Cloud

A cloud is a visible aggregate of minute liquid water droplets, ice crystals, or both, suspended in the atmosphere above the Earth's surface. It forms when moist air cools to saturation and water vapor condenses onto tiny airborne particles. This entry concerns the meteorological sense of the word; the unrelated computing usage is cloud computing.

Clouds matter because they control much of the planet's weather and precipitation, move water through the water cycle, and both reflect incoming sunlight and trap outgoing heat. Because of these competing effects, clouds are among the largest sources of uncertainty in climate models and projections of climate change. Their appearance and height have been systematically classified since 1802, and that classification remains the basis of modern observation.

The article covers the thermodynamics and microphysics of cloud formation, the lifting mechanisms that produce clouds, the classification system and its history, the main cloud families and rarer forms, thunderstorms and precipitation, radiative effects, observation methods, representation in climate models, human influences, clouds on other planets, and the place of clouds in art and culture.

Definition and basic characteristics

A cloud is a visible assemblage of small water droplets, ice crystals, or both, suspended in the atmosphere. The definition separates clouds from related phenomena. Fog is best understood as a cloud whose base touches the ground. Haze and smoke are also visible suspensions of particles, but they do not contain a condensed-water phase, so they are not clouds in the meteorological sense. Water vapour itself is an invisible gas; a cloud becomes visible only once water has condensed into particles large enough to scatter light.

The droplets that make up a cloud are very small. Typical cloud droplets measure a few to several tens of micrometres across. In one cloud the mean diameter is often around 10 to 15 micrometres, but individual droplets can range from about 1 to 100 micrometres. Droplets of this size are far smaller than raindrops, and a cloud may contain roughly 200 droplets per cubic centimetre. Despite their numbers, the droplets occupy a tiny fraction of the air volume. In most clouds, liquid water makes up less than a millionth of the volume, and most of the water in the surrounding air remains as vapour.

The amount of liquid present is usually expressed as liquid water content, the mass of condensed water per unit volume of air, measured in grams per cubic metre. In stratiform clouds it is typically around 0.1 g/m³ to 0.2 g/m³. Fair-weather cumulus commonly reaches about 0.5 g/m³, large cumulus clouds can carry a few grams per cubic metre, and severe thunderstorms can exceed 3 g/m³. Even the densest clouds are therefore mostly empty space in terms of liquid, and the values vary considerably with cloud type and origin. This is one reason cloud type is closely linked to the precipitation it can produce, a relationship that weather forecasters rely on.

Clouds look white in sunlight because their droplets are much larger than the wavelengths of visible light and scatter all visible wavelengths in roughly equal measure, with little selective absorption. The light is scattered many times within a few metres of entering the cloud, so the light that finally emerges has been redirected repeatedly, which makes the cloud appear bright and uniform. Scattering of this kind is what gives sunlit cloud tops their brilliance. Thick clouds appear grey or dark at their bases because very little sunlight penetrates through them to be scattered back toward an observer below. The darkening is a matter of how much light reaches the base, not of the droplets becoming a different colour.

Cloud lifetimes vary enormously. A small fair-weather cumulus, sustained by a rising thermal, may form and disappear within minutes to about an hour. Extensive layers such as stratus or stratocumulus decks can persist for days or weeks, sustained by the moisture supply and the large-scale circulation of the air mass. Whether a cloud lasts depends on the balance between processes that create liquid, such as cooling of rising air, and processes that remove it, such as evaporation as air sinks or precipitation falling out. Those processes are controlled largely by the air mass and the surrounding circulation, which is why the same type of cloud can be fleeting in one setting and long-lived in another.

Thermodynamics of cloud formation

Whether a cloud forms depends first on how close the air is to saturation. The amount of water vapor in air is commonly expressed as relative humidity, the ratio of the actual vapor pressure to the saturation vapor pressure at the same temperature, multiplied by 100 per cent. The dew point is the temperature to which air must be cooled at constant pressure to become saturated, so relative humidity reaches 100 per cent when air temperature falls to the dew point. The difference between air temperature and dew point, called the dew-point depression, is therefore a direct measure of how far the air is from saturation. Saturation vapor pressure rises steeply with temperature, following the Clausius–Clapeyron relation:

Here $e_s$ is the saturation vapor pressure, $L$ the latent heat of vaporization, $R_v$ the gas constant for water vapor, and $T$ the absolute temperature. Near the surface this implies an increase of roughly 6 to 7 per cent in saturation vapor pressure for each kelvin of warming. Warm air can therefore hold far more vapor than cold air, and a modest drop in temperature can bring an air parcel to saturation.

Air that rises ascends into lower pressure and expands. Because this expansion does work on the surrounding air and little heat is exchanged with the environment, the process is close to adiabatic, and the parcel cools. An unsaturated parcel cools at the dry adiabatic lapse rate, $\Gamma_d = g/c_p \approx 9.8\ \mathrm{K/km}$. Its mixing ratio stays constant while its temperature falls, so its relative humidity rises as it climbs. The height at which it reaches saturation is the lifting condensation level (LCL). A common approximation is $z_{\mathrm{LCL}} \approx 125\ \mathrm{m} \times (T - T_d)$, where $T_d$ is the dew point. For example, air at 20 °C with a dew point of 10 °C reaches saturation at roughly 1250 m. The LCL marks the flat base of many cumulus clouds, where condensation begins and visible droplets first form. Condensation nuclei, discussed in a separate section, determine how readily this happens.

Once a parcel is saturated, further ascent involves condensation, which releases latent heat into the parcel. This warming offsets part of the expansional cooling, so the saturated parcel cools more slowly. The moist adiabatic lapse rate $\Gamma_m$ is therefore smaller than $\Gamma_d$. It varies with temperature and pressure, from about 4 K/km in very warm, humid air to near 9 K/km in cold air, and it is roughly 5 to 6 K/km in typical warm lower-tropospheric conditions. The latent heat of vaporization is about $2.5 \times 10^6\ \mathrm{J/kg}$ at 0 °C, so condensing even a few grams of vapor per kilogram of air releases a significant amount of energy. This heat source explains why convective clouds can keep growing well above the LCL, and why cumulonimbus clouds can reach the tropopause.

Whether a rising parcel continues to accelerate depends on atmospheric stability, which compares the observed environmental lapse rate $\Gamma$ with $\Gamma_d$ and $\Gamma_m$. If $\Gamma > \Gamma_d$, the atmosphere is absolutely unstable, and a displaced dry parcel becomes warmer and less dense than its surroundings and keeps rising. If $\Gamma_m < \Gamma < \Gamma_d$, the atmosphere is conditionally unstable: an unsaturated parcel is pulled back toward its starting level, but a parcel lifted far enough to saturate can become buoyant. If $\Gamma < \Gamma_m$, the atmosphere is absolutely stable, and a rising parcel becomes colder and denser than its surroundings and sinks back. Temperature inversions, where temperature increases with height, are an extreme case of stability that can cap vertical growth and favor flat, layered clouds. Whether an air mass is unstable or stable therefore determines whether a given lifting mechanism produces a shallow cumulus or a towering convective cloud.

Condensation nuclei and droplet microphysics

Water vapor does not easily turn into liquid on its own. In a clean container, a droplet needs a supersaturation of several hundred percent before it can form spontaneously from pure water vapor, far beyond anything found in the atmosphere. Real clouds avoid this barrier because the air always contains tiny solid or liquid particles. Those that take up water and become droplets are called cloud condensation nuclei (CCN). Common examples include sea salt from breaking waves, sulfate and other compounds made from sulfur gases, soot from combustion, and mineral dust lifted from deserts. Their number varies enormously: 19-3 Over the Arctic, air can hold only a few tens of CCN per cubic centimetre, while air over industrial cities can contain far more.

Activation and the Köhler curve

Whether a given particle becomes a droplet is described by Köhler theory, which Hilding Köhler published in 1936 (the work builds on earlier thermodynamic ideas). Two effects compete. The first is curvature: a very small droplet has a strongly curved surface, which makes it easier for molecules to escape, so its equilibrium vapour pressure is higher than that over a flat surface. This Kelvin effect pushes toward evaporation. The second is solute: dissolved salts and other solutes lower the vapour pressure of the water in the droplet, which pushes toward growth. In a simplified form, the equilibrium supersaturation over a droplet of radius $r$ can be written as $$S \approx \frac{A}{r} - \frac{B}{r^{3}}$$ where $A$ depends on surface tension and temperature, and $B$ depends on the amount and type of dissolved material. The curve passes through a maximum, the critical supersaturation. A particle whose size and composition give a critical supersaturation below the ambient value grows without limit once it passes that peak, and is said to be activated. Soluble particles of suitable size usually activate at supersaturations of only a fraction of a percent to a few percent, which is why ordinary cloud formation is possible at all. Surface-active organic material can lower the surface tension of a droplet, which in some cases reduces the critical supersaturation further.

Droplet number and cloud properties

For a given amount of liquid water, the number of droplets determines their size. If the same mass of water is spread over more droplets, each droplet is smaller. This relationship is central to aerosol–cloud interactions. Satellite and in situ studies commonly describe a cloud by its droplet number concentration and the droplet effective radius, a weighted mean size defined in 1974 by James Hansen and Larry Travis as the ratio of the third to the second moment of the droplet size distribution. Typical effective radii are around 14 μm globally for liquid clouds, and one commonly cited value for unpolluted boundary-layer clouds is about 16 μm. Adding aerosols to such a cloud lowers this radius. Smaller droplets scatter sunlight more efficiently, so the cloud reflects more light and appears brighter. This is the basis of the so-called Twomey effect, and it is one of the reasons aerosols influence the radiative forcing of clouds. Studies also find that effective droplet radii are roughly 15–20% larger over the ocean than over land, reflecting the lower CCN concentrations over the sea.

Droplet growth and the onset of rain

Once activated, droplets grow mainly by diffusion of water vapour onto their surfaces. Because the vapour supply is shared among all the droplets in a parcel and the growth rate slows as droplets get larger, diffusion alone is too slow to produce raindrops within the lifetime of most clouds. Typical cloud droplets therefore remain at a few to a few tens of micrometres in diameter. The resulting size distribution controls the optical properties of the cloud and how efficiently it can produce precipitation. Drops larger than about 20–25 μm in radius begin to collide and coalesce with their neighbours at appreciable rates, and the conversion of cloud droplets into drizzle-sized drops through this process is called autoconversion. Autoconversion is strongly sensitive to droplet number. Measurements and models indicate that when droplet numbers rise, the conversion rate falls steeply, so polluted clouds tend to hold their water longer before it rains out. This suppression of early precipitation, together with the brightening described above, forms the basis of the second indirect aerosol effect, which is an active subject in climate-change research. Uncertainties in these relationships remain large, particularly for the way turbulence and mixing alter droplet collisions.

Ice formation and phase processes in clouds

Cloud water does not simply freeze at 0 °C. Between about 0 °C and −40 °C, many clouds contain liquid droplets and ice crystals at the same time, and droplets can stay liquid well below the freezing point. Whether they do so depends on the particles available to act as seeds. Freezing needs a surface on which an ice embryo can form, and the particles that can do this, ice-nucleating particles (INPs), are far rarer than the particles that let droplets condense. Because of this scarcity, a cloud can remain almost entirely liquid even at temperatures where ice would be thermodynamically favoured. A very small, pure droplet can freeze spontaneously only at around −35 °C to −40 °C, a limit known as homogeneous freezing. Droplets that are larger or contain impurities usually freeze at warmer temperatures.

Most ice in natural clouds forms heterogeneously, with an external particle acting as the nucleus. Candidate particles include mineral dust such as feldspar, quartz and clay minerals, as well as certain biological materials, including some bacteria, fungal spores, pollen fragments and organic matter from soils and seawater. Nucleation can occur through several modes: deposition directly from vapour onto a particle, immersion of a particle inside a droplet, condensation followed by freezing, and contact with a particle. The efficiency of these particles is strongly temperature dependent. Near −10 °C only a small fraction of INPs are active, and the active fraction rises steeply as the air gets colder. This is why the share of ice in a cloud typically increases quickly with decreasing temperature.

Once ice crystals exist, the Bergeron–Findeisen process often takes over. The saturation vapour pressure over ice is lower than over supercooled water at the same temperature, by roughly 10 % near −10 °C and about 25 % near −30 °C. Air that is saturated with respect to liquid water is therefore strongly supersaturated with respect to ice. Water vapour diffuses from the droplets to the crystals, the droplets evaporate, and the crystals grow at their expense. The process is named after Tor Bergeron and Walter Findeisen, who described it in the 1930s. Growing crystals become heavy enough to fall. They may aggregate with one another, or collect supercooled droplets by riming and form graupel. Once they fall into warmer air below the freezing level, they melt into rain or survive as snow or sleet. In many mid-latitude precipitation systems, this ice-phase pathway is the main source of precipitation.

Clouds that are entirely above freezing (warm clouds) follow a different route. Droplets grow first by condensation, but diffusion alone is too slow to build raindrops of about a millimetre in the time a cloud lasts. Growth becomes faster once droplets reach roughly 15–20 μm in radius, because larger droplets fall more quickly and sweep up smaller ones, which then merge with them. This process, called collision and coalescence, can produce rain within tens of minutes. It is a leading route to rain in many tropical maritime clouds, although deep tropical convection also involves ice processes.

Ice can also multiply in clouds once it has formed. In the Hallett–Mossop process, which operates at roughly −3 °C to −8 °C, riming droplets that are large enough freeze and splinter, releasing many small ice fragments. This secondary production can greatly exceed the number of primary INPs, and it helps explain why some clouds contain far more ice crystals than their nucleating particles would predict. The release of latent heat when water freezes also affects cloud development, as it adds buoyancy to rising air.

Lifting mechanisms and atmospheric stability

A cloud forms when a parcel of air rises, expands and cools until its water vapour condenses. What lifts the air therefore shapes the cloud's type, horizontal extent and lifetime. Broadly, lifting comes from four sources: heating of the surface, forced ascent over terrain, the wedging of one air mass under or over another at a weather front, and large-scale convergence or radiative cooling. How far a lifted parcel keeps rising depends on the stability of the surrounding air. Stable air tends to produce shallow, layered cloud, while unstable air allows clouds to grow vertically.

Surface heating is the most familiar mechanism. Sunlight passes largely unabsorbed through the atmosphere and warms the ground, which in turn heats the air in contact with it. That air becomes less dense than its surroundings and rises as buoyant columns, or thermals. Over land these thermals are strongest in the afternoon, when the surface is warmest. Over the ocean, weaker thermal convection in fair weather produces the scattered, flat-based cumulus familiar from many coastal and tropical skies. The base of each cloud marks the level where rising air first reaches saturation. Under fair conditions the clouds stay small and short-lived. When the air is moist and unstable enough, the same process can build into towering cumulonimbus.

Orographic lifting occurs when air is forced up the slopes of a mountain. Rising air cools, and if it carries enough moisture, cloud and precipitation form on the windward side. The heaviest precipitation often falls there, particularly where the barrier lies close to the sea. Clouds that form over high ground can take several shapes. A band of relatively low but thick cloud, sometimes called a barrage cloud, may lie along the windward slope. Isolated peaks may be crowned by cap clouds or surrounded by collar-shaped clouds. On the lee side, air descends, warms by compression and can hold more moisture, so the cloud thins and dissipates. This produces the rain shadow that leaves downwind areas much drier. Where the flow generates standing waves, lenticular clouds can form downwind of the ridge and mark the wave crests. Hilly terrain also adds to thermal lifting, which is why hills are often cloudier than the surrounding lowlands.

Frontal lifting results from the density contrast between air masses. Warm air is lighter than cold air, so when the two meet it cannot push into the cold air and instead rides up over it. This overrunning happens at a gentle slope at a warm front. Lifting is slow and broad, so the cloud sequence ahead of the front is typically cirrus, then cirrostratus and altostratus, which thicken and lower into nimbostratus. These sheets can extend hundreds of kilometres ahead of the surface boundary and produce steady, widespread rain over many hours. A cold front has a much steeper leading edge, because dense cold air undercuts the warmer air and lifts it quickly. The result is more vigorous, localised uplift, with cumulus and cumulonimbus, showers, and in some cases squall lines and thunderstorms. Where a faster cold front overtakes a warm front, the resulting occluded front can combine both cloud patterns.

Large-scale convergence and radiative cooling produce the broadest cloud cover. In low-pressure systems, air converges near the surface and is forced upward over vast areas, which builds extensive layered cloud and precipitation. Radiative cooling works more locally. On clear nights the ground loses heat by emitting longwave radiation, and the air in contact with it cools. If that air is already close to saturation, it produces stratus or fog near the ground, often by morning. These mechanisms usually lift air far more gently than surface heating or fronts do, which explains why their clouds tend to be widespread, flat and long-lasting.

Classification system and nomenclature

Before the nineteenth century, clouds had no consistent scientific names. Observers described them by colour and shape, using folk terms. The modern system began in 1802, when Luke Howard, a London pharmacist with a strong interest in weather, presented a paper to the Askesian Society titled, in effect, "On the Modification of Clouds," where "modification" meant classification. Howard argued that every cloud belonged to one of three basic families, which he named with Latin words: cirrus ("curl of hair"), cumulus ("heap"), and stratus ("layer"). Because clouds change form as they evolve, he also introduced compound names such as cirrostratus and stratocumulus for transitional forms, along with nimbus for rain-bearing cloud systems. His fuller essay appeared in 1803, which is why both dates are cited in the literature.

Howard was not the only early attempt. Jean-Baptiste Lamarck proposed a French-language scheme in the same period, but it did not spread beyond France. Howard's choice of Latin, the working language of international scholarship, is generally credited with making his names durable.

From Howard's families to the WMO atlas

The current reference is the International Cloud Atlas of the World Meteorological Organization (WMO). It recognises ten basic genera, defined by where they form in the sky and by their approximate appearance. Their names extend Howard's roots with modifiers: stratus/strato for flat, layered, smooth cloud; cumulus/cumulo for heaped, puffy cloud; cirrus/cirro for feathery, wispy forms; and nimbus/nimbo for rain-bearing cloud. The prefix alto- marks mid-level clouds, even though the Latin word means "high." The ten genera are cirrus (Ci), cirrocumulus (Cc), cirrostratus (Cs), altocumulus (Ac), altostratus (As), nimbostratus (Ns), stratocumulus (Sc), stratus (St), cumulus (Cu) and cumulonimbus (Cb). Genera are the top tier of the scheme, and the atlas adds further layers beneath them.

The first subdivision is the species, which describes shape and internal structure. Examples include castellanus, for turrets that resemble a castle battlement, and volutus, a roll cloud added in the 2017 edition. Varieties describe transparency and arrangement; lacunosus, a honeycomb or chequerboard pattern, is one. Beyond these come supplementary features, such as asperitas and cavum, and accessory clouds, which are attached to a parent cloud. The atlas also lists special clouds that form under human or natural influence, including flammagenitus clouds, produced by rising heat from fires or volcanoes, and homogenitus clouds, which arise from human activity such as aircraft contrails. Taken together, the genera, species and varieties yield roughly 100 combinations.

Classification is based on several criteria at once. Visual form and internal structure are central, and the base altitude and vertical extent of a cloud are also important. The atlas has historically grouped genera into low (CL), middle (CM) and high (CH) levels, although the altitudes vary with latitude and season. Because the same cloud can look quite different at different stages of its life, or under different lighting and vantage points, observers may assign it different names on different occasions. Standard definitions reduce this variability but cannot eliminate it.

The main value of a shared vocabulary is comparability. Stations and countries that use the same genus, species and variety names can compare records directly, which supports forecasting, climate records and aviation safety. A fixed set of labelled categories also gives researchers a practical basis for training automated cloud-recognition systems.

Cloud families by altitude

Meteorologists group clouds into families according to the level of the troposphere where each type most often occurs. The World Meteorological Organization (WMO) divides this layer into three conventional levels, formerly called étages: high, middle and low. The boundaries are approximate, they vary with latitude, and neighbouring levels overlap. The levels therefore describe where a genus is usually found rather than a fixed altitude cut-off. The variation with latitude follows from the height of the tropopause, which lies higher over the equator and lower over the poles.

High clouds are the thin, wispy forms that sit highest in the troposphere. Their bases lie in the range shown above, and in the mid-latitudes they typically begin at about 5 km. The air at these heights is far below freezing, so these clouds are composed mostly of ice crystals. Cirrus appears as fibrous filaments, while cirrocumulus and cirrostratus form small rippled patches and whitish sheets. Because they are made of ice and are often thin, high clouds let much sunlight through, and they are frequently visible as halo-producing veils.

Middle-level clouds lie between high and low forms. Altocumulus forms grey or white rounded patches and rolls, and altostratus forms a greyish sheet that can thin enough for the Sun to appear as a dim disc. Because temperatures at these heights are often between the freezing point and well below it, these clouds may contain both supercooled water droplets and ice crystals. Altostratus is usually found in the middle level but frequently extends into higher levels, and nimbostratus, the rain-bearing sheet, is almost always in the middle level and often spans all three.

Low clouds occupy the lowest layer of the atmosphere, from the surface up to about 2 km at every latitude. Stratus and stratocumulus are the typical forms here, and at these temperatures they consist mainly of liquid water. Low clouds are the most common source of drizzle and of persistent grey overcast, and fog is a stratus-type cloud that touches the ground.

Clouds with vertical development are a different case. Cumulus and cumulonimbus generally have their bases in the low level, but their vertical extent is so large that their tops can reach the middle and high levels. Small fair-weather cumulus that barely grows taller than it is wide is still classed as a low cloud because of its limited depth. Only the most powerful cumulonimbus towers can push their anvil tops toward the tropopause, the boundary between the troposphere and the stratosphere described in the atmosphere of Earth article.

Altitude alone is not decisive, so observers combine the height of the cloud base with its form and structure. Because the same cloud can be assigned to different levels depending on latitude and season, the family system works as a convention for consistent observation rather than as a physical law. The weather of a given day, the temperature profile, and the amount of water vapour in the air together determine whether a cloud forms at a high, middle or low level.

Rare and special cloud forms

Mammatus clouds are pouch-like lobes that hang from the underside of a parent cloud, most often the anvil of a cumulonimbus thunderstorm. The name comes from the Latin mamma, meaning "udder" or "breast". Most clouds develop as air rises and cools, but mammatus form where air descends: pockets of cold, cloud-laden air sink into the drier, warmer air beneath the anvil, and the cloud base sags into rounded lobes. The exact mechanism is still debated. One widely discussed idea holds that the evaporation or sublimation of falling ice and precipitation cools the air and makes it denser, which encourages descent, while other accounts stress turbulence and moisture contrasts. Individual pouches typically last only about ten minutes to half an hour. They signal unstable air and often accompany storms, but they are not dangerous in themselves and are not a reliable sign of tornadoes.

Lenticular clouds are smooth, lens- or saucer-shaped clouds that form downwind of mountains and hills. When stable, moist air is pushed over a ridge, it is forced upward, cools, and condenses at the crests of a train of standing waves, known as lee waves or mountain waves. On the downwind side the air sinks and warms, and the droplets evaporate. The cloud therefore forms continuously on the upwind side of each crest and dissipates on the downwind side, so it looks motionless even though air streams rapidly through it. Lenticular clouds are visible signs of mountain waves, which can exist even when no cloud forms. Strong waves may produce long strings of clouds known as wave clouds, and the cap that sits directly above a peak is sometimes called a cap cloud. Pilots must cope with updrafts, downdrafts, and rotor turbulence near them, whereas sailplane pilots often seek out the lift in mountain waves. Depending on altitude they are classified as altocumulus, stratocumulus, or cirrocumulus lenticularis, and their saucer shapes have led to repeated mistaken reports of UFOs.

Nacreous clouds, also called polar stratospheric clouds (PSCs) or mother-of-pearl clouds, form in the stratosphere at roughly 15–25 km, well above the weather layer where ordinary clouds develop. They require extreme cold, generally below about −78 °C, and therefore appear only during the polar winter, when the stratospheric polar vortex is cold enough. They are most often seen from high-latitude regions such as Scandinavia, northern Canada, Alaska, and Antarctica. Their particles are small and uniform, so they diffract sunlight into pastel, iridescent colours. Because they lie high up, they catch sunlight after it has set for observers below, glowing before dawn and after dusk when the sun is only a few degrees below the horizon. Type II PSCs are composed of ice, while Type I particles are droplets of nitric and sulphuric acid. Both types provide surfaces for chemical reactions that release reactive chlorine, which is why these clouds are central to the formation of the ozone hole.

Noctilucent clouds occupy a much higher layer, the mesosphere, at around 80 km, and are the highest clouds in the atmosphere. They consist of water ice that nucleates only at very low temperatures, and they are normally seen in the summer months, mostly from June to July at mid-to-high latitudes. They are visible only after sunset, when the lower sky is dark but the cloud layer, far above the ground, is still lit by the sun. In December 2024, however, clouds resembling summer noctilucent clouds were observed over Siberia in winter at a mean altitude of about 70 km, following a sudden stratospheric warming.

Contrails are another human-made special cloud form. Engines emit water vapour, along with soot and other particles, and in cold air near cruising altitude the vapour condenses and freezes around these particles into a visible trail of ice crystals. Whether the trail vanishes within seconds or persists depends largely on whether the surrounding air is supersaturated with respect to ice. Where it is, a contrail can last for hours and spread under the wind into thin, cirrus-like sheets. Pyrocumulus clouds are formed by a different process: the intense heat of a large fire lifts moist air rapidly, and when the water vapour condenses, the latent heat released drives further ascent, building a cumulus cloud above the blaze. The strongest of these can grow into pyrocumulonimbus clouds.

Thunderstorms, electrification and lightning

A thunderstorm grows from a cumulonimbus cloud, a towering convective cloud whose interior is driven by strong updrafts and downdrafts. Warm, moist air rises rapidly through the cloud, while falling precipitation drags cooler air downward. When the rising air reaches the stratosphere, where temperature increases with height and the air is very stable, the updraft loses its buoyancy and spreads sideways. This produces the flat, fibrous anvil that often crowns mature storms.

Thunderstorms are also electrically active, and the charge separation that powers lightning is thought to arise mainly from collisions in the mixed-phase region of the cloud, where supercooled water, ice crystals and graupel (soft, pellet-like rime ice) coexist. The leading explanation is the non-inductive charging mechanism. When a larger graupel particle strikes a smaller ice crystal, electrons move between them, and the sign of the charge transferred depends on temperature and on the amount of liquid water present. In the typical case, colder ice crystals acquire a positive charge and are carried upward by the updraft, while heavier graupel becomes negatively charged and falls to the middle and lower parts of the cloud. The result is a large electric dipole, with positive charge near the top and negative charge in the middle, often with a smaller positive pocket near the base. Because the detailed microphysics and the role of soluble ions remain under study, this picture is best described as the prevailing model rather than a settled account. Charge is continually re-separated by the storm's circulation, and the process is sensitive to updraft speed, which may partly explain why lightning is more frequent over land than over ocean.

When the potential difference across the cloud grows large enough for the local electric field to exceed the breakdown strength of air, a discharge begins. Many discharges stay inside the cloud as intra-cloud flashes, while others connect to the ground. A cloud-to-ground strike typically starts with a stepped leader, a channel of ionized air that advances downward in short increments, and when it meets an upward streamer rising from the surface, a bright return stroke carries a large current back up the channel. The sudden heating of the air along this channel produces a rapid expansion that we hear as thunder. Lightning is therefore a electrical discharge that neutralizes the charge which the storm has built up, and its visible flash is a large-scale transfer of energy through air that has been briefly converted into a conductive plasma. Lightning is one of the most visible expressions of this process.

Hail forms through a related cycle of repeated lifting. Ice particles are carried upward by strong updrafts, where they collect supercooled droplets that freeze onto their surfaces, then fall back through the cloud, and are lifted again. Each pass adds a layer of ice, and the stones can grow to considerable size before the updraft can no longer support them and they reach the ground. Hailstones therefore often show concentric layers when cut open. Their maximum size depends on the strength and persistence of the updraft and on the height of the freezing level, which is why the largest hail is usually associated with the most intense convection. Hail is one of several precipitation types that reach the surface from clouds.

Supercells are the most severe form of thunderstorm. A supercell is a long-lived storm, typically lasting more than an hour, built around a deep, persistently rotating updraft known as a mesocyclone. Wind shear, in which wind speed or direction changes with height, causes horizontal rolls of air to be tilted into vertical rotation within the storm. The mesocyclone is generally a few kilometres across and can be detected by Doppler radar as a hook-shaped echo. Although supercells are the least common of the principal thunderstorm types, they can produce the largest hail, damaging straight-line winds, and the strongest tornadoes. A tornado typically forms when the rotation near the base of the mesocyclone tightens and extends down to the ground, and most of the most violent tornadoes come from supercells. Not every supercell produces a tornado, however, and the factors that separate tornadic from non-tornadic storms remain an active research question.

Precipitation and the hydrological cycle

Clouds are the source of nearly all precipitation. Water vapour condenses onto particles and grows into droplets or ice crystals, and once these become heavy enough to overcome the updrafts that hold them aloft, they fall. The form that reaches the ground depends mainly on the temperature profile the particles pass through. Drizzle consists of droplets smaller than about 0.5 mm and typically falls from low stratiform clouds. Rain forms either by collision and coalescence in warm clouds or through the ice phase in colder ones. Snow falls when the whole column between cloud and ground stays below freezing. Sleet consists of raindrops that refreeze into small ice pellets in a sub-freezing layer near the surface. Freezing rain occurs when snowflakes melt in a warm layer aloft and the resulting liquid drops become supercooled before striking a surface below freezing, where they freeze on contact. Hail is built up inside strong convective storms and is discussed in the thunderstorm section.

Clouds are also the main link in the water cycle. Water is lifted into the atmosphere by evaporation from the ocean, lakes and rivers, soil, and vegetation, and the vapour is carried by atmospheric circulation until it condenses into cloud. Precipitation returns that water to the surface, and globally the amount falling each year balances the amount evaporating, roughly half a million cubic kilometres in a typical year. Most of this precipitation falls over the oceans, but precipitation over land is essential for freshwater supply. Snow and ice that accumulate on mountains and in polar regions feed glaciers and seasonal snowpacks, which store water and release it slowly through the year.

Not all precipitation reaches the ground. Virga consists of streaks of rain or snow that visibly fall from a cloud base but evaporate before touching the surface. Virga moistens the air layer beneath the cloud without wetting the ground, and it is common in deserts and other regions where the air near the surface is dry. Evaporation absorbs heat from the surrounding air, so the process cools the air and can make it denser. In strong cases this produces downdrafts that spread outward at the surface as gusty winds.

The fraction of water that condenses in a cloud and eventually falls as precipitation is called precipitation efficiency. Much of the condensed water is lost before reaching the ground, through re-evaporation in the sub-cloud layer, through detrainment into the surrounding air, or simply by remaining suspended as cloud. The efficiency varies with cloud type, the humidity of the air below the cloud, the intensity of convection, and the concentration of aerosols that influence droplet growth. Because these factors differ from one region to another, the conversion of cloud water into surface water has a direct effect on regional water availability. Changes in this efficiency are one reason that projections of changing precipitation patterns remain uncertain in some regions, even though a warmer atmosphere can hold more water vapour and generally intensifies the hydrological cycle.

Radiative effects and the energy budget

Clouds are among the main controls on how much solar energy the Earth keeps and how much it radiates back to space. Their influence is usually expressed as cloud radiative effect (CRE), the difference between the net radiative flux with clouds present and the flux that would occur under hypothetical clear skies, typically evaluated at the top of the atmosphere. Averaged over the globe, where cloud cover hovers near 60%, the combined effect is a net cooling of roughly 20 W m⁻².

That net figure hides two opposing processes. The first is reflection. Water droplets and ice particles scatter a large share of incoming shortwave sunlight back out to space, and the fraction returned is governed by the cloud's albedo. Global-mean estimates of this shortwave effect are near $-50\ \mathrm{W\,m^{-2}}$, meaning clouds remove that much energy from the sunlight the planet would otherwise absorb. The second process is trapping. Clouds absorb infrared radiation coming up from the surface and lower atmosphere, then re-emit it in all directions, including back downward. Because cloud tops are colder than the surface, they also send less energy to space than a clear sky would. The longwave effect is about $+30\ \mathrm{W\,m^{-2}}$ globally, and the difference between the two terms gives the net cooling.

The sign of a single cloud's effect depends on how these two terms balance, which in turn depends on its altitude, thickness, phase and the surface beneath it. Thick, low stratocumulus decks are bright and sit at temperatures close to the surface, so they reflect far more sunlight than they add to the greenhouse effect, and they tend to cool. High cirrus clouds work the other way. They are cold and fairly transparent to sunlight, so they reflect little while still blocking a substantial amount of outgoing infrared, and they typically warm. Deep convective clouds, whose bright upper parts and cold anvils offset each other, are often treated as close to neutral. Thin cirrus is the most delicate case: studies show that the net sign can change with assumptions about ice-crystal shape and size, because those assumptions shift the balance between reflection and trapping, and the net effect of such thin layers is small in magnitude.

The same cloud can also give different numbers depending on the level at which the energy balance is evaluated. Its effect at the top of the atmosphere, at the surface and within the atmospheric column are not identical, and when these are averaged over the globe, positive and negative contributions from high and low clouds partly cancel. This is one reason the distinction between cloud effects on the surface and on the atmosphere matters when interpreting regional energy budgets.

Because the net cloud term is large relative to other components of the planetary energy balance, small changes in it can matter for climate. For scale, a recent satellite-based estimate put the longwave effect of the partly cloudy "twilight zone" around shallow ocean clouds at about $0.75\ \mathrm{W\,m^{-2}}$, which is comparable to the forcing from raising atmospheric CO₂ by 75 ppm. Such comparisons show why cloud radiative forcing is treated as a major component of the Earth's energy budget, and why its accurate representation in climate models is so demanding. The response of clouds to warming, which can amplify or dampen the change in radiative forcing that greenhouse gases produce, is discussed in a later section.

Observation and measurement

Ground-based lidar and ceilometers

Cloud base height at airports and weather stations is usually measured with a ceilometer, a lidar instrument that sends short laser pulses straight up. The instrument times the echo returning from droplets and ice crystals. Because light travels at a known speed, that delay gives the height of the lowest cloud boundary, and the strength of the echo reveals how dense the layer is. Modern units can track several layers at once. When precipitation or fog hides the cloud base, the instrument reports vertical visibility instead. The main limitation is attenuation: a thick cloud or heavy rain absorbs the beam quickly, so upper layers may never be detected. For clouds with high optical depth, roughly above 3, radar is needed to locate the top. Human observers still report cloud type, amount, and approximate height as part of standard surface weather observations.

Satellite imagery

Satellites view clouds through two main kinds of channel. Visible channels record sunlight reflected from cloud tops and the surface. They give the sharpest image but work only in daylight. Infrared channels record thermal emission, so they operate day and night. Because temperature in the lower troposphere generally falls with height, the infrared brightness temperature of a cloud top acts as a proxy for its altitude, and very cold tops indicate high cloud. Converting this into a height requires an assumption that the temperature sensed at the top equals the ambient air temperature there. Analysts then match that temperature to a sounding or model profile to find the corresponding pressure. For low clouds over the ocean, where the sea surface temperature is known, a simpler estimate is

where $T_{B}$ is the infrared brightness temperature, $T_{\text{sfc}}$ the corrected sea surface temperature, and $\Gamma$ the local lapse rate. Split-window methods compare two nearby infrared wavelengths, near 10.8 µm and 12 µm, to estimate both the height and the optical thickness of upper-level cirrus. Retrievals have known weaknesses. Thin clouds let radiation from the warmer surface contaminate the signal, and thick, inhomogeneous clouds viewed at high zenith angles tend to have their tops underestimated.

Cloud radar

Millimetre-wavelength cloud radars occupy a different niche. Most operate near 35 GHz or 94 GHz, which correspond to the Ka and W bands, and they are commonly zenith-pointing ground instruments. Their short wavelength makes them sensitive to small droplets and to individual ice particles, which conventional centimetre-wavelength weather radars detect poorly. Doppler processing also yields the vertical velocities of cloud particles, which helps distinguish updrafts and the fall speeds of ice. Such radars have been used to resolve fine structure in thin cirrus and marine stratus, and to estimate liquid water and ice content profiles. The trade-off is that the signal weakens rapidly in heavy rain, so these instruments are most useful for non-precipitating and lightly precipitating clouds. In practice they are often combined with ceilometers and microwave radiometers to relate a cloud's geometry to its microphysical properties.

Aircraft measurements

Aircraft in-situ instruments fly through clouds and sample them directly. Probes can measure droplet size distributions, ice content, and aerosol concentrations, quantities that remote sensors can only infer indirectly. These measurements are essential for testing retrieval algorithms and for validating models. However, flights are expensive and cover only limited regions and time periods. Sampling remote oceans, polar regions, and other sparsely observed areas is especially difficult, so the in-situ record is much thinner than the satellite and ground-based datasets it is used to check.

Clouds in climate models and feedbacks

Global climate models divide the atmosphere into a lattice of grid boxes. Each box typically spans about 200 to 1,000 km horizontally, far wider than the cumulus towers or stratocumulus decks that control how much sunlight the planet reflects. These boxes can vary in size from 2°–10°, latitudinally and longitudinally (approximately 200–1,000 kilometers), and tens of kilometers in height. Because most cloud processes operate below that scale, models approximate their average effect on each box through parameterizations rather than simulating individual clouds. Cloud formation processes are mostly at sub-grid scales and are therefore represented by approximate parameterizations.

Clouds matter for climate because they change as the surface warms. Their amount, height, and optical thickness can shift, and each shift alters both the sunlight reflected to space and the infrared radiation that escapes. Clouds respond to CO2-driven environmental changes, such as rising surface temperature, in ways that can amplify or dampen climate change. A positive cloud feedback adds to the warming produced by CO2 and other greenhouse gases, while a negative one offsets part of it. The main yardstick is equilibrium climate sensitivity (ECS), the steady-state warming that a model reaches after atmospheric CO2 doubles, with ice sheets and vegetation held fixed. ECS is defined as the steady-state warming in response to a doubling of atmospheric CO_2 in a model with fixed ice sheets and vegetation. The Intergovernmental Panel on Climate Change (IPCC) fifth assessment report of 2013 gave a range of 1.5 to 4.5 °C for ECS.

Assessments have judged the net cloud feedback to be positive, but its size is poorly pinned down. The AR5 report concluded that the "net cloud feedback" is likely positive. The report produced a central estimate of +0.6 Watts per metre squared per degree C (W/m2/C), with an uncertainty range between -0.2 and +2.0 W/m2/C. The spread is largely a low-cloud problem: the IPCC stated that differences in the response of low clouds to a warming are responsible for most of the spread in model-based estimates of equilibrium climate sensitivity. Models in the sixth Coupled Model Intercomparison Project (CMIP6) generally ran warmer than their predecessors, and this was traced largely to stronger positive cloud feedbacks, linked to a larger loss of low cloud cover in extratropical regions. The higher climate sensitivity in the CMIP6 models is primarily attributable to the strengthened positive cloud feedbacks associated with the greater reduction in low cloud cover in the extratropical regions.

Attempts to narrow the range now combine model output with observations. A 2026 study by Wilson Kemsley, Nowack and Ceppi analysed high and low cloud feedbacks separately, using relationships between cloud properties and their controlling factors in satellite data. It finds a robustly positive cloud feedback (i.e., a negative feedback is <0.5% probable), indicating that clouds amplify global warming. Two of its three methods also indicate an upward ECS shift with reduced uncertainty. An independent 2020 assessment drawing on several lines of evidence gave a 66% probability that ECS lies between 2.6 K and 3.9 K. A recent assessment of climate sensitivity using multiple lines of independent evidence produces a probability of 66% that the ECS lies between 2.6 K and 3.9 K.

Cloud effects are also hard to isolate because clouds act in several directions at once. In one experiment with the MPI-ESM1.2 Earth system model, making clouds transparent to radiation raised ECS from 2.84 K to 4.29 K. The result reflects clouds' masking of CO2 forcing and their influence on other feedbacks, such as the surface albedo feedback, rather than the cloud feedback term alone. Clouds have a masking effect on CO2 forcing and can influence other feedback mechanisms such as the surface albedo feedback.

Modellers are trying to reduce reliance on parameterization. Convection-permitting models, with horizontal grid spacings of roughly 1 to 4 km, can explicitly resolve cumulus-scale motions. Convection-permitting models with approximately 1–4 km horizontal grid spacing can explicitly resolve cumulus-scale atmospheric motions and therefore reduce or eliminate the need for conventional convective parameterization. Their advantages come with costs. The drawback of CPMs is the high demand on computational resources. They also carry their own biases: CPMs continue to exhibit persistent biases, notably the overestimation of precipitation intensity and cloud depth with underestimation of stratiform precipitation area. Global models that can run at these resolutions over long climate timescales remain limited, so parameterized clouds are likely to remain central to projections of future warming for some time.

Human influences on clouds

Human activity changes clouds mainly by adding particles to the air. Because aerosols provide the condensation nuclei on which water condenses, extra particles from fossil fuel combustion and other sources can alter cloud droplet numbers, liquid water and cloud cover, and these changes are subject to large uncertainties. The overall group of responses is known as aerosol–cloud interactions. For a fixed amount of water, splitting it among more, smaller droplets makes a cloud more reflective, a brightening usually called the Twomey effect. The same process can also shift when and where precipitation falls, since smaller droplets are less efficient at producing rain.

Ship tracks are the most visible example of this process. Ship tracks, long lines of polluted clouds that are visible in satellite images, are one of the main tools for quantifying aerosol–cloud interactions. They usually sit within marine low clouds and appear brighter than the surrounding cloud, because the ship-derived particles increase droplet counts. An individual track can persist for a few hours to more than 24 hours before diffusing and mixing back into the atmosphere. Timing studies show that the response unfolds over different timescales: in one analysis, the number of droplets in a track rose within an hour, while the actual amount of water in the cloud continued to change over hours, likely beyond the 20-hour limit of the study. Visible tracks are also an incomplete record, since only a small fraction of the clouds polluted by shipping show ship tracks. Including the tracks that cannot be seen, researchers infer a substantial cooling from changes in cloud water, which could mean that the climate is more sensitive to warming than observed temperature trends alone would suggest.

Cloud seeding applies a similar physical principle deliberately, but with the goal of producing precipitation. Vonnegut conducted experiments with silver iodide as a seeding agent. He found that silver iodide's physical structure closely resembles the structure of ice, and thus it is effective at forming ice crystals in supercooled clouds. Silver iodide became the most common seeding material, and it is often cited as remarkably efficient, since merely one gram of silver iodide can precipitate up to 10 trillion artificial ice crystals. Its effectiveness at the scale of real weather remains contested. A 2019 review concluded that seeding supercooled orographic clouds produces ice-particle plumes that drift downwind and reach the ground. Other claims are more controversial: analyses of Australian experiments reported apparent effects of cloud seeding with silver iodide that last not only for many days but even for many months, yet one critique argues that there appears to be a flaw in some (but not all) of the analysis which exaggerates the time span. The underlying mechanism is also incompletely settled, as scientists have not yet established the exact mechanism of the freezing process. Environmental concerns have been raised about its potential toxicity to aquatic life and the broader ecosystem, while proponents point out that silver iodide retains its form in water and does not break down into the potentially toxic silver ion.

Aircraft contrails are a human-made cloud type with a measurable climate effect. They form behind aircraft at around an altitude of 10-11km. They only form in sufficiently cold and humid regions of the atmosphere, where water vapour condenses on to the soot particles emitted by aircraft engines. Persistent contrails result when airplanes fly through layers of high humidity conditions (known as super saturated air). They can last for a few minutes or longer than a day. Long-lived contrails are most common where cirrus already exists: one analysis of humidity data found that such conditions are met ~90% over the Northern midlatitudes and almost 100% in the Southeast Asian subtropics, mostly within subvisible or visible cirrus. Their warming adds to the effect of the carbon dioxide that aviation emits. Although contrails reflect sunlight by day, the scientific consensus is that contrails overall have a net warming effect, but the magnitude is uncertain, with around 70% uncertainty in how much warming contrails contribute. One modelling study estimated that the 2019 global net radiative forcing from contrails lies between $34.8$ and $74.8\ \text{mW m}^{-2}$. Simulations also suggest that altitude is a practical lever: shifting the altitudes of 1.7 per cent of flights by about 2,000 feet could cut the contrail forcing by 59 per cent in the airspace studied, with less than a tenth of a per cent increase in fuel consumption.

Clouds on other planets and moons

Venus is wrapped in a planet-wide deck of clouds that is unlike anything on Earth. The droplets are mainly concentrated sulfuric acid solution, along with minor contaminants, and their acid content is estimated at 75% to 92% by weight. The cloud tops lie roughly 80 kilometres above the surface. Because this layer is highly reflective and strongly absorbs infrared radiation, it controls much of the planet's radiative balance. Photochemistry is thought to create the acid: sulfuric acid vapour is first produced above about 70 km, and the vapour then condenses into droplets. Near the cloud tops, winds exceed 300 kilometres per hour and circle the planet in about four days. The droplets are also extremely dry; calculations of their water activity suggest they could not support a habitable environment, although the question of microbial life in the clouds is still debated.

Mars has a thin atmosphere, predominantly composed of CO2 (around 95%), and its clouds are made of two kinds of ice. Water-ice clouds are the more familiar type. They gather over the flanks of large volcanoes, and the north polar hood forms during the northern winter as a cap of water and carbon dioxide ice clouds over the pole. As the seasons change, an equatorial belt of water ice clouds forms and has become as characteristic of the Martian climate as the southern summer dust-storm season. Most Martian clouds stay below about 60 km, and they are composed of water ice. Carbon dioxide ice clouds are higher and more elusive. The first clear detection came from orbit when the OMEGA instrument aboard ESA's Mars Express detected them for the first time in late June 2004. Some are dense enough that they can reduce the Sun's apparent brightness by up to 40 per cent, and the shaded ground can be up to 10°C cooler than its surroundings. In March 2021, the Curiosity rover photographed twilight clouds at high altitude, which were likely made of frozen carbon dioxide because of the very cold conditions there. Only this type of cloud has been seen to produce iridescence on Mars, and such clouds appear near the top of the images at altitudes of about 60 to 80 kilometres.

The atmospheres of Jupiter and Saturn are organised into cloud layers set by temperature and composition. In the upper layer, the visible clouds are mostly ammonia ice; deeper down, ammonium hydrosulfide clouds form, and water clouds lie deeper still where pressures are higher. The banded appearance of both planets reflects their zonal circulation: brighter zones are associated with rising air and darker belts with sinking air, and the contrast is much stronger on Jupiter than on Saturn, where haze mutes the bands.

Titan, Saturn's largest moon, is the only place in the solar system besides Earth with a surface-level hydrological cycle. Titan has a dense atmosphere, which supports an Earth-like hydrological cycle of methane clouds, rain, and liquid flowing across the surface to fill lakes and seas. Methane is the working fluid: methane rises to form clouds and then sometimes falls back to the surface as rain, which is Titan's methane cycle, analogous to Earth's water cycle. With surface temperatures around -180 C, methane remains liquid, and the surface's frozen water ice acts like bedrock. Rain is infrequent: it may be centuries between showers at a given location. Titan's dense air and low gravity slow the fall of its raindrops. In 2025, observations from the James Webb Space Telescope and the Keck II telescope, taken in 2022 and 2023, revealed clouds at mid and high northern latitudes, where Titan is currently experiencing summer, and the clouds appeared to rise to higher altitudes over time, suggesting active convection. These findings were published in Nature Astronomy in May 2025.

History of the study of clouds

Aristotle's Meteorology, written in ancient Greece, treated clouds, rain, and winds as products of exhalations rising from the earth. It placed them inside a broader natural-philosophical system rather than an empirical catalogue, and that framework shaped European thinking on the subject for centuries. Aristotle's account stayed influential long after observers had begun to measure the sky.

Before the nineteenth century, observers often regarded each cloud as a unique, essentially unclassifiable event. Their records tended to describe colour or personal impression rather than fixed categories. Before the 19th century, many meteorologists thought of each cloud as unique, unclassifiable and in a state of temporary existence. Instead of strict descriptions clouds were recorded by colour or individual interpretation.

The change came from Luke Howard, a London pharmacist and amateur meteorologist. In December 1802 he presented a paper, "On the Modifications of Clouds," to the Askesian Society, a London discussion group of amateur scientists. He argued that every cloud belonged to one of three principal families, to which he had given the Latin names: cirrus (meaning "fiber" or "hair"), cumulus ("heap" or "pile"), and stratus ("layer" or "sheet"). To cover transitional forms, he added compound names such as cirrostratus and stratocumulus, borrowing the binomial logic of Linnaean taxonomy. Howard was not the first to attempt a classification of clouds—only the previous year Jean-Baptiste Lamarck had proposed a list of descriptive terms in French—but the success of Howard's system was due to his use of universal Latin. The essay appeared in print in 1803 in the Philosophical Magazine. Once published, the cloud classification was soon in use around the world, and in 1896 it was officially adopted (with minor amendments) by the World Meteorological Organization. Howard's scheme also drew literary responses in Europe; the German poet Johann Wolfgang von Goethe, for one, wrote verses in Howard's honour.

Cloud chambers and the physics of condensation

In the early twentieth century, clouds became an instrument of physical research. The Scottish physicist C. T. R. Wilson had originally set out to study atmospheric phenomena. Wilson won the 1927 Nobel Prize for his development of the cloud chamber, which he originally undertook to study atmospheric phenomena. He found that ions could act as centers for water droplet formation. His first chamber, built in 1911, expanded saturated air to cool it. When an ionizing particle passes through the chamber, water vapor condenses on the resulting ions and the trail of the particle is visible in the vapor cloud. This made the paths of charged particles photographable, and during the first half of the 20th century, experiments that looked at cosmic rays passing through cloud chambers revealed the existence of several fundamental particles, including the positron, the muon and the first strange particles. Wilson shared the 1927 Nobel Prize in Physics with Arthur Compton for this work. The device linked the study of cloud droplets to particle physics, and it used the same condensation principles that govern natural clouds.

The satellite era

Observation of clouds moved into orbit in 1960. The first experimental weather satellite, TIROS 1, was launched from Cape Canaveral on April 1, 1960, by a Thor-Able launch vehicle. Its two television cameras provided weather forecasters their first ever view of cloud formations as they developed around the globe. The early pictures showed spiral formations associated with large storms. Although the satellite operated for only 78 days, TIROS-1 sent back more than 19,000 usable pictures, proving the worth of weather observing satellites to the world and opening the door for weather technology of the future. Continuous coverage followed in 1962, and the program's data supported the first accurate weather forecasts based on data gathered from space. The shift from ground-based description to remote sensing from orbit created the observational basis for modern weather forecasting.

Clouds in art, literature and culture

In East Asian art, cloud motifs have long carried meanings beyond weather. Stylized clouds, often shown as curling scrolls, appear in painting, calligraphy, ceramics, textiles and lacquer. The so-called auspicious cloud (xiangyun) signals good fortune, and clouds that surround mountains and pavilions in landscape scrolls suggest the realm of immortals. In such works, mist and cloud are not incidental: they organize space, separate near from far, and mark the transition between the earthly and the spiritual.

European painting took a different route. In the early nineteenth century, John Constable made a long series of small oil studies of the sky over Hampstead, near London, and many of them were annotated with the date, the time of day and the wind direction. He treated clouds as a subject of direct, repeated observation rather than as a decorative backdrop. Those studies helped change how landscape painters approached the sky, and they remain a useful record of atmospheric conditions for historians of meteorology as well as of art.

In poetry and prose, clouds usually work as images of impermanence and mystery. Their shapes change constantly and they cannot be held, so writers have used them to stand for transience, for dreams and for the unreachable. Percy Bysshe Shelley's poem "The Cloud" (1820) gives the cloud a voice that speaks of its endless cycle of rising, falling and returning, while William Wordsworth's celebrated lyric of 1807 uses a drifting cloud as a figure of solitude. Across traditions, the cloud often sits on a boundary, between heaven and earth, between the visible and the invisible, or between one state of being and another.

The word has also crossed into technology, where it has become a metaphor for something that is remote, shared and not directly seen. Network diagrams had used a cloud-shaped outline to represent an unspecified network for years before the late 2000s, when the term cloud computing spread widely to describe software and storage delivered over the internet. The borrowing is a small but clear example of how a term from weather science can acquire entirely new meanings in culture and industry, while the original sense of a visible mass of condensed water remains in the background of its use.

Unresolved questions in cloud science

Ice nucleation between −40 °C and 0 °C

Pure water can remain liquid down to roughly −38 °C before it freezes spontaneously, so ice in the real atmosphere depends on foreign particles that lower the energy barrier to freezing. Particles that work at the warm end of this range are best understood. Biological materials are the most effective natural ice nucleators known: bacterial proteins can trigger freezing near −2 °C, and fungi, plants and invertebrates also produce active material at about −2 to −7 °C. Below roughly −15 °C, where mineral dust and other mineral particles become the relevant seeds, the picture is far less complete. The biological particles themselves are also hard to study, since some ice nuclei are unstable when heated and remain poorly characterised, and why organisms produce them is not clear. Field observations and models both indicate that biological ice nuclei influence precipitation and possibly regional climate, yet simulated contributions fall well short of the observed ones. The temperature dependence of ice nucleation in the colder part of the range, and the role of biological particles across it, therefore remain incompletely explained.

Aerosol effects on clouds and radiative forcing

Aerosols modify clouds by acting as condensation nuclei and, in some cases, as ice nuclei. More particles generally produce more and smaller droplets, which can brighten clouds and reduce precipitation efficiency. The size of this effect, however, depends on aerosol composition, the cloud regime, humidity, and the background meteorology, and these factors are difficult to separate from natural variability in observations. Aerosol–cloud interactions are therefore among the least certain parts of the forcing that drives the climate system. Because aerosol cooling has partly offset the warming from greenhouse gases, uncertainty about this offset feeds directly into estimates of how strongly the climate responds to greenhouse gases, and into how much of the observed warming can be attributed to each cause.

Low-cloud feedback under warming

Marine low clouds exert a strong cooling effect on the planet, and how that cooling changes as the climate warms is a leading source of uncertainty in climate sensitivity, meaning the warming produced by a doubling of carbon dioxide. A 2021 study that used satellite observations to constrain this feedback estimated a near-global marine low-cloud feedback of 0.19 ± 0.12 W m⁻² K⁻¹, which implied a moderate climate sensitivity of about 3 K. The same study found that models with erroneously positive trade-cumulus feedback produce unrealistically high sensitivities, while models with overly weak low-cloud feedback yield unrealistically low ones. Not all analyses agree on the implications. A 2026 paper in Geophysical Research Letters argues that recent cloud-controlling-factor analyses point to a higher sensitivity. In the tropics, the low-cloud feedback is closely tied to deep convection and to the response of the overturning circulation to warming, and observational constraints for the tropical Atlantic and Pacific can pull in opposing directions.

Kilometre-scale representation and observation over remote oceans

Global climate models use grids far too coarse to resolve individual clouds, so microphysics and convection must be approximated through parameterisations. Higher-resolution simulations with explicit convection can represent these processes more directly, but their computational cost keeps them from being used routinely for long climate projections. Progress on this front depends on better observations to test the models. Remote oceans, where extensive low clouds form, remain sparsely sampled by surface stations, aircraft and ships, so the observational record there is thin. Closing these gaps is therefore a prerequisite for narrowing the uncertainties in Earth's energy budget and in projections of future climate change.

References

1. Cloud Droplet - an overviewsciencedirect.com
2. Liquid water contenten.wikipedia.org
3. Clouds fog and water Dropletsatoptics.co.uk
4. VISIBILITY AND LIQUID WATER CONTENT IN CLOUDS IN THE FREE ATMOSPHERE in: Journal of the Atmospheric Sciences Volume 7 Issue 1 (1950)journals.ametsoc.org
5. Cloudeducation.nationalgeographic.org
6. Cloud liquid water content, drop sizes, and number of dropletswww-das.uwyo.edu
7. Cloudsatmo.arizona.edu
8. Surfactants and cloud droplet activation: A systematic extension of Köhler theory based on analysis of droplet stabilitypubs.aip.org
9. Cloud Droplet - an overviewsciencedirect.com
10. Dynamical regimes of CCN activation in adiabatic air parcelsarxiv.org
11. Cloud Condensation Nucleus - an overviewsciencedirect.com
12. Cloud drop effective radiusen.wikipedia.org
13. Differences in liquid cloud droplet effective radius and number concentration estimates between MODIS Collections 5.1 and 6 over global oceans - PMCpmc.ncbi.nlm.nih.gov
14. On the Relation among Satellite-Observed Liquid Water Path ...vandenheever.atmos.colostate.edu
15. ACP - Importance of aerosols and shape of the cloud droplet size distribution for convective clouds and precipitationacp.copernicus.org
16. 7.2: Nucleation of Liquid Droplets - Geosciences LibreTextsgeo.libretexts.org
17. Orographic influence on the windward sidecloudatlas.wmo.int
18. Orographic cloudsuser.eumetsat.int
19. Orographic Lift, Explainedopensnow.com
20. Front Fundamentals - AOPAaopa.org
21. Cloud Formation Processes I. Introductionuvm.edu
22. Lifting Along Fronts: when air masses interact - WW2010ww2010.atmos.uiuc.edu
23. Chapter 12: Fronts and Airmassespressbooks-dev.oer.hawaii.edu
24. Weather fronten.wikipedia.org
25. Luke Howardencyclopedia.com
26. Luke Howard and Cloud Namesarchive.rmets.org
27. Gathering clouds - Science Museum Blogblog.sciencemuseum.org.uk
28. Clouden.wikipedia.org
29. Luke Howard: namer of the clouds · Creation.comcreation.com
30. Classifying cloudswmo.int
31. International Cloud Atlas and new cloud classificationsrmets.org
32. International Cloud Atlas translated into official UN languagespublic.wmo.int
33. Generacloudatlas.wmo.int
34. 6.4: Cloud Classification - Geosciences LibreTextsgeo.libretexts.org
35. WORLD METEOROLOGICAL ORGANIZATION INTERNATIONAL CL·OUD ATLAS Volume Icloudatlas.wmo.int
36. Definitions of cloudscloudatlas.wmo.int
37. The classification and nomenclature of cloudshko.gov.hk
38. Mammatus clouds: Pouch-like formations and what they meantimsweather.au
39. Weather Words: Mammatus Cloudsweather.com
40. The beauty and mystery of mammatus cloudssevere-weather.eu
41. Unusual cloud formations - Met Officeweather.metoffice.gov.uk
42. UBC ATSC 113 - Lenticular cloudseoas.ubc.ca
43. Lenticular clouden.wikipedia.org
44. EarthSkyearthsky.org
45. Nacreous Clouds - Atmospheric Opticsatoptics.co.uk
46. Nacreous - Cloud Appreciation Societycloudappreciationsociety.org
47. Polar stratospheric cloudsantarctica.gov.au
48. Polar stratospheric clouden.wikipedia.org
49. Winter Noctilucent Clouds Following Sudden Stratospheric Warming: First Observationsarxiv.org
50. Supercellen.wikipedia.org
51. Supercell Thunderstorms: thunderstorms with deep rotating updraftsww2010.atmos.uiuc.edu
52. Multicell and Supercell Thunderstormscourses.ems.psu.edu
53. What makes some thunderstorm cells 'super'? - Cardinal Newscardinalnews.org
54. Cosmic Rays and Climatearxiv.org
55. Thunderstorms, lightning, sprites and magnetospheric whistler-mode radio wavesarxiv.org
56. Five Miles of Electricity in a Flash: The Physics of Lightning Explainedsciencedigest.org
57. What Is Lightning? How It Forms and Strikeslightningtracker.app
58. Physics and technology of Laser Lightning Controlarxiv.org
59. Supercell low-level mesocyclones: Origins of inflow and vorticityarxiv.org
60. ACP - Sensitivities of cloud radiative effects to large-scale meteorology and aerosols from global observationsacp.copernicus.org
61. Longwave radiative effect of the cloud twilight zonenature.com
62. 1 Response of shortwave cloud radiative effect to greenhouse gases andacp.copernicus.org
63. Satellite observed trends of global mean net atmospheric shortwave and longwave irradiances and diabatic heating by precipitationscience.org
64. Clouds and Radiation - NASA Scienceearthobservatory.nasa.gov
65. ACP - Radiative effect of thin cirrus clouds in the extratropical lowermost stratosphere and tropopause regionacp.copernicus.org
66. Global warming in figures and the question of its treatment: some historical and epistemological viewsarxiv.org
67. Clouds, Precipitation, and Climate Changescied.ucar.edu
68. Vaisala Lidar Ceilometer CL61 with Depolarization Measurementefmlab.nd.edu
69. Ceilometer CL31 and CL51 for Meteorologyvaisala.com
70. Weather and Climate Physics Branch: Laser Ceilometer: NOAA Physical Sciences Laboratorypsl.noaa.gov
71. Cloud boundary height measurements using lidar and radararxiv.org
72. ceilo [Meteo Observations at IGMK]obs.meteo.uni-koeln.de
73. Cloud Top Height Determination From 10.7 um Imageryrammb.cira.colostate.edu
74. Infrared Satellite Imagerycourses.ems.psu.edu
75. Satellite Cloud Imagery: Infrared, Water Vapor & Cloud-Top Temperaturedropcast.io
76. 1 CLOUD TOP PROPERTIES AND CLOUD PHASE ALGORITHM THEORETICAL BASIS DOCUMENTmodis-images.gsfc.nasa.gov
77. Development of a Cloud-Top Height Estimation Method by Geostationary Satellite Split-Window Measurements Trained with CloudSat Data in: Journal of Applied Meteorology and Climatology Volume 49 Issue 9 (2010)journals.ametsoc.org
78. Using Synthetic Brightness Temperatures to Address Uncertainties in Cloud-Top-Height Verification in: Journal of Applied Meteorology and Climatology Volume 56 Issue 2 (2017)journals.ametsoc.org
79. Cloud radarmmwave.wp.st-andrews.ac.uk
80. millimeter wavelength cloud radar (MMCR)arm.gov
81. An Unattended Cloud-Profiling Radar for Use in Climate ...psl.noaa.gov
82. Cloud physics studies with 8 mm wavelength radar - ScienceDirectsciencedirect.com
83. Rain‐rate estimation algorithm using signal attenuation of Ka‐band cloud radar - Oh - 2020 - Meteorological Applications - Wiley Online Libraryrmets.onlinelibrary.wiley.com
84. Cloud Radar - Radartutorialradartutorial.eu
85. Parameterizationebsco.com
86. A perspective on climate sensitivity - ScienceDirectsciencedirect.com
87. publikationen.bibliothek.kit.edupublikationen.bibliothek.kit.edu
88. Clouds study finds that low climate sensitivity is ‘extremely unlikely’ - Carbon Briefcarbonbrief.org
89. ACP - The presence of clouds lowers climate sensitivity in the MPI-ESM1.2 climate modelacp.copernicus.org
90. A review on regional convection‐permitting climate modeling: Demonstrations, prospects, and challengesncbi.nlm.nih.gov
91. An optimized parameterization of sub-grid scale advection for convection permitting modelspnnl.gov
92. Tropical High Cloud Feedback Relationships to Climate Sensitivityjournals.ametsoc.org
93. Recent Cloud Controlling Factor Analyses Indicate Higher Climate Sensitivity - Wilson Kemsley - 2026 - Geophysical Research Letters - Wiley Online Libraryagupubs.onlinelibrary.wiley.com
94. Invisible ship tracks show large cloud sensitivity to aerosolnature.com
95. Simulating cloud-aerosol interactions made by ship emissionsarxiv.org
96. Ship tracks show how aerosols affect clouds fast and slowimperial.ac.uk
97. Weather Modification and Local Climate Management in the United States: A Review of Its Technological Evolution, Operations, Governance, and Local Implementation Challengesmdpi.com
98. Unexpected Effects of Cloud Seeding with Silver Iodideresearchgate.net
99. Review of Persistence Effects of Silver Iodide Cloud Seedingjournalofweathermodification.org
100. Does cloud seeding really work?cen.acs.org
101. Sodium Iodide versus Silver Iodide for Cloud Seeding: A Comparative Look - The Environmental Blogtheenvironmentalblog.org
102. Making it Snow: A Brief History and Review of the Science Behind Cloud-Seeding - DRIdri.edu
103. Here’s what we know about the climate cost of white trails aircraft leave in the skytheconversation.com
104. Information on Contrails from Aircraftepa.gov
105. Most long-lived contrails form within cirrus clouds with uncertain climate impactncbi.nlm.nih.gov
106. Reduce Airplane Contrailsdrawdown.org
107. Are contrails harmful to the environment?nationalacademies.org
108. Small altitude changes could cut climate impact of aircraft by up to 59%sciencedaily.com
109. Detectable ship tracks account for just 5% of aerosol indirect forcing from ship emissions - PMCpmc.ncbi.nlm.nih.gov
110. ACP - Global aviation contrail climate effects from 2019 to 2021acp.copernicus.org
111. Aircraft propulsion systemimage-ppubs.uspto.gov
112. A Simple Condensation Model for the H2SO4-H2O Gas-cloud System on Venusarxiv.org
113. Venus's atmosphereebsco.com
114. Mars carbon dioxide ice clouden.wikipedia.org
115. Water Ice Clouds over the Northern Plainsjpl.nasa.gov
116. Curiosity Spots Carbon Dioxide-Ice Clouds in Martian Skiessci.news
117. ESA Science & Technology - Carbon dioxide ice cloud on Marssci.esa.int
118. NASA’s Curiosity Rover Captures Colorful Clouds Drifting Over Marsjpl.nasa.gov
119. Why Titan? - Dragonfly - JHUAPLdragonfly.jhuapl.edu
120. Frigid Titan has soaring, bubbling clouds with unusual rainsearthsky.org
121. Webb Observations Reveal Active Methane Cycle and Atmospheric Dynamics on Titanspacedaily.com
122. Venus Life Finder Mission Studyarxiv.org
123. ESA - Ice clouds put Mars in the shadeesa.int
124. The Man Who Named The Clouds - Science Museum Blogblog.sciencemuseum.org.uk
125. Charles Thomson Rees Wilson sees particle trackstimeline.web.cern.ch
126. Cloud chamberen.wikipedia.org
127. TIROS 1 Becomes the First Experimental Weather Reconnaissance Satelliteebsco.com
128. Celebrating the World’s First Meteorological Satellite: TIROS-1nesdis.noaa.gov
129. First Photo of Earth From a Weather Satellite, TIROS-1 - NASAnasa.gov
130. Celebrating 60 Years of the World’s First Weather Satellitenesdis.noaa.gov
131. TIROS-1: The World's First Weather Satellitemonmouthtimeline.org
132. Guide to the Luke Howard Collection 1865 - UChicago Librarylib.uchicago.edu
133. Luke Howard (meteorologist)en.wikipedia.org
134. Cirrus, cumulus and stratus clouds - Luke Howard (1772-1864) — Google Arts & Cultureartsandculture.google.com
135. Cavendish Laboratory : C.T.R. Wilson's cloud chamber of 1911cudl.lib.cam.ac.uk
136. Cloud Chamber - an overviewsciencedirect.com
137. Cloud Chambericts.res.in
138. Snow Crystalsarxiv.org
139. Overview of biological ice nucleating particles in the atmosphere - ScienceDirectsciencedirect.com
140. Geographic, seasonal, and precipitation chemistry influence on the abundance and activity of biological ice nucleators in rain and snowpnas.org
141. Toward Understanding Bacterial Ice Nucleationpubs.acs.org
142. Forming ice: There's a fungal protein for thatnsf.gov
143. AMT - Analyzing the chemical composition, morphology, and size of ice-nucleating particles by coupling a scanning electron microscope to an offline diffusion chamberamt.copernicus.org
144. Ice Nuclei - an overviewsciencedirect.com
145. observational constraints low cloud feedback reduce uncertainty climate sensitivityeesm.science.energy.gov
146. Model spread in tropical low cloud feedback tied to overturning circulation response to warmingnature.com
147. Multi-objective observational constraint of tropical Atlantic and Pacific low-cloud variability narrows uncertainty in cloud feedbacknature.com