Cloud radar is a specialized type of radar designed to detect the tiny water droplets and ice crystals that make up clouds, particles far too small for conventional weather radar to see. Operating at millimeter wavelengths, typically around 35 or 94 gigahertz, these instruments fill a critical gap in atmospheric observation by profiling the vertical structure of clouds from the ground, from aircraft, and from orbit. The technology has reshaped how scientists study cloud behavior, from thin cirrus veils in the upper atmosphere to drizzle forming inside low marine fog, and its measurements feed directly into the climate models that project future warming.
Why Ordinary Weather Radar Cannot See Clouds
Standard weather radars operate at longer wavelengths, in the centimeter range, and are built to track rain, hail, and snow. These wavelengths interact strongly with large precipitation-sized drops but pass right through the microscopic droplets suspended inside most clouds. A typical cloud droplet is roughly 10 to 20 micrometers across, thousands of times smaller than a raindrop, so it scatters almost no energy back to a centimeter-wave radar. The physics behind this is straightforward: the amount of energy a small particle scatters back is proportional to the sixth power of its diameter. Double the size of a droplet and it returns 64 times more signal. Cloud droplets are so small that they are essentially invisible at conventional radar frequencies.
Cloud radar solves this by using much shorter wavelengths, in the millimeter range. Shrinking the wavelength boosts the backscattered signal from tiny particles enough for the radar to detect them. The most common frequency bands in use are Ka-band (around 35 GHz, with a wavelength near 8 mm) and W-band (around 94 GHz, wavelength near 3 mm). Some research setups also use X-band (around 10 GHz) for studying shallow precipitating clouds, though X-band sits closer to conventional weather radar territory and is better suited to precipitation-sized particles.
The Attenuation Trade-Off
Moving to shorter wavelengths comes with a significant penalty: the radar signal gets absorbed and scattered much more aggressively as it travels through the atmosphere, especially through liquid water. A 1-kilometer-thick liquid cloud containing a moderate amount of water produces negligible signal loss at longer wavelengths like Ku-band, but at W-band the one-way attenuation can reach about 4 decibels per kilometer. Rain makes things worse. A modest rainfall rate of 1 mm per hour causes roughly 1.2 dB per kilometer of one-way loss at Ka-band and about 3.4 dB per kilometer at W-band during the melting layer, where snowflakes transition to rain. Even ice-phase clouds can cause measurable attenuation at W-band when large snowflakes are present, with losses on the order of 0.5 to 2 dB per kilometer depending on how reflective the ice particles are.1Atmospheric Measurement Techniques. Estimating total attenuation using Rayleigh targets at cloud top: applications in multilayer and mixed-phase clouds observed by ground-based multifrequency radars
This means cloud radar sees the finest cloud detail close to the instrument but struggles to peer through thick, wet layers. A ground-based W-band radar looking up through heavy rain or a deep convective storm will lose its signal partway up. Researchers work around this limitation in several ways: using Ka-band where attenuation is lower and accepting reduced sensitivity to the smallest particles, combining measurements from multiple frequency bands, or correcting for attenuation using algorithms that estimate how much signal has been lost based on the cloud properties themselves.
What Doppler Measurements Add
Most modern cloud radars are Doppler-capable, meaning they measure not just how much energy comes back from a cloud but also the velocity of the particles producing the echo. When the radar points straight up, it captures the vertical motion of cloud droplets and ice crystals. That vertical velocity is a combination of two things: the air itself moving up or down, and the particles falling through that air under gravity. Separating these two contributions is one of the trickier problems in cloud radar science, because what researchers often want is the air motion, which controls how clouds grow and decay.
One approach pairs a cloud radar with a radar wind profiler. In a study combining a 35 GHz cloud radar with a 482 MHz wind profiler, researchers merged the Doppler spectra from both instruments to isolate the vertical air motion from the particle fall speed. The wind profiler is sensitive to clear-air turbulence and refractive-index gradients (essentially, it sees the air itself), while the cloud radar sees the hydrometeors. By comparing the two, the method achieved an estimated accuracy better than 0.1 meters per second for both the vertical air velocity and the terminal fall velocity of the particles.2Atmospheric Measurement Techniques. Combining cloud radar and radar wind profiler for a value added estimate of vertical air motion and particle terminal velocity within clouds
Doppler data from cloud radar has become especially valuable now that it is available from space. The EarthCARE satellite, launched in 2024, carries a 94 GHz radar with Doppler capability. Early validation against ground-based radars shows that EarthCARE’s Doppler velocities have minimal bias, within a few centimeters per second, in ice clouds. Performance remains reliable even in low-level mixed-phase clouds, though the radar’s long pulse length smooths out some of the finer vertical detail.3Atmospheric Chemistry and Physics. Evaluation of the EarthCARE Cloud Profiling Radar (CPR) Doppler velocity measurements using surface-based observations First-light observations have already captured Doppler velocity fields inside deep convective towers, offering a window into the internal dynamics of storm systems from orbit.4Atmospheric Measurement Techniques. First insights into deep convection by the Doppler velocity measurements of the EarthCARE Cloud Profiling Radar
Identifying Particle Shape with Polarimetry
Clouds contain a wide variety of ice crystal habits: thin plates, long columns, fat graupel, fluffy aggregates. These shapes matter for how clouds produce precipitation, how they scatter and absorb radiation, and how they interact with aircraft. Polarimetric cloud radar can distinguish among these habits by transmitting and receiving radar pulses in different polarization states and comparing the returns.
At 95 GHz, one classification scheme uses polarimetric variables and air temperature to sort ice crystals into categories: columnar crystals, planar crystals, mixtures of planar crystals with aggregates or lightly rimed particles, dense graupel-like snow, and graupel larger than about 2 mm.5Journal of Atmospheric and Oceanic Technology. Cloud Ice Crystal Classification Using a 95-GHz Polarimetric Radar At Ka-band, polarimetric observations have been used to track how ice crystal shape changes with temperature. Measurements during one European field campaign found that cloud layers near −5°C contained elongated (prolate) particles, layers near −15°C held flat (oblate) particles, and layers near −8°C and −20°C contained roughly spherical shapes.6Atmospheric Measurement Techniques. Relationship between temperature and apparent shape of pristine ice crystals derived from polarimetric cloud radar observations during the ACCEPT campaign These temperature-shape relationships align with what laboratory ice-growth studies predict, giving researchers confidence that the radar retrievals are capturing real microphysics.
In the Arctic, X-band polarimetric radar has been applied to identify ice particle types in shallow mixed-phase clouds near Barrow, Alaska. A Bayesian classification algorithm using reflectivity, differential reflectivity, differential phase, and cross-correlation measurements was able to detect and characterize the variability of cloud microphysics, with results consistent with ground-based in situ observations.7Atmospheric Research. Ice particle type identification for shallow Arctic mixed-phase clouds using X-band polarimetric radar
Ground-Based Observatories
The most extensive long-term cloud radar network belongs to the Atmospheric Radiation Measurement (ARM) program, run by the U.S. Department of Energy. ARM began deploying profiling millimeter-wavelength cloud radars at its observatories in the early 1990s, initially pointing straight up to profile the column of atmosphere above each site. Starting around 2009, the program expanded considerably, adding scanning radars in multiple frequency bands and deploying additional sites around the world. The network now spans fixed and mobile installations totaling over 43 years of combined observations across 18 different sites, covering environments from the tropical western Pacific to the Arctic North Slope of Alaska.8Bulletin of the American Meteorological Society. The ARM Radar Network: At the Leading Edge of Cloud and Precipitation Observations
These ground-based systems generate continuous time-height profiles of cloud structure, capturing how clouds form, evolve, precipitate, and dissipate over hours and days. The long records they accumulate are irreplaceable for evaluating whether satellites and models are getting cloud behavior right. Comparisons between ARM ground radars and the CloudSat spaceborne radar found differences that in most cases exceeded the 1 to 2 dB uncertainty of the comparison technique itself, highlighting that calibration and consistency across platforms remain active challenges.9Bulletin of the American Meteorological Society. The ARM Radar Network: At the Leading Edge of Cloud and Precipitation Observations
Cloud Radar in Space
The first dedicated spaceborne cloud radar flew on CloudSat, launched in 2006 as part of NASA’s A-Train constellation of Earth-observing satellites. Its 94 GHz Cloud Profiling Radar provided the first global, vertically resolved view of cloud structure. Using those vertical profiles, classification algorithms sort each observation into cloud types: thin cirrus, deep convective towers, anvil clouds, boundary-layer cumulus, and others. The initial year of data produced global distributions of cloud types over land and ocean that were broadly consistent with earlier estimates from passive instruments but revealed differences that exposed the limitations of both approaches.10Geophysical Research Letters. Classifying clouds around the globe with the CloudSat radar: 1‐year of results
CloudSat’s radar could not measure Doppler velocity, so it told you where cloud particles were but not how fast they were moving. EarthCARE, a joint European Space Agency and Japan Aerospace Exploration Agency mission, addresses that gap. Its 94 GHz radar adds Doppler capability, giving scientists the first global dataset of how fast hydrometeors are falling inside clouds.11Atmospheric Chemistry and Physics. Evaluation of the EarthCARE Cloud Profiling Radar (CPR) Doppler velocity measurements using surface-based observations These sedimentation velocities are directly linked to how efficiently clouds produce precipitation and how long cloud particles persist before falling out, both factors that strongly influence Earth’s energy budget.
Airborne Cloud Radar
Between ground stations and satellites sit airborne platforms, which fly radar instruments on research aircraft to sample clouds at close range across different climate zones. The HIAPER Cloud Radar, mounted on a high-altitude research aircraft, has been used to classify clouds into categories based on whether they are convective or stratiform across three distinct regions: the subtropical easterlies off the California coast, the Southern Ocean, and the tropics near Central America.12Journal of Geophysical Research: Atmospheres. Cloud Properties Derived From Airborne Cloud Radar Observations Collected in Three Climatic Regions Airborne radars fill an observational niche: they can fly targeted patterns around specific cloud systems, providing spatial context that a ground-based profiler (which stares at a single column) cannot, while offering better sensitivity and resolution than an orbiting radar passing overhead at 7 kilometers per second.
Combining Cloud Radar with Other Instruments
Cloud radar alone has blind spots. It loses sensitivity to the thinnest ice clouds because the particles are too small to return a detectable signal, and it saturates in heavy precipitation because of attenuation. Lidar has the opposite profile: it excels at detecting thin cloud layers and aerosol but is quickly extinguished by optically thick clouds. Combining the two gives a far more complete picture than either alone.
The most productive example of this pairing has been CloudSat’s radar flying in close formation with the CALIPSO lidar. A combined radar-lidar geometrical profile product merges the two datasets to characterize the vertical and spatial structure of hydrometeor layers, filling in the layers where one instrument loses sensitivity.13Journal of Geophysical Research: Atmospheres. The CloudSat radar‐lidar geometrical profile product (RL‐GeoProf): Updates, improvements, and selected results Variational retrieval algorithms that fuse radar, lidar, and infrared radiometer data from the A-Train satellites produce seamless profiles of ice cloud properties, transitioning smoothly between regions where one sensor detects the cloud and regions where both do.14Journal of Geophysical Research: Atmospheres. Combined CloudSat‐CALIPSO‐MODIS retrievals of the properties of ice clouds
On the ground, similar synergies are routine. ARM sites typically co-locate cloud radar with lidar, microwave radiometers, and radiosonde launches. The combination allows retrieval of quantities that no single instrument can provide alone, such as liquid water content profiles in mixed-phase clouds where ice crystals dominate the radar signal but liquid droplets dominate the lidar return.
Discoveries in Marine and Polar Clouds
Cloud radar has been particularly productive in two environments that matter enormously for climate but are difficult to study with other tools: marine boundary-layer clouds and polar mixed-phase clouds.
Low-lying marine stratocumulus and stratiform clouds cover vast stretches of the subtropical oceans and have an outsized cooling effect because they reflect sunlight back to space. Whether these clouds drizzle affects how long they last and how much area they cover. A machine-learning algorithm applied to ARM cloud radar Doppler spectra found that drizzle is far more common in these clouds than previously estimated. Traditional approaches based on radar reflectivity alone significantly underestimated drizzle occurrence, especially in thin clouds with low liquid water paths.15Atmospheric Chemistry and Physics. New insights on the prevalence of drizzle in marine stratocumulus clouds based on a machine learning algorithm applied to radar Doppler spectra Since drizzle formation regulates cloud lifetime and coverage, underestimating it means misrepresenting one of the largest terms in Earth’s radiation budget.
In polar regions, mixed-phase clouds persist for surprisingly long periods even in conditions where theory suggests the ice phase should quickly consume all available liquid water. During the MOSAiC expedition, which drifted with Arctic sea ice from October 2019 to September 2020, a combination of lidar and radar tracked the evolution of mixed-phase clouds over the central Arctic for the first time with robust phase discrimination. Two long-lasting events, one in mid-winter near the North Pole and another in late summer, provided new insight into how liquid and ice phases coexist and interact over extended periods.16Atmospheric Chemistry and Physics. MOSAiC studies of long-lasting mixed-phase cloud events and analysis of the liquid-phase properties of Arctic clouds Earlier work at Barrow, Alaska, used the morphological features of cloud radar Doppler spectra to detect supercooled liquid droplets even when ice crystals dominated the radar signal, training the algorithm against lidar depolarization measurements that are sensitive to liquid water.17Journal of Geophysical Research: Atmospheres. Detection of supercooled liquid in mixed‐phase clouds using radar Doppler spectra These mixed-phase cloud processes are central to the Arctic amplification of climate warming, because the balance between liquid and ice in polar clouds controls how much longwave radiation reaches the surface during the dark winter months.
Feeding Climate Models
One of the highest-impact applications of cloud radar data is evaluating and improving the cloud representations inside numerical weather prediction and climate models. Clouds remain the single largest source of uncertainty in climate projections, largely because models cannot resolve individual cloud processes and must approximate them with simplified schemes. A significant part of that uncertainty comes from a shortage of observations detailed enough to challenge the assumptions baked into those schemes. Multi-frequency cloud radar observations, combined with polarimetric weather radar and other instruments at research supersites, now enable a much more thorough test of microphysical parameterizations than was possible a decade ago.18Copernicus Publications. Overview: Fusion of radar polarimetry and numerical atmospheric modelling towards an improved understanding of cloud and precipitation processes
The process works roughly like this: a model predicts how much liquid and ice should be present at each altitude, what particle sizes should exist, and how quickly precipitation should form. Cloud radar measures those same quantities in the real atmosphere. Where the two diverge, modelers know their parameterization needs work. This feedback loop has been especially productive for ice cloud microphysics, where models historically had very little observational constraint and could produce wildly different answers depending on which assumptions they used.
The Insect Problem
A surprisingly persistent headache for ground-based cloud radar operators is biological contamination. Insects, birds, and bats are strong scatterers at millimeter wavelengths, and they occupy the same altitudes as many cloud types. On warm days in summer, insect echoes can extend several kilometers above the surface and overlap with genuine cloud returns, corrupting retrievals of cloud boundaries and microphysical properties. A filtering algorithm developed for 35 GHz radar uses theoretical sensitivity curves and statistical variance in the echo to separate biological scatterers from cloud returns. When insect density is high, incorporating the radar’s linear depolarization ratio improves separation; in turbulent shallow convective clouds, combining depolarization ratio with spectral width performs best.19Atmospheric Measurement Techniques. A simple biota removal algorithm for 35 GHz cloud radar measurements The fact that researchers continue to develop and refine these removal techniques speaks to how stubborn the contamination problem remains, particularly in tropical and midlatitude continental environments where insect activity peaks during the warm season.
Dual-Wavelength and Multi-Wavelength Techniques
Using two or more radar frequencies simultaneously provides information that no single frequency can offer. The principle exploits the fact that large particles scatter differently at different wavelengths, while small particles scatter identically (relative to wavelength) at any frequency. By comparing the reflectivity measured at two frequencies, researchers can estimate the size of the dominant scatterers in a cloud volume.
A dual-wavelength approach has been tested for retrieving cloud liquid water and ice content by measuring how much extra attenuation the shorter wavelength suffers compared to the longer one. In principle, the difference isolates the liquid water because liquid absorbs millimeter waves much more efficiently than ice does. In practice, when non-Rayleigh-sized particles are present, such as drizzle drops or large ice crystals, the reflectivity difference includes contributions from both attenuation and particle size effects, creating an ambiguity that biases the liquid water estimate regardless of which frequency pair is used.20Journal of Atmospheric and Oceanic Technology. Cloud Liquid Water and Ice Content Retrieval by Multiwavelength Radar This is why modern retrieval algorithms increasingly combine radar with other measurements, like microwave radiometers that constrain the total liquid water path, rather than relying on the radar alone.
Triple-frequency setups, pairing something like S-band, Ka-band, and W-band, push the technique further by sampling three distinct scattering regimes simultaneously. The ARM program’s expansion to four frequency bands at some sites reflects this trend toward richer multi-wavelength datasets that can tease apart the contributions of different particle populations within the same cloud volume.21Bulletin of the American Meteorological Society. The ARM Radar Network: At the Leading Edge of Cloud and Precipitation Observations