The Water Cycle

The water in the glass#

Fill a glass of water and drink it. Every molecule you just swallowed is older than the Sun's light that grew the food you ate this morning — older than the continents, older than life. Not one of those molecules was manufactured for you. Water is essentially never created and essentially never destroyed on Earth; the same finite pool has been in circulation for something like four billion years.

Which means the water in that glass has a history. Statistically, some of it has been steam venting from a volcano, ice a kilometre deep in an Antarctic glacier, blood in a dinosaur, fog over a Carboniferous swamp, and black water three kilometres down in the abyssal Pacific that had not touched the sky since before the pyramids. It passed through all of that, and it is passing through you now, and in a few days most of it will be back in the air.

This is the single most important thing to understand about the water cycle, and it is the thing most people have backwards. Water is not consumed. It is borrowed and moved. When we say a city "uses" a billion litres of water, none of that water is gone — it has been shifted from one place, and one form, to another. The whole subject is bookkeeping: tracking a fixed quantity of a conserved substance as it moves between a handful of storage tanks, driven around the loop by the Sun. Get the bookkeeping right and everything from why the atmosphere carries all the rain despite holding almost none of the water, to why a warming world floods and droughts more fiercely, falls straight out of it.

The reservoirs, and their startling proportions#

The cycle is best understood exactly like the carbon cycle: a box model of reservoirs (stocks, in km³) connected by fluxes (flows, in km³/yr). The first surprise is how lopsided the stocks are.

| Reservoir | Volume (km³) | Share of all water | | --- | --- | --- | | Oceans | ~1,338,000,000 | 96.5% | | Ice caps & glaciers | ~24,000,000 | 1.74% | | Groundwater | ~23,400,000 | 1.69% | | Lakes | ~176,000 | 0.013% | | Soil moisture | ~16,500 | 0.0012% | | Atmosphere | ~12,900 | 0.001% | | Rivers | ~2,120 | 0.0002% |

Two facts in that table overturn most people's mental picture.

First, the planet is not a water world with some salt in it — it is a salt-water world with a rounding error of fresh. About 97% of all water is in the oceans and undrinkable. Of the ~2.5% that is fresh, the overwhelming majority is not in any lake or river you could point to: roughly 69% of fresh water is frozen into ice caps and glaciers, and almost all of the rest — about 30% — is groundwater, much of it deep, ancient, and slow to recharge. Everything humanity actually drinks, farms with, and fights over — the rivers and accessible lakes and shallow soil moisture — is a fraction of a percent of a fraction of the whole. The intuition that "most of Earth's water is fresh and available" is wrong by orders of magnitude.

Second, the atmosphere — which delivers every drop of rain and snow that has ever fallen — holds essentially none of the water. About 0.001% at any instant. If you condensed the entire atmosphere's moisture onto the surface it would make a layer about 2.5 cm deep worldwide. That thimbleful is the reservoir the entire terrestrial water supply passes through. How a reservoir that small can be responsible for all the rain is the puzzle the rest of the article resolves — and the answer is not "the table is wrong."

Watching the loop turn#

The fluxes between these boxes are driven by one thing: sunlight. The Sun evaporates water from the ocean and transpires it from plants; the vapour rises, cools, and condenses into cloud; it falls as rain or snow; and gravity returns it to the sea as runoff and groundwater flow. A continuous, closed loop, powered from above.

The widget is that loop, drawn as choreography rather than a fluid solve — the fluxes and the way they respond to solar input are the honest part; the individual raindrops are illustrative. The Sun evaporates parcels off the ocean (cyan) and transpires them from the trees (green); they rise, condense into cloud (violet), fall as rain over the sea and land or as snow on the high ground (white), and the water that lands on the mountain either runs off to the sea (blue) or soaks in and later transpires again.

Things to try:

• Let it run at the default setting first. Notice that the traffic is overwhelmingly between the ocean and the sky. Most evaporation is oceanic and most rain falls straight back onto the ocean; the land is a comparatively small side-loop. Yet that side-loop — precipitation over land minus what evaporates back off it — is every river, aquifer, and glacier on the planet.
• Now drive the solar input up. This is the key experiment. As you turn up the Sun (and with it the temperature), evaporation and precipitation intensify together — more parcels lift off the ocean, the clouds thicken, and the rain and snow come down harder. Turn it down and the whole machine slows. The cycle has no engine of its own; it runs at exactly the rate the Sun drives it. A warmer atmosphere runs it faster and holds more vapour — the readout shows both the evaporation flux and the water held aloft climbing as the temperature rises.
• Watch a molecule's journey. Follow a single parcel: it does not fall back where it evaporated. It drifts a long way in the cloud before it precipitates — which is why rain over a continent can be ocean water that evaporated a thousand kilometres upwind. The idea that "rain falls near where the water evaporated" is another comfortable falsehood; water vapour is transported large distances before it comes down.

The mathematics: mass balance, residence time, and a colossal heat engine#

Everything the widget shows follows from one equation. For any reservoir holding a volume $V$ of water, conservation of mass says the stock changes at exactly the rate of inflows minus outflows:

$$\frac{dV}{dt} \;=\; \sum F_{\text{in}} \;-\; \sum F_{\text{out}}$$

That is the entire model. In long-term steady state the left side is zero, so each reservoir's inflows and outflows must balance — global evaporation equals global precipitation at about 486,000 km³/yr, and the water that rains on land in excess of what evaporates back (about 40,000 km³/yr) is precisely the runoff that rivers and groundwater return to the sea.

Residence time = reservoir ÷ flux#

Now the quantity that makes sense of the whole lopsided table. The residence time of a reservoir is its stock divided by the rate at which water flows through it:

$$\tau \;=\; \frac{V}{F}$$

It answers: how long, on average, does a water molecule stay here before moving on? Apply it to the atmosphere, with a stock of 12,900 km³ and a through-flux (global evaporation ≈ precipitation) of about 495,000 km³/yr:

$$\tau_{\text{atm}} \;=\; \frac{12{,}900\ \text{km}^3}{495{,}000\ \text{km}^3\,\text{yr}^{-1}} \;\approx\; 0.026\ \text{yr} \;\approx\; 9\ \text{days}$$

A water molecule spends about nine days in the air. That single number dissolves the paradox of the tiny reservoir that carries all the rain. The atmosphere holds almost no water — but it turns that stock over about 40 times a year. A small tank flushed forty times delivers a large annual flux. It is exactly the bank-account logic of the carbon cycle: throughput, not standing balance, is what moves water around the planet.

Run the same division for the deep ocean — a stock of ~1.34 billion km³ against an evaporative flux of ~419,000 km³/yr — and you get

$$\tau_{\text{ocean}} \;=\; \frac{1.338 \times 10^{9}\ \text{km}^3}{4.19 \times 10^{5}\ \text{km}^3\,\text{yr}^{-1}} \;\approx\; 3{,}200\ \text{years.}$$

The bigger reservoir has the vastly longer residence time, because residence time is a ratio, not a size. This is the same distinction the carbon cycle turns on, and it is worth over-learning: a reservoir can be enormous and slow, or minuscule and fast, and it is the flux relative to the volume — not the volume itself — that decides which.

Latent heat: why the cycle is a planetary heat engine#

There is a second bookkeeping running alongside the water: energy. Evaporating water is not free. It takes a large amount of energy to break the hydrogen bonds holding liquid water together and send a molecule off as vapour — the latent heat of vaporisation, $L_v \approx 2.5 \times 10^{6}\ \mathrm{J/kg}$. That energy does not vanish; it is carried, hidden, inside the vapour, and released again the instant the vapour condenses back into cloud droplets kilometres up.

So every gram of water that evaporates is a tiny charged battery. The Sun charges it at the surface; the atmosphere carries it aloft; condensation discharges it high above the ground. Multiply by the global evaporation flux and the numbers are staggering. Evaporating 486,000 km³ of water a year consumes on the order of

$$P \;=\; L_v \,\frac{dm}{dt} \;\approx\; \left(2.5\times10^{6}\ \tfrac{\text{J}}{\text{kg}}\right)\left(4.9\times10^{17}\ \tfrac{\text{kg}}{\text{yr}}\right) \;\approx\; 4\times10^{16}\ \text{W} \;=\; 40\ \text{PW,}$$

roughly half of all the sunlight the surface absorbs, and hundreds of times humanity's entire power consumption. The water cycle is not just a plumbing diagram; it is the largest heat engine on the planet, and the latent heat it moves is the fuel that builds thunderstorms and drives atmospheric convection. The rising, condensing branch of the widget above and the rising, condensing parcel in the convection article are the same physics seen from two sides: the water cycle supplies the moisture, and its latent heat is what makes a fair-weather cloud tower into a storm.

Reservoirs and residence times, side by side#

The atmosphere's nine days and the deep ocean's three thousand years are only two points in a range that spans five orders of magnitude. The second widget lays out every reservoir at once — sized to its true proportion, and placed on a residence-time axis by the same $\tau = V/F$ division.

What to try:

• Read the to-scale bar along the top. The ocean is the whole bar; ice and groundwater are thin bands; and everything you can actually drink — lakes, rivers, soil, and the atmosphere combined — is narrower than the marker line. Seeing the fresh-water sliver at true scale is the fastest cure for the "most water is fresh" misconception.
• Select each reservoir and read off $\tau = V/F$. The card shows the division explicitly. The atmosphere turns over in about 9 days; rivers in a couple of weeks; soil moisture in a couple of months; lakes in decades; groundwater and ice in thousands of years; the deep ocean in millennia. The residence-time axis at the bottom stretches from days to a million years — an almost unimaginable range for the same substance.
• Watch the turnover rings. Each reservoir carries a small orbiting marker spinning at a rate set by $1/\tau$. The tiny atmospheric reservoir whirs around; the rivers tick along; and the giant ocean, ice, and groundwater reservoirs sit visibly frozen in place. That contrast is the whole lesson made physical: the smallest reservoir is the fast, restless one that runs the weather, while the reservoirs that hold nearly all the water barely move on any human timescale.

That last point has a hard practical edge. Groundwater and ice have residence times of millennia because their fluxes are small. When we pump a deep aquifer faster than its trickle of natural recharge, or melt an ice cap, we are drawing down a reservoir that refills on a thousand-year clock. It is conserved water — it is not destroyed — but on any timescale that matters to us, it is effectively fossil water, and spending it is closer to mining than to using a renewable flow.

Why this matters: a warming world runs the cycle harder#

The water cycle is not an abstraction; it is the supply chain for every drop of fresh water on Earth, and climate change is reaching straight into the physics we just derived.

Return to the latent-heat argument, but run it through the atmosphere's carrying capacity. The amount of water vapour air can hold before saturating is governed by the Clausius–Clapeyron relation, and over the range of Earth's surface temperatures it rises steeply — about 7% more water vapour per degree Celsius of warming. A warmer atmosphere is a thirstier and wetter one. This is the most robust prediction in all of climate science because it is thermodynamics, not modelling: warm the air and it holds, transports, and precipitates more water.

The consequence is an intensification of the whole cycle — exactly the behaviour you drove by hand with the solar slider. Evaporation strengthens, the atmospheric moisture load grows, and precipitation comes in heavier bursts. The rule of thumb hydrologists use is "wet gets wetter, dry gets drier": regions and seasons of convergence see more extreme rainfall and flooding, while regions of subsidence and evaporation see deeper drought — the dry-descending and wet-rising branches of the same convective circulation that sets the planet's climate zones, now turned up. The total water is unchanged, as it always is. What changes is how violently it is moved and where it is delivered.

And because the reservoirs that buffer us — glaciers, snowpack, deep aquifers — are the slow ones, the disruption is asymmetric. A snowpack that melts earlier each spring, or a glacier that shrinks, removes a natural reservoir that used to release its water gradually through the dry season, replacing a smoothed supply with a flashier one. The ocean, the dominant reservoir and the planet's thermal and hydrological flywheel, absorbs and redistributes the changes on its own millennial schedule, long after the forcing that caused them.

The water in your glass, then, is a loan from a fixed and ancient pool, routed to you through a nine-day reservoir riding on top of a three-thousand-year one, and moved the whole way by the Sun. We can neither make more of it nor destroy any of it. All we can do — and increasingly are doing — is change how fast, how fiercely, and how unevenly it moves.