Why do cumulus clouds have well defined boundaries? In other words, what are the physical mechanisms that hold a cloud together, as an entity separate from other clouds, that prevent it from spreading, etc.

Naively, one could expect the atmospheric vapour to spread homogeneously or (in presence of nonlinearities inherent in hydrodynamics) form periodic structures or even vortices - these are indeed observed, but does not seem to explain the cumulus clouds. Dust clouds in outer space, held together by gravity seem closer phenomenologically, but physically gravity seems as a less plausible explanation than fluid/gas dynamics

  • 4
    $\begingroup$ as you are interested in clouds have you seen this site ? weather.gov/source/zhu/ZHU_Training_Page/clouds/… $\endgroup$
    – anna v
    Jun 19 at 8:32
  • 3
    $\begingroup$ that prevent it from spreading having seen and even made quite a few cloud time lapses in my time, I question where this is really the case. Over longer time scales, depending on atmospheric condition clouds can be quite fluid, and in some cases "evaporate" off the top or sides even as more moisture "condenses" as a replacement. $\endgroup$
    – Michael
    Jun 19 at 22:45
  • 2
    $\begingroup$ A related question: Wu do clouds form in lumps? $\endgroup$ Jun 20 at 9:05
  • 1
    $\begingroup$ @Michael +1 What I mean, in more technical sense, is that the clouds are compact entities with larges spaces between them; and they remain stable over long periods of times (hours or days). It is understood that they evolve, and that the thickness of their boundaries is rather large, compared, e.g., to a size of human. The answers also make important distinction between the actual and observed boundaries - but both are real. $\endgroup$ Jun 21 at 9:25
  • $\begingroup$ Is this the same as asking, "Why do clouds exist?"? $\endgroup$
    – Dan
    Jun 22 at 10:56

4 Answers 4


Clouds are fuzzier than they look.

Clouds get their white colour from Mie scattering of light from water droplets of size comparable to the wavelength of light. But for smaller droplets Rayleigh scattering is the best approximation. The formula for the intensity of that radiation seen by an observer at distance $R$, scattering angle $\theta$, wavelength $\lambda$ from a particle of refractive index $n$ with diameter $d$ is $$ I=I_0 \left(\frac{1+\cos^2\theta}{2R^2}\right)\left(\frac{2\pi}{\lambda}\right)^4\left(\frac{n^2-1}{n^2+2}\right)^2\left(\frac{d}{2}\right)^6 $$

Note the last term: the scattering intensity increases with the sixth power of the diameter. That means that as the vapour density increases in the air and droplets start forming, even a completely smooth gradient of droplet sizes will look like it has a sharp edge where the scattering goes from minuscule to dominant.

Once droplets are $\sim 10$% of the light wavelength the full Mie theory is needed, but the effect is roughly the same (and less wavelength dependent).

There are doubtless other forces keeping cumulus clouds sharp, like the upper boundary often corresponding to the top of an upwelling convective flow into drier air and hence having a strong vapour gradient, and the cloud base being set by temperature, pressure and the dew point.

  • 3
  • 4
    $\begingroup$ "for smaller droplets Rayleigh scattering is the best approximation" — this is misleading. Such claims should always note that Mie solution is the universal solution for scattering of waves by spherical particles, regardless of particle size. It works fine (with full precision) in Rayleigh regime, as well as for liquid-water clouds and raindrops. The only reason why Rayleigh description is preferred in its domain is that it's much simpler and works well enough (even though it's not "exact") in case of very small scattering particles such as air molecules. $\endgroup$
    – Ruslan
    Jun 20 at 8:59
  • 2
    $\begingroup$ @Ruslan - Note that this is the meaning of the previous sentence. Given that the question wasn't about the finer points of physical approximation theory, I think this update would be overly pedantic and only confuse the reader. $\endgroup$ Jun 20 at 18:23
  • 3
    $\begingroup$ This argument isn't completely clear. You refer to a "smooth gradient of droplet sizes". However, the general situation is that in any given small (mesoscopic) volume, there is a distribution of droplet sizes, and this distribution can expected to vary smoothly over space. You seem to imply that the distribution is narrow (sharply peaked at a particular size), and varies over space mainly by shifting the "location" of its peak (causing the distribution itself to rise and fall rapidly) -- in which case, why is that? ... $\endgroup$
    – nanoman
    Jun 20 at 18:26
  • 4
    $\begingroup$ ... Since total incoherent scattering from droplets of any given size is directly proportional to their density, the brightness of a small volume should be your Rayleigh scattering formula convolved with the droplet size distribution. Now, your observation about the $d^6$ factor implies that this is dominated by the largest droplets. But this just reduces the question to: How does the density of the largest droplets vary over space? And I'm not clear how you have reached the conclusion that this variation is sudden. $\endgroup$
    – nanoman
    Jun 20 at 18:27

I just want to expand on Anders’s otherwise great answer, because I think based on the text of the question this paragraph is more what you're asking for:

There are doubtless other forces keeping cumulus clouds sharp, like the upper boundary often corresponding to the top of an upwelling convective flow into drier air and hence having a strong vapour gradient, and the cloud base being set by temperature, pressure and the dew point.

What the layperson needs to understand about a cloud, is that it is not like a rock with a defined set of grains/crystals in a fixed configuration. It is more like a river or a waterfall. Stuff is coming in one end and going out the other.

At the bottom you have moist air coming upwards, where the water inside is fully vaporized and so is transparent. But the pressure is dropping and so is the temperature, and at a particular virtual boundary which is the flat bottom of the cloud, this vaporized water condenses into little droplets. Those droplets are visible now.

All of the little droplets that you are seeing in the cloud are drifting, and on average they are drifting upwards.

Partly this is driven by outside factors, different masses of air moving across the landscape collide and one of them shifts upwards while the other one burrows underneath. But it is also self-sustaining. As a consequence of condensing, droplets release heat which warms the air which expands it. The warm air rises and drags the droplet with it upwards. Because the flows involved are way slower than the speed of sound, the air behaves approximately as an incompressible fluid. That just means that if there is this sudden updraft from the condensation warming the air, then air gets sucked in from the bottom to prevent a vacuum. (Incompressible also means not expandable, the vacuum can be interpreted as “stretching out the air” and at lower velocities air doesn't like to stretch.)

Eventually the droplet leaves the moist air that it is in. If it leaves out of the sides of the cloud, or the bottom, it may re-evaporate. On the other hand if it leaves out of the top of the cloud, which usually requires some pretty considerable convective wind transport, then it may get to the upper atmosphere where it can spontaneously freeze into “diamond dust”, tiny little flecks of supercooled ice. These are cirrus clouds, and they look a lot more fuzzy and flowy. A cumulus cloud that does this with those strong convective currents is a cumulonimbus cloud, a rain cloud. They have a characteristic anvil-like shape from the diamond dust top spreading out while the bottom is so firm and solid. (And the rain comes of course from the super-cooled ice crystals coming back down into the cloud, growing into big snowflakes, falling out of the cloud, and melting into raindrops.)

If you've been out on a foggy day then you have walked around in a cloud before, they do not have extremely sharp boundaries when you are up close in person. But, these very common cumulus clouds have this distinctive flat bottom and puffy top because those represent properties of the surrounding air where the droplets are appearing and disappearing. The bottom is flat because the dominant temperature and pressure gradient is vertical, the top is poofy with a sharp-looking boundary because you are looking at this boundary where this moist air mass from below collides with a dry air mass which it is pushing, roiling, bubbling into. The droplets at the one boundary are either turned away or evaporated into the dry air, the droplets at the other boundary are formed and are pulling more moist air up after them.

  • $\begingroup$ Thank you, I didn't make the distinction between actual boundaries and visible boundaries in the OP: @AndersSandberg addressed only the latter (I do agree that their answer is very good), so this answer is a necessary addition. I think one can drive through a wall of fog - so I'd say that in this case a cloud still has pretty clear boundaries (though their scale is bigger than the size of a human). $\endgroup$ Jun 21 at 7:12
  • $\begingroup$ THIS is the main reason. The liquid water really is mostly concentrated in these sharply bounded regions as can be seen in numerical simulations such as researchgate.net/publication/333090507/figure/fig1/… $\endgroup$ Jun 21 at 19:54

Convection in meteorology is often assumed to be more or less adiabatic, and convection is what creates cumulus clouds. You can imagine clouds as bubbles rising inside a water tank, they also have a quite well defined boundary. Warm humid air rises up, cools (reduction of pressure) and condensation begins.



Furthermore the wind in the upper atmosphere is less turbulent than wind close to the ground. The clouds are advected by the general motion of the wind. So the warm humid air bubble which rose up stays together. With strong shear winds clouds some times get thorn apart.

To add to the question, there are cumulus clouds that look fuzzy as well. For example the Pileus cloud phenomenon.


Edit: To add to your question about periodic structures. There are periodic clouds. They are created by gravity waves in the upper atmosphere.


New contributor
smichel is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.

Cumulus clouds are opaque (optically thick). The interior of a large cloud is very dark (as you've experienced if you've flown through one in an airliner in the daytime during ascent or descent). When you view a cloud from outside, the white scattered light you see comes from the outer layer of the cloud where sunlight can reach.

Starting from the outside, as the droplet density increases heading into the cloud, the scattered intensity becomes greater and greater, until so much light has been scattered out that the remaining inward-bound light is significantly attenuated. Around this point, the scattered intensity reaches a maximum and then quickly dies out. Even if the droplet density continues to increase toward the center of the cloud, this does not contribute to the cloud's brightness.

(This is also why sunlit cumulus clouds are very bright -- they scatter nearly all of the incident sunlight, similar to an ordinary bright-white diffusely reflecting object, like a sheet of paper.)

When viewing a cloud from outside, you see mainly the "skin" at the depth of maximum scattering. Thus, you see a nearly uniform brightness on any line of sight that intersects this skin, despite potentially large variations in depth and density under the skin. As your line of sight starts to miss the skin, the brightness you see decreases along with the droplet density; if this occurs over a distance that is small compared to the size of the cloud, it will appear as a sharp boundary.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.