Ceilometer Algorithm Reads the Sky’s Aerosol Story in Real Time

Every day, invisible plumes of Saharan dust, Canadian wildfire smoke, and urban pollution drift across Europe, shaping air quality, cloud formation, and the planet’s radiation balance. Identifying what is floating overhead, and at what altitude, has traditionally required expensive multi-wavelength research lidars or satellites that pass over only a few times a day. Now, researchers at the Royal Meteorological Institute of Belgium have shown that a workhorse instrument already installed at hundreds of weather stations worldwide can do far more than anyone expected. Their new algorithm, called CONIOPOL, transforms a standard commercial ceilometer into a real-time classifier of aerosols, clouds, and precipitation, opening a path to continuous, low-cost atmospheric monitoring across entire continents.

CONIOPOL, whose name blends ‘coniology’ (the study of atmospheric dust) with ‘polarization’, was developed by Quentin Laffineur, Alexander Mangold, and Andy W. Delcloo and is described in the journal Atmospheric Measurement Techniques. The algorithm runs on the Vaisala CL61, a single-wavelength automatic lidar ceilometer that fires a 910.55-nanometer infrared laser straight up and records the light scattered back by particles in the air. What sets the CL61 apart from ordinary ceilometers is its polarization capability: it measures both the parallel and cross-polarized components of the returning signal. Spherical particles such as liquid droplets preserve the laser’s polarization, while irregular particles like mineral dust, volcanic ash, and ice crystals scramble it. That difference, quantified as the volume linear depolarization ratio, becomes the key that unlocks particle identification.

The algorithm weaves together three strands of information: the attenuated backscatter profile, the depolarization ratio, and the cloud-base height determined by the instrument itself. After careful signal conditioning, including noise filtering based on the signal-to-noise ratio and a two-dimensional median filter that suppresses isolated noisy pixels without blurring real atmospheric layers, a decision tree sorts every measurement into categories. A threshold on backscatter, determined statistically from a full year of data at about 3 x 10^-6 per meter per steradian, separates clouds and precipitation from aerosols. Depolarization values then split the targets further: values above 0.2 point to mineral dust, values above 0.37 suggest volcanic ash, and values below 0.1 indicate spherical particles such as smoke, marine salt, or liquid droplets.

One of the algorithm’s most distinctive features is its honesty about ambiguity. Rather than forcing every observation into a single pure category, CONIOPOL includes transitional classes that capture mixtures, such as dust blended with boundary-layer pollution or pollen mixed with continental aerosols. A two-dimensional sliding-window consistency filter then examines neighborhoods of five consecutive profiles and five adjacent range gates, spanning roughly 150 seconds and 24 meters, and suppresses isolated detections that do not form coherent structures. This approach acknowledges that the atmosphere is a continuum, not a set of neat boxes, and it makes the resulting classification maps far more robust against noise and instrument artifacts.

Calibration was a central challenge. The CL61 has no internal depolarization calibrator, so the team adopted a clever workaround borrowed from previous work on Doppler lidars: monitoring the depolarization signal at the base of liquid water clouds, whose spherical droplets provide a natural reference with a known, near-zero depolarization. Over the full year of observations, from February 2024 to January 2025, the median depolarization at liquid cloud bases stayed stable between 0.01 and 0.02, confirming that the instrument’s polarization channel did not drift. The backscatter calibration, anchored to a factory-calibrated reference using the well-established liquid-cloud calibration method, also remained stable throughout the study period.

To test whether the classifications made physical sense, the researchers compared them against the Copernicus Atmosphere Monitoring Service (CAMS) global forecast, which models aerosol mixing ratios for dust, sea salt, soot, organic carbon, and sulfate. Because the two datasets differ fundamentally in resolution and nature, the comparison was conditional rather than point-by-point. The results were striking: whenever CONIOPOL detected dust, CAMS dust mixing ratios were significantly elevated below about 3 kilometers, and the intermediate ‘aerosol mix’ class showed intermediate dust levels, exactly as expected for diluted or partially mixed dust intrusions. Similarly, smoke and marine detections coincided with enhanced modeled concentrations of soot, organic carbon, sulfate, and sea salt, particularly during summer when wildfire plumes crossed Europe at higher altitudes.

The year-long record also revealed clear seasonal rhythms. Continental aerosols dominated, detected in more than half of all hours and peaking near 85 percent in July and August, when photochemical production of secondary particles intensifies. Smoke and marine aerosols appeared 10 to 27 percent of the time, mostly confined below one kilometer, consistent with sea-salt aerosols trapped in the marine boundary layer. Dust appeared mainly in spring and autumn, with winter dust layers reaching above 3 kilometers. Ice crystals showed a pronounced seasonal cycle with maxima in late winter and autumn, while liquid cloud droplets were by far the most common hydrometeor, occurring 20 to 70 percent of the time depending on the month.

A showcase example from February 2024 illustrates the algorithm’s power. During a single night, CONIOPOL tracked an elevated dust plume near the top of the boundary layer, distinguished an upper cloud layer of ice crystals from a lower layer of supercooled liquid water, and even captured a snow-to-rain transition near 2,200 meters altitude. Precipitation classification relies on depolarization thresholds derived from the data themselves: values below 0.01 indicate fog, drizzle, or mist; between 0.01 and 0.23, liquid rain; and above 0.41, heavily rimed graupel. Compared with a conventional rain gauge, the algorithm detected precipitation 5 to 13 percent more often, a positive bias partly attributed to the difficulty of separating drizzle and fog from true precipitation.

Sensitivity tests confirmed the algorithm’s resilience. Perturbing the backscatter calibration by up to 10 percent left most classifications unchanged, while depolarization variability proved to be the dominant source of uncertainty, mainly shifting observations between neighboring categories near fixed thresholds rather than producing arbitrary reassignments. The authors caution that the thresholds were developed for regional background conditions and long-range transported aerosols, and that fresh plumes near strong sources, such as erupting volcanoes or intense wildfires, may behave differently. They also note that a single site and a single year cannot fully validate the approach.

Even with those caveats, the implications are considerable. Polarization-enabled ceilometers are cheap, rugged, and already deployed in operational weather networks across Europe and beyond. CONIOPOL demonstrates that they can deliver scientifically meaningful aerosol typing around the clock, bridging the gaps between sparse satellite overpasses, campaign-based research lidars, and surface air-quality stations. The team envisions extending the algorithm to other CL61 sites and combining its output with the ACTRIS research infrastructure and the new EarthCARE satellite mission. If that happens, the humble ceilometer on the roof of a weather station may become one of the most valuable eyes humanity has on the ever-changing chemistry of the sky.

Subject of Research: Real-time aerosol and hydrometeor classification using single-wavelength polarization lidar

Article Title: New operational perspective to identify aerosols and hydrometeors in real-time with a pioneering algorithm (CONIOPOL) based on single-wavelength polarization lidar (CL61)

Article References: Laffineur, Q., Mangold, A., & Delcloo, A. W. (2026). New operational perspective to identify aerosols and hydrometeors in real-time with a pioneering algorithm (CONIOPOL) based on single-wavelength polarization lidar (CL61). Atmospheric Measurement Techniques, 19(19), 6415-6446. https://doi.org/10.5194/amt-19-6415-2026

Image Credits: AI Generated

DOI: 10.5194/amt-19-6415-2026

Keywords: aerosols, lidar, ceilometer, depolarization, CONIOPOL, cloud phase, precipitation, Saharan dust, wildfire smoke, air quality, CAMS, atmospheric monitoring