Why weather happens

Stand outside on a warm afternoon and watch one cloud for five minutes. Not a time-lapse on a screen. The actual thing, with your actual eyes.

It grows upwards. The top boils over itself like milk climbing a pan, and if the afternoon is right it keeps climbing until it hits something invisible, flattens, and spreads sideways into a great lopsided anvil. Somewhere under that anvil it's raining hard on somebody who didn't bring a coat.

That's most of weather in a single object. Air went up.

Nearly everything else — the rain, the wind, the arrows on the forecast map, the reason the washing dried — falls out of air rising in one place and sinking in another. It isn't random. It never was.

All of it happens in a very thin layer

Begin with how shallow this is, because almost everyone pictures the atmosphere as deeper than it is.

Air doesn't end at a line. It thins out until there's effectively nothing left, and the fading takes hundreds of kilometres. But the weather doesn't use most of that. Cloud, rain, wind, storms — practically all of it is stuck in the bottom layer, the troposphere, which is roughly ten kilometres deep over Britain and sixteen or so over the tropics. Set that against a planet 12,700 kilometres across and you've got a film thinner than the skin on an apple. Half the mass of the whole atmosphere sits below five and a half kilometres, which is lower than plenty of mountains.

The lid on it is real, and that flattened anvil is you looking straight at it. Through the troposphere, air gets colder as you climb — about 6.5°C for every kilometre on average. Then it stops. Above the tropopause the temperature holds steady and then starts rising again, and warm air sitting on top of cold air is a stable arrangement that nothing much wants to punch through. A thunderstorm's updraught charges upward for miles, meets that, and has nowhere left to go but sideways.

Which is why storm clouds have flat tops. They ran out of sky.

The engine is the Sun, and it heats unevenly

All of this is solar-powered, but the power isn't the interesting part. The unevenness is.

Sunlight hits the equator nearly head-on and hits the poles at a slant, spreading the same energy over a much larger patch of ground and forcing it through more atmosphere on the way in. Snow and ice then throw most of what's left straight back out. The result is a permanent, enormous heat imbalance between the middle of the planet and the ends of it, and the atmosphere spends its entire existence failing to fix that imbalance while trying very hard. Weather is the failure. Every storm, every wind, every front is heat being shifted poleward by an inefficient machine.

George Hadley worked out the first sensible version of this in 1735. Warm air rises at the equator, drifts poleward high up, cools, and sinks again — Hadley put the sinking too far north, but the shape of the idea survived. The real circulation breaks into three loops per hemisphere rather than one, and the middle loop is messy, but the tropical cell is very close to what he described and still carries his name.

The sinking half of it has left fingerprints all over the world map. Air that rose over the equator has already dumped its moisture as tropical rain. It comes back down around 30° north and south — dry, warm, compressed — and that descending band is where you find the Sahara, the Arabian deserts, the Kalahari, the Australian interior and the dry belt of northern Mexico. The great deserts aren't scattered at random. They sit in a ring, at a latitude, because that's where the air goes down.

Why air rises, and what happens when it does

Air goes up for four reasons and there aren't any others worth learning. The ground heats it from below and it becomes buoyant. A mountain gets in the way and shoves it up. A colder, denser air mass undercuts it. Or air converging into a low-pressure area has nowhere left to go but upward.

What happens next is the same in every case, and it's the single most useful piece of physics in this whole business. Rising air expands, because there's less pressure squeezing it the higher it goes. Expanding costs energy. The energy comes out of the air's own heat, so the parcel cools without losing a single joule to its surroundings — roughly 10°C per kilometre while it stays dry.

Cool it enough and it can't hold its water any more. Warm air carries far more vapour than cold air does — about seven per cent more for each extra degree — so cooling a moist parcel eventually brings it to the point where the vapour has to come out. It condenses onto tiny floating specks of dust, sea salt, pollen and soot, because pure vapour is surprisingly reluctant to condense onto nothing. Millions of droplets, each a hundredth of a millimetre across, hanging in the air.

That's a cloud. That's all a cloud is.

Then comes the bit that isn't obvious. Condensing releases heat — exactly the heat that was absorbed when the water evaporated off an ocean somewhere days earlier. So the moment cloud starts forming inside a rising parcel, the parcel gets a shot of warmth, which makes it more buoyant, which makes it rise faster, which makes more cloud. A cumulus cloud is running on the energy of water that evaporated somewhere else. Hurricanes are the same trick at a scale that's difficult to think about.

Rain is harder than it looks. A cloud droplet is about a millionth of the volume of a raindrop, and it can't fall — the air resistance on something that small beats gravity easily, which is why clouds sit there. Droplets have to merge, either by colliding and sticking on the way past each other, or by the route that dominates in colder clouds, where ice crystals steal vapour from the liquid droplets around them and grow at their expense until they're heavy enough to drop. Melt those on the way down and you get rain. Don't, and you get snow. Send them up and down a few times through a violent storm, adding a layer of ice each pass, and you get hail.

Highs, lows and the weight of the sky

Air weighs something. Roughly ten tonnes of it are pressing down on every square metre of ground, and the only reason you're not flattened is that the same pressure is pushing back out from inside you.

That weight isn't evenly spread, and the map of where it's heavy and where it's light is the map forecasters actually work from. Standard sea-level pressure is 1013 hectopascals. Anything much under 1000 counts as a decent low; a serious Atlantic storm can go below 950.

The link to your washing is direct. Where air is sinking, it warms and compresses as it descends, cloud evaporates away, and you get high pressure with clear settled skies. Where air is rising, cloud forms, and you get low pressure with wind and rain. A barometer isn't measuring weather. It's measuring whether the air above you is on its way up or on its way down, and everything else follows from that.

High pressure isn't automatically good news, mind. In winter a stubborn high can trap cold air and pollution under a temperature inversion for a fortnight, which is grey, still, freezing and grim.

Why the wind refuses to go straight

Wind is air sliding from high pressure to low pressure. On a planet that stood still it would go straight there, the pressure difference would even out, and that would be that.

The planet doesn't stand still. Anything moving freely across a rotating surface gets deflected — right in the northern hemisphere, left in the southern — and this is the Coriolis effect, named for Gaspard-Gustave de Coriolis, who published the maths in 1835 while thinking about waterwheels and other rotating machinery rather than about weather at all.

The consequence is that wind never gets to the low. It's bent sideways on the way, and ends up circling instead — anticlockwise around a northern-hemisphere low, clockwise around a high, with the southern hemisphere doing the reverse. That's why weather systems on a satellite image are spirals rather than sensible straight flows collapsing inward. It also gives you a free trick, first written down by the Dutch meteorologist Buys Ballot in 1857: stand with your back to the wind in the northern hemisphere and low pressure is off to your left. Sailors were using that generations before anyone could explain why it worked.

The effect can't do your bath. It's far too feeble at that scale to beat the shape of the basin and whatever swirl the water already had. It runs systems hundreds of kilometres wide. It doesn't run the sink, and anyone who's demonstrated otherwise on holiday was doing a magic trick.

Air masses, and the fronts where they collide

Leave a large body of air sitting over one place long enough and it takes on the character of what's underneath. Parked over the Arctic it goes cold and dry. Parked over the tropical Atlantic it goes warm and wet. These are air masses, and they can be the size of a continent.

They don't blend readily. Warm and cold air of different densities meet along a sharp sloping boundary and stay stubbornly separate, and that boundary is a front — a word with a specific and slightly grim origin. The Norwegian physicist Vilhelm Bjerknes and the group around him at Bergen developed the modern picture of mid-latitude storms in 1919, in the immediate aftermath of a war, and borrowed the vocabulary of battle lines for the places where two air masses fought it out. The name stuck.

The sequence they described is still what happens over your head. A warm front arrives first: its boundary slopes gently, so the warm air slides up over the cold in a long shallow ramp, and you get high cirrus hours or a day ahead, thickening and lowering into flat grey sheets, then steady unremarkable rain that goes on for ages. Then a mild, damp, dull spell in the warm sector behind it. Then the cold front, which is a much steeper affair — dense cold air bulldozing underneath and forcing the warm air up sharply, giving a short violent burst of heavy rain, maybe thunder, a sudden wind shift, and then that clean rinsed feeling with towering cumulus scattered about behind it.

Cold fronts move faster than warm ones. Eventually the cold front catches the warm front up, lifts the warm sector off the ground entirely, and the storm winds down. That's an occlusion, and it's why a low-pressure system doesn't last forever.

The river in the sky

High above all of this, where cold polar air meets warmer air from the south, the sharp temperature contrast drives a narrow ribbon of very fast wind that snakes right around the planet from west to east. That's the jet stream, sitting at roughly the height an airliner cruises, commonly running at 150 to 250 kilometres per hour and capable of far more in winter.

It was found by a man almost nobody has heard of. In the 1920s the Japanese meteorologist Wasaburo Ooishi tracked pilot balloons from a station near Mount Fuji and documented a ferocious westerly current in the upper air. He published his results in Esperanto, which he believed would make them universally accessible, and the effect was the exact opposite — the work went largely unread outside Japan for two decades. American bomber crews rediscovered the thing the hard way in 1944, flying west and finding themselves nearly stationary over the ground.

For forecasting, the jet is the steering wheel. Storms travel along underneath it. When it's straight and quick, systems get shunted through and the weather changes every day or two. When it buckles into big lazy meanders, systems stall, and whatever you've got settles in for a week or more. A long dry spell and a long wet one are frequently the same phenomenon wearing different clothes: a jet stream that's stopped moving things along.

Thunderstorms and what lightning actually is

A thunderstorm is what you get when the updraught gets out of hand. You need warm moist air near the ground, something to lift it, and an atmosphere unstable enough that once a parcel starts rising it keeps going. Hot afternoons do it. So do cold fronts.

Inside the cloud, ice particles of different sizes are being flung around and colliding constantly, and those collisions separate electric charge — the small light ice crystals carried up top end up positive, the heavier soft hail sinking towards the base ends up negative. Nobody has the microphysics of that exchange completely nailed down, but the outcome is a cloud with its charge sorted top and bottom, and a growing voltage that dry air is only just able to hold back.

When it can't hold any longer, the charge equalises through a channel a few centimetres wide, and that channel heats to something like 30,000°C — five times the surface temperature of the Sun, for a few ten-thousandths of a second. Thunder is the sound of that air exploding outward and then collapsing back. Light gets to you effectively instantly and sound covers about a kilometre every three seconds, so counting the gap and dividing by three gives you the distance in kilometres. Five seconds to the mile, if you'd rather.

Lightning does strike the same place twice, and the saying is simply wrong. Tall conductive objects get hit repeatedly, on purpose, every year — which is exactly why they have conductors bolted to them.

Somebody had to name the clouds

Clouds went unnamed for most of human history. People described them, painted them, farmed by them, and had no shared vocabulary for them at all, which made comparing observations between two places almost pointless.

Luke Howard fixed that. He was a London pharmacist and an amateur of the weather, and in 1802 he read a paper to a small scientific society proposing Latin names based on shape: cirrus for the wispy ones, cumulus for the heaped ones, stratus for the layered ones, and nimbus for the ones producing rain. Combine and modify them and you can label anything up there. Goethe admired the scheme enough to write poems about it. Constable wrote Howard's terms on the backs of his cloud studies.

Two centuries on, the international classification is still Howard's, extended but not replaced, and it's genuinely useful to a person standing in a field. Cirrus that thickens and lowers through the afternoon is a warm front on its way, often a good twelve hours ahead of the rain. Small fair-weather cumulus that grow tall and hard-edged by midday mean the atmosphere is unstable and someone's getting a storm. A flat featureless grey sheet means nothing is rising anywhere and you can expect drizzle and boredom.

Why forecasts run out

Forecasts are very good a day or two ahead and distinctly woolly a week out. That gap isn't a lack of computing power, and buying more of it won't close the gap.

The idea of computing the weather at all belongs to Lewis Fry Richardson, who spent the First World War driving an ambulance and filling his spare hours with a hand-calculated forecast. It took him weeks, it covered six hours, and the answer was wildly wrong — his method predicted a pressure change that never occurred. He published it anyway in 1922, along with a daydream about a vast hall filled with 64,000 human calculators, each handling one square of the globe, conducted from a podium in the middle so that they collectively kept pace with the real weather. He'd got the method essentially right and was several decades ahead of any machine that could run it. The first successful numerical forecast came off the ENIAC computer in 1950.

The hard limit turned up in 1961, and it was an accident. Edward Lorenz was rerunning a simplified weather model and, to save time, restarted it partway through by typing in a number from a printout — 0.506 rather than the 0.506127 the machine was holding internally. He expected the same run back. What he got diverged slowly, then completely, until the two forecasts had nothing whatever to do with each other.

That's chaos, and it's structural rather than a solvable engineering problem. The atmosphere is so sensitive to where it starts that any imperfection in today's measurements grows into a large error about next week. We can't measure every parcel of air on the planet, so there's a horizon past which specific forecasts can't go, however good the model gets. Lorenz's own image for it, from a talk title in 1972, was a butterfly's wings.

Forecasters deal with this honestly, and the way they do it is buried in a number most people misread. The model gets run dozens of times with slightly jiggled starting conditions. If all the runs agree, confidence is high. If they scatter, it's genuinely uncertain. A 30 per cent chance of rain isn't the forecaster hedging — it means that in something like 30 per cent of those runs, it rained on you.

Weather is not climate, and the difference isn't pedantry

One distinction causes more confusion than everything else here put together, and it's worth getting straight.

Weather is the state of the atmosphere right now, in one place. This afternoon's wind. Tomorrow's rain. Climate is the statistics of that weather over a long period — the World Meteorological Organization works in thirty-year blocks, which is why the reference period gets updated once a decade rather than once a year.

The old line is that climate is what you expect and weather is what you get, and it does the job. A brutal cold snap tells you nothing about the climate, in the same way that one unusually long queue tells you nothing about a shop's average waiting time. You need the whole distribution, and you need years of it.

It works the other way too. A single hot week isn't evidence of anything by itself either, and treating it that way concedes the argument to anyone pointing at a blizzard. What changes with the climate is the shape of the distribution — where the middle sits, how fat the tails are, how often the extremes turn up. That's a statement about decades, and it's measured in decades, which is exactly why it's so easy to argue about from a window.

None of the rest of this needs instruments. Watch the cumulus pile up on a hot afternoon and you're watching buoyancy. Feel the wind swing round and drop as the rain clears and you've just had a cold front go through. Notice cirrus thickening from the west at breakfast and you know roughly what the evening holds. You won't beat the Met Office, and there's no reason you'd want to. But the sky stops being scenery.