warm liquid droplet layer otherwise.
- Strong, non-liquid signal is cloud or precipitation: ice below the 0 °C level, drizzle/rain above it. The backscatter threshold separating this from weaker signal is picked per file from the backscatter histogram, so it adapts to each site/day's aerosol load.
- Remaining signal is aerosol; gates with no signal are clear.
- With a CL61, the depolarization ratio splits ice from liquid (non-spherical ice depolarizes strongly) and anchors the freezing level to the observed ice.
MIT