Cloud height can be measured from the ground using laser-based instruments called ceilometers, cloud radar, and simple thermodynamic calculations, while from above it is determined by satellite lidar, infrared sensors, stereoscopic imaging, and even neural networks trained on satellite data. The choice of method depends on whether you need the height of a cloud’s base or its top, how accurate you need the answer to be, and whether you are an airport weather observer, a climate scientist, or someone standing in a field with a thermometer.
Ceilometers and Cloud Radar From the Ground
The workhorse instrument for measuring cloud base height at airports and weather stations is the ceilometer, a device that fires a short pulse of laser light straight up and times how long it takes for the reflection to return from the cloud base. The principle is identical to how a bat echolocates: send a signal, wait for the echo, divide by two, and you have distance. Modern ceilometers operate continuously, generating a new cloud base reading every few seconds, and they are accurate enough that aviation weather reports worldwide rely on them. A long-term study of ceilometer data found that cloud base height varies dramatically by season: during summer months, for instance, more than a quarter of detected clouds had bases near 1,400 meters, and roughly 80% had bases below 3,000 meters.1Atmospheric Research. Behavior of cloud base height from ceilometer measurements
Cloud radar works on the same time-of-flight principle but uses microwave pulses instead of light, which lets it see deeper into the cloud and measure not just the base but also the internal structure and the top. Ka-band millimeter-wave cloud radars, operating at around 35 GHz, can profile clouds from the ground all the way up to about 15 kilometers. A study using this type of radar over the Tianshan mountains in central Asia verified cloud top heights by comparing radar readings against observations from the Fengyun-4A weather satellite, and the two generally agreed well.2Meteorological Applications. Fine vertical structures of cloud from a ground‐based cloud radar over the western Tianshan mountains Radar also captures something ceilometers miss: the fine internal layering within a cloud, including thin embedded layers that a laser pulse might not detect.
Detecting weak cloud signals from radar data is trickier than it sounds, because the returned signals can be faint enough to blend into instrument noise. One approach uses three kinds of continuity to separate real clouds from random noise: continuity across altitude, continuity across the Doppler velocity spectrum, and continuity over time. The idea is that real clouds persist in a coherent way across all three dimensions, while noise does not.3PubMed Central. A Cloud Detection Method for Vertically Pointing Millimeter-Wavelength Cloud Radar
The Temperature-Dewpoint Method
You do not need a laser or a radar dish to get a rough estimate of cloud base height. The simplest field method uses the air temperature and the dewpoint, which is the temperature at which air becomes saturated and water begins to condense. As a parcel of air rises, it cools at a predictable rate, and the dewpoint drops at a slower rate. The altitude where the two converge is approximately where clouds form. The rule of thumb used by pilots and meteorologists is to take the difference between the surface temperature and the dewpoint in degrees Celsius, multiply by about 125, and you get the cloud base height in meters. On a warm day with dry air, that gap might be large, pushing the cloud base several kilometers up. On a humid day, the gap shrinks and clouds can sit close to the ground.
This calculation, sometimes called the Espy method or the lifted condensation level estimate, works best for fair-weather cumulus clouds that form from surface heating. It is less reliable for layered clouds pushed in by weather fronts, fog that forms by radiative cooling at night, or clouds that develop aloft from entirely different air masses. Research into thunderstorm prediction illustrates the practical significance of cloud base height: on days when thunderstorms developed, cloud bases averaged around 890 meters, while on non-thunderstorm days, cloud bases averaged roughly 2,000 meters.4Journal of Atmospheric and Solar-Terrestrial Physics. A new thermodynamic index for thunderstorm detection based on cloud base height and equivalent potential temperature Low cloud bases, combined with high atmospheric instability, are a strong indicator that storms are likely to fire.
Satellite Lidar for True Cloud Top Height
From space, the most trusted measurement of cloud top height comes from profiling lidars, which are essentially space-based ceilometers pointing downward. The CALIPSO satellite, operating from 2006 to 2023, carried a lidar that sent pulses toward Earth and measured the backscatter profile of the atmosphere, detecting both cloud layers and aerosol layers with sharp vertical resolution. Space profiling lidars are widely considered the most reliable source for total cloud amount and true geometric cloud top height.5Atmospheric Measurement Techniques. Impact of the revisit frequency on cloud climatology for CALIPSO, EarthCARE, Aeolus, and ICESat-2 satellite lidar missions
CALIPSO’s successor is the ATLID lidar aboard the EarthCARE satellite, launched in 2024. ATLID is a high-spectral-resolution lidar, meaning it can separate the different types of scattering that occur when laser light hits molecules versus particles. It identifies cloud boundaries using a technique called the wavelet covariance transform, which detects sharp changes in the backscatter signal. Strong features are classified at roughly one-kilometer horizontal resolution, while thinner or more diffuse cloud layers are detected by averaging over a wider swath.6Atmospheric Measurement Techniques. Cloud top heights and aerosol layer properties from EarthCARE lidar observations: the A-CTH and A-ALD products The trade-off with any satellite lidar is coverage: a single satellite in a polar orbit only passes over a given location once every several days, so the global picture is built up over time rather than captured in a snapshot.
Infrared and Geometric Methods From Satellites
Satellite lidars are accurate but expensive and limited in coverage. Most operational cloud height estimates from space rely on passive sensors that observe the thermal infrared radiation Earth emits. The basic idea is that cloud tops are colder than the surface below, and if you know how temperature decreases with altitude in the atmosphere, you can work backward from the cloud’s brightness temperature to its approximate height. This is straightforward for thick, opaque clouds that completely block the radiation from below, but it breaks down for thin cirrus clouds that are semi-transparent.
A widely used refinement is the CO₂ slicing technique, which exploits the fact that carbon dioxide absorbs infrared radiation at slightly different altitudes depending on the wavelength. By comparing measurements in several CO₂ absorption bands, you can triangulate the altitude where the cloud is emitting most of its signal. CO₂ slicing has been generally accepted as a useful method for determining cloud top pressure for clouds above about 600 hPa, which corresponds to roughly the upper half of the troposphere.7Journal of Geophysical Research: Atmospheres. Improvement in thin cirrus retrievals using an emissivity‐adjusted CO2 slicing algorithm An adaptation of this technique applied to volcanic ash clouds from the 2010 Eyjafjallajökull eruption returned height estimates with errors of about 2.2 kilometers compared to lidar validation, making it useful as a quick first approximation for aviation hazard alerts.8Atmospheric Measurement Techniques. An adaptation of the CO2 slicing technique for the Infrared Atmospheric Sounding Interferometer to obtain the height of tropospheric volcanic ash clouds
A completely different satellite approach avoids temperature altogether and uses geometry. If an instrument views the same cloud from two angles, either because the satellite has a dual-angle scanner or because two separate instruments observe the same scene, the apparent shift of the cloud between views can be converted into a height measurement using parallax, the same principle your brain uses for depth perception. The Along Track Scanning Radiometer (ATSR) aboard the European ERS-1 satellite demonstrated this by comparing nadir and forward-looking images of the same clouds, achieving height accuracy of roughly plus or minus one kilometer.9Remote Sensing of Environment. Cloud-top height determination using ATSR data The Multi-angle Imaging SpectroRadiometer (MISR) on NASA’s Terra satellite later refined stereoscopic cloud height retrieval on a global scale. A recent study examined the errors these stereoscopic methods introduce and found that they depend on wavelength, viewing geometry, spatial resolution, and how bumpy the cloud top is, with optically thin clouds and oblique viewing angles introducing the largest biases.10Atmospheric Measurement Techniques. Errors in stereoscopic retrievals of cloud top height for single-layer clouds
Yet another passive approach uses the oxygen A-band, a set of absorption lines near 760 nanometers in the visible-to-near-infrared range. Oxygen is well mixed throughout the atmosphere, so the depth of these absorption features in sunlight reflected off a cloud encodes how much atmosphere the light passed through on its way back to the sensor, which in turn reveals the cloud top height. Oxygen A-band measurements provide information about cloud top height, cloud physical and optical thickness, and even surface atmospheric pressure.11Frontiers in Remote Sensing. Oxygen A-band absorption spectroscopy with solar photon counting and lithium niobate nanophotonic circuits This method works during daytime only, since it depends on reflected sunlight, but it neatly complements thermal infrared techniques that work day and night.
Balloons and Aircraft
Sometimes you just fly through the cloud and measure where it begins and ends. Radiosondes, the instrument packages carried aloft by weather balloons twice a day from hundreds of stations worldwide, record temperature and humidity as they ascend. A sharp jump in relative humidity to near saturation marks the cloud base; a drop back to dry conditions marks the top. This gives a direct profile of where clouds sit, but it is a one-shot measurement over one location rather than a continuous picture.
For research purposes, balloons can carry more specialized instruments. A campaign over Switzerland used radiosondes paired with a frost point hygrometer and an optical backscatter detector to track cirrus clouds. Two balloons launched about an hour apart, one from Payerne and one from near Zurich about 120 kilometers downwind, detected the same thick cirrus cloud, confirming that the cloud had drifted intact between the two sites.12Atmospheric Chemistry and Physics. Balloon-borne match measurements of midlatitude cirrus clouds This kind of “match” technique lets researchers study how a cloud’s properties evolve as it moves.
Remotely piloted aircraft offer another in-situ option, especially for lower clouds. A study using a small research drone measured vertical wind speeds inside and near marine stratocumulus clouds and compared the results to a ground-based Doppler cloud radar, which provided measurements at roughly 29-meter vertical resolution up to 15 kilometers.13Atmospheric Measurement Techniques. Vertical wind velocity measurements using a five-hole probe with remotely piloted aircraft to study aerosol–cloud interactions The drone’s wind measurements agreed well with the radar’s, validating both approaches and showing that small unmanned aircraft can serve as flying cloud probes in environments where sending a crewed aircraft would be impractical or unnecessary.
Automated Reporting for Aviation
For pilots, knowing cloud base height and how much of the sky is covered is a safety-critical piece of information. Airports encode this data in METAR reports, the standardized weather observations transmitted every half hour or hour. In many countries, the cloud information in a METAR comes from a human observer reading a ceilometer display, but there is growing interest in fully automated reporting. An open-source algorithm called ampycloud was developed specifically for this purpose. It ingests raw ceilometer data, clusters the hits into distinct cloud layers, and assigns a sky coverage fraction to each layer.
Testing ampycloud against human-generated METARs showed that it agreed with the official report in about 58% of cases, and was off by at most one sky-coverage category in roughly 96% of cases. Its identification of a ceiling, meaning the lowest layer that covers more than half the sky and determines whether an airport is operating under visual or instrument flight rules, matched the METAR about 88% of the time.14Atmospheric Measurement Techniques. ampycloud: an open-source algorithm to determine cloud base heights and sky coverage fractions from ceilometer data The remaining disagreements tended to involve ambiguous situations where a human observer and the algorithm interpreted the same data differently, such as when scattered clouds were borderline between categories. Making this code open-source means airports and national weather services can adapt and improve it for their own conditions.
Machine Learning and Three-Dimensional Cloud Mapping
All the methods described so far produce either a single-point profile (ceilometer, radar, balloon) or a two-dimensional map of cloud top heights (satellite). The atmosphere, of course, is three-dimensional, and clouds have complex vertical structure: multiple layers, gaps, thin veils above thick decks. Reconstructing the full 3D cloud field from satellite imagery alone has long been a challenge, but convolutional neural networks are starting to crack it.
A recent approach trained a compact neural network on paired data: geostationary satellite images from Himawari-8 on one side, and joint CALIPSO/CloudSat vertical profiles on the other. The network learned to predict a per-pixel 38-layer cloud mask at 500-meter vertical resolution from the satellite imagery alone. Performance was strong, with a cloud top height bias of about 450 meters and a mean error in cloud thickness under half a kilometer.15Geophysical Research Letters. CNN‐Based Retrieval of 3D Cloud Structures Solely From Geostationary Satellite Imagery Because geostationary satellites observe the same region every 10 minutes rather than every few days, this method could eventually provide near-real-time 3D cloud structure over large areas, something no single instrument can deliver on its own.
Citizen Science and Visual Estimation
You do not need a satellite or a ceilometer to contribute useful cloud height observations. NASA’s GLOBE Observer app lets anyone with a smartphone photograph the sky and report cloud type and estimated coverage. Matching these ground observations against satellite overpasses helps validate satellite cloud products. The program has shown that engaged participants can collect large volumes of data quickly: during one month-long challenge, citizen scientists submitted over 55,000 cloud observations.16Humanities and Social Sciences Communications. The role of citizen science mobile apps in facilitating a contemporary digital agora
Visual estimation of cloud height without instruments is inherently rough, but knowing cloud type helps. Low clouds like stratus and stratocumulus typically sit below about 2,000 meters. Medium-altitude clouds prefixed with “alto” generally range from about 2,000 to 6,000 meters. High clouds, including cirrus and cirrostratus, live above 6,000 meters and can reach well beyond 10,000 meters in the tropics. If you can identify the cloud genus, you can bracket its altitude within a few thousand meters before doing any calculation. Combining that visual identification with the temperature-dewpoint method described earlier gets you surprisingly close to what a ceilometer would report for fair-weather cumulus.
Noctilucent Clouds at the Edge of Space
At the far extreme of cloud altitude sit noctilucent clouds, thin, electric-blue wisps that form near 82 kilometers altitude in the mesosphere, far above where weather happens. These are made of ice crystals that nucleate on meteor dust, and they are visible only during summer twilight at high latitudes, when the sun illuminates them from below the horizon while the lower atmosphere is already dark. Measuring their height requires specialized lidar systems that can reach the mesosphere.
A 22-year observational record from the ALOMAR observatory in northern Norway tracked long-term changes in noctilucent cloud properties. The data showed that brighter noctilucent clouds became more frequent over time, increasing in occurrence by roughly 9% per decade, while the altitude of the faintest detectable clouds dropped by about 108 meters per decade. Stronger clouds, by contrast, rose by about 76 meters per decade.17Journal of Atmospheric and Solar-Terrestrial Physics. Long-term variations of noctilucent clouds at ALOMAR These shifts may be connected to changes in upper-atmospheric temperature and water vapor driven by rising greenhouse gas concentrations far below, making noctilucent clouds an unexpectedly sensitive indicator of climate change in a part of the atmosphere most people never think about.
Why Different Methods Give Different Answers
If you point a ceilometer, a satellite infrared sensor, and a stereoscopic camera at the same cloud at the same time, you will not necessarily get the same number. Each instrument measures a slightly different thing. The ceilometer detects the first altitude where the laser hits enough droplets to scatter light back, which corresponds to the physical cloud base. An infrared sensor on a satellite infers the altitude of the radiating surface, which for a thin cirrus layer might be somewhere in the middle of the cloud rather than at its true top. A stereoscopic method measures the altitude of the brightest visible surface, which for a bumpy cumulus cloud is the top of the highest turret rather than the average cloud top.
These differences are not errors in the colloquial sense; they reflect fundamentally different definitions of “cloud height.” The aviation community cares almost exclusively about cloud base, because that determines ceiling and visibility for landing aircraft. Climate scientists often need cloud top height, because the radiative effect of a cloud on Earth’s energy budget depends heavily on how high and how cold its top is. Thunderstorm forecasters care about both. When reading any cloud height measurement, understanding which definition is being used matters more than the number of decimal places reported.
Multi-layer cloud scenes add another wrinkle. A ceilometer sees only the lowest cloud layer, because the laser cannot penetrate it. A satellite looking down sees only the top layer for the same reason. Radar can sometimes detect multiple layers, but signal attenuation through thick clouds limits how many layers it resolves. The most complete picture comes from combining instruments: radar for vertical structure, lidar for precise boundaries, satellite for spatial coverage. The 3D neural network approach described earlier is, in a sense, an attempt to fuse the strengths of different instruments into a single product without actually flying them all at once.