Deserts sit where the air comes down dry
The world's great hot deserts are not scattered at random. Most of them sit in two belts on either side of the tropics, reaching out to about 30 degrees north and south of the equator. They are there because of where the air goes, not because of how hot the ground gets. Air rises and rains near the equator, flows toward the poles high up, and comes down again over those two belts. Air that is sinking cannot make rain, and this air lost almost all of its water on the way up.
Rain needs rising air
Air that rises meets lower pressure, so it expands and cools. Cool air can hold much less water vapour than warm air. Roland Stull's open textbook Practical Meteorology works the numbers: at 30 °C, the vapour in saturated air pushes with about 4.36 kilopascals of pressure, and at 0 °C with only about 0.61. By the same formula, each drop of 10 °C roughly halves the amount of water air can hold. So when air rises and cools far enough, the extra water has to come out as cloud droplets, and some of it falls as rain or snow.
The same rule works backwards. Air that sinks is squeezed by rising pressure and warms. Every degree it warms raises the amount of water it could hold, while the amount it actually carries stays the same. Clouds that drift into sinking air evaporate. A place where the air is sinking most of the time is a place that seldom gets rain, however close it is to the sea.
The loop that sets the belts
Sunlight heats the tropics more than anywhere else. Warm, moist air near the equator rises in towering thunderstorms, which drop heavy rain in a belt called the Intertropical Convergence Zone. Stull notes that this convection is so strong that it pushes the top of the lower atmosphere, the tropopause, up to 15–18 km in the tropics.
At the top the air has to go somewhere, so it spreads toward the poles. As it goes, Earth's spin turns it more and more toward the east, until near 30° latitude it is running as a fast west-to-east wind called the subtropical jet. Stull explains that the spin keeps this high air from getting much farther poleward than about 30°. It cannot rise into the stable stratosphere above, and it cannot push back toward the equator against the air still arriving. So it piles up and sinks. Where it reaches the ground it forms belts of high pressure, the subtropical highs. Near the ground the air then flows back toward the equator as the trade winds, and the loop closes. This whole loop is called the Hadley cell.
Stull gives its speeds: about 6 mm per second up in the rising core, and about 4 mm per second down in the centres of the broad sinking zone. Four millimetres a second is about a third of a kilometre a day. The loop shifts with the seasons, too. The bigger of the two cells crosses the equator, with its rising branch in the summer hemisphere and its sinking branch in the winter one. Stull places the rain belt at about 9° N in August and September and about 6° S in February and March.
Interactive One parcel of air on its trip around the Hadley circulation, from the rainy equator to the dry belt near 25° latitude. Set where the air was last saturated on the way up, which is where it dropped the last of the water it could not hold, and how far it has sunk on the way down. The figure works out how much water the air kept, how much it could hold where it is now, and so its relative humidity. Press Let it sink to watch the humidity fall as the air comes down. Every control works with Tab, Enter and the arrow keys.
Relative humidity 4%
Why sinking air is so dry
Atmospheric scientists have a tidy way to put this, which Sherwood and colleagues set out in a 2010 review. Follow a parcel of air back along its path until you reach the last place where it was saturated, the coldest point of its trip. Since then, if it has not mixed with other air, it has carried exactly the water vapour that air could hold there and no more. Its relative humidity now is just the ratio of what it could hold then to what it could hold now (with a small correction for the change in pressure).
Because the amount air can hold rises so steeply with temperature, this ratio gets small fast. Sherwood's team give an example: air at −13 °C (260 K) whose last saturation was at about −48 °C (225 K) has a relative humidity of only about 5%. Set the figure above to 11.5 km and 6 km, which are close to those two temperatures in its atmosphere, and it gives the same. Air that came out of the top of a tropical thunderstorm was last saturated very high and very cold. Bring it down to the middle of the atmosphere and it is bone dry, and the lower it sinks, the warmer it gets and the lower its relative humidity falls.
Where exactly the driest subtropical air was last saturated is still argued. Sherwood's review describes a 2005 study by Galewsky, Sobel and Held, which found that in northern winter most of the air arriving at the driest spot had last been dried out in the middle latitudes, carried in by large weather systems, rather than at the top of the tropical storms. Another study, by Dessler and Minschwaner, found that most of it was dried out high in the tropics. Either way, the air reaching the subtropics was last saturated somewhere far colder than where it ends up. Sinking air also leaves a layer of warmer air sitting on top of cooler air near the ground. Stephen Leroy and Sara Vannah describe the strong "capping inversions" over the sinking branches of the Hadley cells as what holds back the deep storms that would make rain.
Not every desert is a sinking-air desert
The US Geological Survey's booklet Deserts: Geology and Resources, by Alta Walker, sorts deserts by what makes them dry. The sinking-air deserts it calls trade wind deserts, and it counts the Sahara, the world's largest desert, among them. There are others:
Rain shadow deserts lie behind tall mountain ranges. Air forced up over the mountains drops its water on the near side, then comes down the far side dry. The booklet's example is the Turpan Depression, in the rain shadow of the Tian Shan in China. Coastal deserts sit on the western edges of continents near the tropics, beside cold ocean currents that run along the coast. Fog from the cold upwelling water often blankets them in winter, but rain is rare. The Atacama in South America is the driest of all: the booklet says measurable rain, a millimetre or more, may fall there as rarely as once every 5 to 20 years. Midlatitude deserts, such as the Sonoran Desert of southwestern North America, lie between 30° and 50°, in basins in the interior far from the sea. Monsoon deserts, such as the Rajasthan and Thar deserts of India and Pakistan, lie west of the Aravalli Range, where the summer monsoon has already lost its moisture on the range's eastern slopes. And polar deserts, like the Dry Valleys of Antarctica, get less than 250 mm of precipitation a year.
Is the dry belt moving?
Climate models predict that as Earth warms, the tropics widen, which would push the edges of the dry belts toward the poles. Measuring that turns out to be hard, because the edge of the Hadley cell cannot be seen directly. Scientists track it through stand-ins: where the surface winds change from easterly to westerly, where the jet stream sits, where the tropopause drops. These measures can disagree, and natural swings like El Niño move them from year to year. A 2025 review led by Isla R. Simpson concluded that, with newer data and better measures based on surface winds, the widening seen so far now matches what models give for the same period. In 2025 Annika Reiter, Julia Danzer and Andrea K. Steiner found that over 15-year windows the measures they tested gave trends that could point either way, and over 25 years they were still mostly inconclusive. Only over 35 years did most of them clearly show a poleward shift.
In 2026 Leroy and Vannah used the latitude where the average surface wind over the oceans changes from easterly to westerly as a marker of the Hadley cells' edges. They found the tropics had widened by 2.16 ± 0.50 degrees of latitude from 1995 to 2024, most of it in the Northern Hemisphere. Earlier studies using other measures had not found a significant widening in the north, and the atmosphere models they tested reproduced the southern widening but struggled to give any widening in the north. They write that if the tropics are widening, the subtropical deserts should be expected to spread toward higher latitudes. So the dry belts are probably creeping poleward, but how fast, and in which hemisphere, is still argued.
In short
Rain needs air that rises and cools. Near the equator warm, moist air rises in storms and rains out its water. High up it flows toward the poles until Earth's spin stops it near 30°, and there it sinks. Sinking air warms, and since it was last saturated somewhere very cold, it carries only a tiny share of the water it could hold, so clouds cannot form. That is why most of the great hot deserts sit in belts reaching out to about 30° from the equator. Mountains, cold coastal currents and distance from the sea make the other kinds. Measurements suggest the dry belts are widening toward the poles as the climate warms, though by how much is still being worked out.
Where this comes from
- Practical Meteorology: An Algebra-based Survey of Atmospheric Science, ch. 11, General Circulation (v1.02) linked only, not reproduced
www.eoas.ubc.ca/books/Practical_Meteorology/prmet102/Ch11-global-v102.pdf - Practical Meteorology: An Algebra-based Survey of Atmospheric Science, ch. 4, Water Vapor (v1.02b) linked only, not reproduced
www.eoas.ubc.ca/books/Practical_Meteorology/prmet102/Ch04-watervapor-v102b.pdf - Tropospheric water vapor, convection, and climate (Reviews of Geophysics 48, RG2001) linked only, not reproduced
doi.org/10.1029/2009RG000301 - Deserts: Geology and Resources, Types of Deserts linked only, not reproduced
pubs.usgs.gov/gip/deserts/types/ - Widening of the tropics in global surface-air winds (Scientific Reports) linked only, not reproduced
doi.org/10.1038/s41598-026-43234-z - The potential of GNSS radio occultation data for the analysis of the tropical width: a comparison with reanalyses (Weather and Climate Dynamics 6) linked only, not reproduced
doi.org/10.5194/wcd-6-1683-2025 - Confronting Earth System Model trends with observations (Science Advances) linked only, not reproduced
pmc.ncbi.nlm.nih.gov/articles/PMC11900882/