A rose bought in a supermarket in London or Berlin in February was, two days before, on a plant in a greenhouse on the shore of a lake in the Kenyan Rift Valley, and it carries, in the terms this site uses, about ten litres of the lake's water in its stem. Kenya grows about a third of the roses sold in Europe, most of them on a few dozen farms around Lake Naivasha, and the flower is the product on this site with the least substance and the most travel: water, a little sugar, air freight and a week in a vase. This article is about where a cut flower's water comes from, why the industry moved to the equator, what the lake has lost and recovered, what the farms send back, and what the labels on the wrapper cover.

The flower

A rose plant under glass or plastic produces flowers continuously, every stem cut at a few weeks old, and it is grown in beds or in bags of substrate under drip irrigation, with the water carrying the fertiliser to the root, in a greenhouse that holds the humidity and keeps the pests out. A hectare of roses gives about a million stems a year and uses, in the Naivasha studies, about ten thousand cubic metres of water, which is a swimming pool a week, and the arithmetic divides to seven to thirteen litres a stem. Most of that water leaves the greenhouse as vapour from the plant, the green and blue shares of the footprint method in a place where nearly all of it is blue, because the greenhouse keeps the rain out and the drip line brings the lake in. The grey share, the water needed to dilute the fertiliser and the pesticide that leave the beds, is about a third of the total in the older farms and much less in the ones that recycle their drainage.

A bunch of twenty carries two hundred litres, a Valentine's Day dozen a hundred and twenty, and the day itself, on which Europe buys a large share of its year's roses, moves a few billion litres of Rift Valley water in a week.

Why the equator

The cut flower trade was, until the 1990s, Dutch: grown under glass in the Westland, heated with gas through the winter, lit with lamps in the short days, and sold through the auctions at Aalsmeer, which still set the world price. It moved to Kenya, Ethiopia, Colombia and Ecuador for the reasons every crop on this site moves: twelve hours of light every day of the year, a climate at altitude that needs no heating, land and water that cost a fraction of Holland's, and wages a tenth. A Kenyan rose, cut, chilled and flown to Amsterdam, carries less carbon than a Dutch one grown in February under lamps, by the studies that have compared them, because the flight is cheaper in carbon than the gas. The Dutch auctions still sell it; the Dutch greenhouses now grow the flowers that will not travel.

Naivasha was the place because it is a freshwater lake, one of the few in the Rift Valley, at 1,900 metres, two hours from an international airport with nightly flights to Europe, and because the land around it was ranches that could be bought. The first farms arrived in the 1980s. By 2010 there were about fifty, employing tens of thousands, and the town that had been a few thousand people was a few hundred thousand.

The lake

Lake Naivasha has no outlet on the surface; it is fed by two rivers from the Aberdare hills and the rain, and it loses water to evaporation and to seepage into the volcanic rock beneath, and its level has always swung by metres with the rains. In the 2000s it fell further than that. The farms drew from it directly and from boreholes in the shore aquifer that the lake feeds; the town drew for its people; the rivers that fed it were drawn down by the farms in the hills for the vegetables that go to the same supermarkets; and the rains failed in the middle of the decade. By 2009 the lake was at its lowest in a century, the shore had retreated a kilometre in places, the fishery had collapsed, the hippos were in the farms, and the flower industry, which the newspapers had found, was blamed for all of it. The studies that apportioned the loss put the farms at a share, the town and the upstream farms at shares, and the rain at most, which is the finding every lake on this site that has been blamed on one user has produced.

The rains of 2010 and after refilled it, and in 2020 it rose to its highest in decades and flooded the farms that had been built on the shore it had left. The farms, under the buyers' audits and the basin's new rules, have since moved to metered abstraction, drip on every bed, and the recycling of their drainage, and the lake's water level is a matter of the rain again.

What goes back

The water that leaves a flower farm is the drainage from the beds and the wash water from the packhouse, and it carries the fertiliser, the pesticide and the fungicide that a crop grown in a humid greenhouse for its appearance receives more of than almost any food. In the older farms it ran to the lake, and the lake's fish carried the residues; the studies of the 2000s found the pesticides in the lake sediments and the water at levels that the fish and the birds registered. The farms that supply the European chains have since built wetlands and lined lagoons to treat the drainage, closed loops that catch and reuse it, and integrated pest management that swaps the sprays for predatory mites, because the chains' auditors and the labels on the wrapper require it, and the residues in the lake have fallen. The farms that supply the Middle East and the domestic market, with no auditor, have not all followed.

The label

The fair trade and the ethical labels on a bunch of Kenyan roses cover the workers first, their wages, their protective clothing, the crèche and the premium that goes to the estate's community, and they cover water and pesticide in the standard's criteria on abstraction, drainage and the list of chemicals a farm may not use. What they do not cover is the flight, which is the rose's carbon, or the arithmetic of the lake, which no label on a single farm can. A certified rose is a rose from a farm that meters its water and treats its drainage, which is what the buyer can ask for, and the basin's level is a matter of how many such farms there are and how much rain fell, which is what the buyer cannot.

Ethiopia and the rest

Kenya's neighbour has followed it. Ethiopia's flower farms, on the lakes and the highlands south of Addis Ababa, grew from nothing in 2000 to the second largest in Africa in a decade, on the same model: greenhouses on a lake shore, drip from the lake, night flights to Amsterdam, and a government that leased the land and built the cold store at the airport. The lakes of the Rift, Ziway above all, have shown the Naivasha pattern with fewer studies to apportion it. Colombia and Ecuador, which supply the American market from the savanna around Bogotá and the valleys under Cotopaxi, use less lake and more well, and the aquifer under the Bogotá savanna is falling. Every one of these is a place where a farm can grow a flower on cheap water and cheap labour under a constant sun and get it to a rich city by morning, and the water arithmetic is the same in all of them: ten litres a stem, from a source that the buyer never sees and the auditor has only recently begun to ask about.

What it teaches

The rose is the product on this site that has the least in it and travels furthest: ten litres of Rift Valley water, a gram of sugar, a flight to Amsterdam and a week on a table, for a thing that is thrown away when it wilts. The lake it comes from fell in a dry decade and was blamed on the farms, which were part of the reason, and rose again in a wet one, and the farms now meter, drip and recycle because the buyers in Europe asked them to, which is the only mechanism this site has found that works from the shelf back to the source. The water stays in Kenya. The flower does not.

Ten litres a stem, two hundred in a bunch, and a lake in the Rift Valley that fell and rose with the rain.

Sources

1. Mekonnen, M.M., Hoekstra, A.Y. and Becht, R. (2012). Mitigating the water footprint of export cut flowers from the Lake Naivasha basin, Kenya. Water Resources Management 26. About 7 to 13 litres per stem; the basin's flower farms and their share of abstraction.
2. Kenya Flower Council and Kenya National Bureau of Statistics, cut flower exports: about 200,000 tonnes a year; Kenya about 38 percent of EU rose imports; Naivasha about 70 percent of Kenya's flower production.
3. Becht, R., Odada, E.O. and Higgins, S. (2006). Lake Naivasha: experience and lessons learned brief. International Lake Environment Committee. Lake level decline of the 2000s and abstraction by farms and the town.
4. Williams, A. (2007). Comparative study of cut roses for the British market produced in Kenya and the Netherlands. Cranfield University. Carbon footprint of Kenyan versus Dutch roses.
5. Fairtrade International, Flowers and Plants Standard; water and pesticide criteria on certified farms.
6. Photographs: opener: Orange cut roses in Bro Church graveyard 2 by W.carter (CC BY-SA) via Wikimedia Commons; inline: Lake Naivasha Shoreline by Kenenock (CC BY-SA) via Wikimedia Commons; inline: Naivasha ast 2008033 lrg by NASA (Public domain) via Wikimedia Commons.

Los Angeles is a city of four million people, and a region of eighteen, on a coastal plain that gets about 380 millimetres of rain a year and has a river that the city paved into a concrete channel after a flood in 1938 and that is dry for most of the year. It is on this site because it is the clearest case in the world of a city built on water from somewhere else: it took a river from a valley three hundred and seventy five kilometres away in 1913, a share of the Colorado in 1941, and the rivers of northern California in 1972, and every one of those sources is now shrinking, so that the city that has lived for a century on other people's water is planning, for the first time, to make its own.

This article is about how the water was taken and what it did to the places it came from, what the shrinking of the Colorado means for a city at the end of the pipe, how the city cut its use by a third while growing by a million, and what it means to recycle everything.

The Owens valley

In 1904 the superintendent of the city's water department, William Mulholland, and a former mayor, Fred Eaton, went to the Owens valley, a farming valley on the dry side of the Sierra Nevada two hundred miles to the north, and Eaton began buying land and water rights along the Owens river, telling the sellers he was buying for a federal irrigation project. He was buying for the city. By 1907 the city had the rights, the federal project was cancelled, and the aqueduct, three hundred and seventy five kilometres of channel, tunnel and pipe running by gravity from the valley to the San Fernando valley on the city's edge, was built in five years and opened in November 1913 with Mulholland's five word speech: there it is, take it. The land in the San Fernando valley, which the aqueduct's water made worth farming and then worth building on, had been bought in advance by a syndicate that included the men who had promoted the aqueduct, which is the part of the story that the film got right.

The Owens valley's farms went dry, its people fought, dynamited the aqueduct in the 1920s and lost, and by 1926 Owens Lake, a hundred square miles of shallow water at the valley's end, had been drained to a salt flat. The city bought the rest of the valley through the following decades and owns most of it still, which is why it is empty.

The dust

The salt flat that had been Owens Lake became, over the following seventy years, the largest single source of dust pollution in the United States. Its bed of fine alkaline sediment, laced with arsenic, lifted in the valley's winds into storms that closed the road, put the towns along it over every air quality limit, and reached the city that had made it. In 1997 the air quality district ordered the city to control the dust, and the city has since spent more than two billion dollars flooding parts of the lakebed in shallow ponds, planting salt grass on other parts, and spreading gravel on the rest, at a cost in water of about a hundred million cubic metres a year, which is more than the aqueduct now takes from the valley in a dry year. The city that drained a lake to grow now waters its bed to keep the dust down, which is the most expensive lesson in the history of American water and it was learned in the same place it was taught.

Mono Lake, further north, which the aqueduct's second stage began draining in 1941, was saved by a court in 1983 that ruled the state held the lake in trust for the public and the city's diversions had to stop at a level that kept it alive. The city lost a sixth of its aqueduct supply and the lake has been slowly rising since.

The Colorado

The region's second source is the Colorado, delivered through the regional wholesaler's aqueduct from Lake Havasu on the Arizona border, and it is the river that the Las Vegas and Phoenix articles on this site describe as allocated, in 1922, on the basis of a wet decade, to seven states and Mexico in quantities that add up to more than it has carried in any year since. The reservoirs that hold it, Mead and Powell, fell in the drought of 2000 to 2023 to a third of their capacity, the federal government has cut the states' allocations twice, and the negotiations over the rules that take effect in 2026 are about how much less each state will get and in what order. California's share is the largest and, under the law of the river, the last to be cut, which is why Arizona's farms went dry before Los Angeles's lawns, and why Arizona's negotiators are asking for a different order. A city at the end of the longest pipe is a city with the least control over what comes through it.

Using less

The city's use of water peaked in 1990 and has fallen by about a third since, while the population grew by a million. The fall came from the fixtures, the tariff, the drought campaigns of 2015 and 2022, and from the lawns: the city paid households to tear out turf, restricted sprinklers to two days a week and then, in 2022, to one, and the front gardens of the San Fernando valley are gravel and succulents where they were grass. A resident of the city now uses about four hundred litres a day, which is high by the standards of this site and half of what it was, and the region's per capita figure is among the lowest of any American city in a dry climate. The city has grown into its water by using less of it, which is the Melbourne and Cape Town pattern on a larger scale and over a longer time.

Making its own

The city's plan for the next century is to import less and make more. The largest sewage works on the coast, at Hyperion by the airport, treats about 850 million litres a day and has, since 1950, discharged it to the sea through a pipe eight kilometres out. The plan is to rebuild it, by 2035, to purify all of that to drinking standard, by the membranes, reverse osmosis and ultraviolet that the Orange County and Singapore articles on this site describe, and to put it into the aquifer under the San Fernando valley and the pipes that come out of it, which would give the city about a third of its supply from within its own borders for the first time since 1913. Alongside it, the city is capturing the stormwater that its concrete channels were built to send to the sea, in basins that let it soak into the aquifer, and cleaning the aquifer of the solvents the aerospace industry left in it. The Colorado will still come through the pipe. The city will need less of it.

The river in the channel

The Los Angeles river runs for eighty kilometres through the middle of the region in a concrete channel that the Army engineers built after the flood of 1938 drowned a hundred people, and for most of the year it carries a few inches of treated effluent from the works upstream and, in a storm, a torrent that reaches the sea in hours. The channel was built to get the rain out of the city as fast as possible, and it does, sending to the ocean each winter a volume of stormwater that the city's engineers now put at enough, if captured, for a tenth of the region's use. The plan to unpave parts of it, to slow the water and let it soak, has been argued since the 1980s and begun in a few reaches, and the basins and the wetlands the county has dug beside it are the stormwater capture that the recycling plan depends on. The river that the city buried to make room for itself is the source it is now trying to dig up.

What it teaches

Los Angeles is the city on this site that took its water furthest and paid for it longest. The valley it drained became a dust bowl it now waters, the lake it began to drain was saved by a court, the river it shares is shrinking under seven states, and the delta it pumps is protected for fish. The city's answer, a century after Mulholland, is to recycle what it flushes and catch what falls on it, which is the answer every dry city on this site reaches when the pipes from elsewhere run short. It will still be a city of four million in a semi desert, and it will still drink from the Colorado. It will simply, at last, drink some of its own.

Three hundred and seventy five kilometres of aqueduct, a lake turned to dust, and a sewage works being rebuilt to make a third of the city's water.

Sources

1. Los Angeles Department of Water and Power (LADWP), Urban Water Management Plan 2020 and annual water supply reports: about 40 percent imported through the Metropolitan Water District, about 35 percent the LA Aqueduct in an average year, the rest groundwater and recycled water.
2. Reisner, M. (1986). Cadillac Desert. Viking. The Owens valley acquisition and the aqueduct of 1913.
3. Great Basin Unified Air Pollution Control District, Owens Lake dust control: about 100 million cubic metres of water a year applied; the largest dust source in the United States before 2000.
4. US Bureau of Reclamation, Colorado River Basin: Lake Mead and Lake Powell operations; the 2026 operating guidelines negotiations.
5. City of Los Angeles, Hyperion 2035 programme: full recycling of about 850 million litres a day of treated wastewater to potable use.
6. Photographs: opener: Los Angeles, Winter 2016 by salewskia (CC BY-SA) via Wikimedia Commons; inline: Los Angeles Aqueduct (29081069995) by Tony Webster (CC BY-SA) via Wikimedia Commons; inline: Blowing-alkali-dust-Owens-Lake by Richard Ellis (CC BY) via Wikimedia Commons.

Nepal has more fresh water per person than almost any country on Earth, and its capital gets water a few hours a week. The Kathmandu valley is a bowl of hills at 1,400 metres with the Himalaya on its northern horizon, and it holds four million people on the bed of a lake that drained in the Pleistocene, drinking from rivers that are small, from springs that are drying, from wells that are falling, and from a fleet of tankers that queue at the wells before dawn and sell what they pump by the litre. The tunnel that was meant to end this took twenty three years to dig and was blocked by a flood a month after it opened.

This article is about why a city under the largest snowpack on Earth is short of water, what the valley's stone spouts and its wells say about the ground, what Melamchi was and what happened to it, and what a tanker costs.

The valley

The valley is drained by the Bagmati and its tributaries, rivers that rise in the hills on its rim, not in the Himalaya, and that carry, in the dry months from October to May, a few cubic metres a second between them, and below the city, the city's sewage. The snow of the high mountains drains north and east into other valleys and does not reach Kathmandu. The utility draws from the rivers above the city, from a few dozen deep wells, and from springs on the valley's rim, and its supply is about 150 million litres a day in the dry season against a demand it puts at about 450, so that most of the city's pipes carry water for a few hours on a few days a week and the rest of the demand is met by whatever a household can find.

The rain comes in the monsoon, from June to September, about 1,400 millimetres of it, and it floods the low districts of the valley and runs off through the Bagmati's gorge at the southern end, taking the sewage with it. The problem is storage, and the valley has almost none.

The spouts and the wells

The old cities of the valley, Kathmandu, Patan and Bhaktapur, were watered for a thousand years by stone spouts, dhunge dhara, sunk in courtyards below street level and fed by channels and brick lined canals from ponds on the valley's edge that collected the monsoon and released it slowly through the shallow aquifer. There were several hundred, they were the city's public water until the 1900s, and about half have dried in the last thirty years, because the ponds have been built over, the channels broken by construction, and the shallow aquifer they drew on pulled down by the deep wells around them. The ones that still run are the places where the poor of the old cities queue with jerry cans.

The deep wells are the modern version, and they are the groundwater article's story on this site at a valley's scale. The utility's wells, and the thousands of private ones sunk by hotels, hospitals, factories and apartment blocks through the 1990s and 2000s, draw from a deep aquifer under the lake sediments that fills slowly through the clay above it, and the water table in the central valley has fallen by fifteen to twenty metres in three decades. The tankers draw from wells on the valley's edge, where the aquifer is shallower and the water is closer, and the villages there have watched their own wells fall as the tankers' pumps run through the night.

Melamchi

The Melamchi is a river in the next valley to the north east, fed by the snow the Kathmandu valley never gets, and the plan to bring it through the hills was drawn in the 1980s: a weir on the river, a tunnel twenty six kilometres through the ridge, and a treatment works at Sundarijal on the city's edge, delivering 170 million litres a day, which would more than double the utility's dry season supply. The development banks financed it in 1998. Then the Maoist insurgency closed the site for years, the first contractor left, the second went bankrupt, the earthquake of 2015 cracked the tunnel and the works, the third contractor left in a dispute over payment in 2018, and a fourth finished the tunnel in 2020. Water reached the city in March 2021, twenty three years after the loan and a generation after the design.

In June 2021 a monsoon cloudburst on the Melamchi's upper catchment sent a debris flow down the river that buried the intake, the weir and the access road under metres of boulders and mud, killed several people in the villages along it, and cut the tunnel off from the river. It has since run for a few months each dry season, when the river is low enough to clear a channel to the intake, and stopped each monsoon when the debris moves again, while the engineers argue over a new intake further upstream and the banks over who pays for it. The tunnel that was the answer for forty years delivers, in a good year, half of what it was built for, in the months the valley needs it least.

The tanker

The gap between the utility and the demand is filled by tankers: about five hundred of them, drawing from wells on the valley's edge and selling a five thousand litre load to apartment blocks, hotels and households at a price that varies with the season and, in the dry months, doubles. A household that buys a load a week is paying, per litre, several times the utility's tariff, and the water is untreated, tested by nobody and, in the surveys that have sampled it, contaminated in a large share of loads with the bacteria of the wells it came from. Bottled water in twenty litre jars is the other supply, sold from every corner shop, and it is the Lagos sachet at Himalayan altitude. The valley's water economy, in the absence of the pipe, is a private one run from the back of a lorry, which is the pattern of every city on this site where the utility supplies a third of the need.

The river below

The Bagmati leaves the valley as a sewer. The city has sewers for a share of its people and treatment works for almost none of them, and the river through the old city, past the temples of Pashupatinath where the dead are cremated on its bank, is black in the dry season and carries the city's waste to the plains below. The campaign to clean it, which has run every Saturday for a decade with volunteers pulling plastic from the banks, has not been matched by the treatment works, of which the first at any scale has been under construction for most of that time. The river that could be the city's water, stored behind a dam on the valley rim, is the river the city has made undrinkable, which is the Dhaka and Lahore arithmetic in a mountain valley.

The monsoon that leaves

The valley gets nearly a metre and a half of rain, most of it in four months, and keeps almost none. The lake bed the city sits on is clay for much of its depth, so the rain that falls on the paved valley runs off rather than soaking in, and the ponds and the rice terraces that once held it through the dry season have been built over in the thirty years in which the valley's population tripled. The monsoon floods the low districts along the Bagmati and the Bishnumati every year, the floods carry the sewage and the plastic, and by December the rivers are back to their trickle and the tankers are back at the wells. The recharge that the Bangalore and Chennai articles on this site describe, sending the roof's rain into the ground on purpose, has been made a rule in the valley and is fitted on a few hundred buildings out of a few hundred thousand. The reservoir on the valley rim that would hold the monsoon for the dry season has been in the plans as long as Melamchi and has no financier.

What it teaches

Kathmandu is the city on this site that shows that a country's water and a city's are different things. Nepal has the snow and the rivers; the valley has a small catchment, no storage, a falling aquifer and a tunnel that took a generation and was buried in a month. The stone spouts that watered the old cities dried when the ponds were built over, the wells that replaced them are being emptied, and the tankers that fill the gap sell the valley's edge to its centre by the litre. The tunnel will be repaired. The valley will still have a monsoon it cannot keep and a dry season it cannot fill.

Twenty six kilometres of tunnel, twenty three years to dig it, and a flood that buried the intake a month after the water came.

Sources

1. Kathmandu Upatyaka Khanepani Limited (KUKL), annual reports: demand about 450 million litres a day; supply about 150 in the dry season and 200 in the wet; non revenue water about 40 percent.
2. Asian Development Bank, Melamchi Water Supply Project completion and evaluation reports (2021 to 2023): 26.5 kilometre tunnel; 170 million litres a day; the June 2021 flood damage.
3. Shrestha, S. et al. (2017). Groundwater environment in Kathmandu Valley. In Groundwater Environment in Asian Cities. Elsevier. Water table decline of 15 to 20 metres in the central valley.
4. UN Habitat and NGO Forum, Water Movements in Patan: the stone spouts (dhunge dhara) of the valley and their decline.
5. Kathmandu Valley Water Supply Management Board, tanker water supply surveys: about 500 tankers; prices per 5,000 litre load.
6. Photographs: opener: Panoramic view of Kathmandu Valley from the Chandragiri Cable Car (12) by Saddam Hussain Ansari (CC BY) via Wikimedia Commons; inline: Dhunge Dhara Stone Spout Godawari Kunda Godawari Lalitpur Nepal Rajesh DhunganaJPG (3) by Rajesh Dhungana (CC BY-SA) via Wikimedia Commons; inline: Bagmati River 1 by Gaurav Dhwaj Khadka (CC BY-SA) via Wikimedia Commons.

In Amman the water comes one day a week. Each district of the city has its delivery day, the utility pressurises the pipes for it, and the tanks on every roof fill and are drawn down through the six days until the next; a household that runs its tank dry on a Thursday waits until its day. It has been the system since the 1980s, it is not a drought measure and it is not an emergency, and it is how a city of four million lives in a country with about a hundred cubic metres of water per person per year, which is a tenth of the level that the international agencies call scarcity and among the lowest figures on Earth. Every building in the city carries a row of black plastic tanks on its roof, and the skyline is made of them.

This article is about where Amman's water comes from and why the rota is permanent, what the fossil aquifer under the southern desert is and how long it lasts, what a million refugees did to the arithmetic, why half the water is lost, and what the pipe from the Red Sea is meant to do.

The plateau

Amman sits on a plateau at nine hundred metres, east of the Jordan valley and west of the desert, in a country whose rain falls in a strip along the western hills and almost nowhere else, and whose rivers, the Jordan and the Yarmouk, are shared with neighbours who take most of them upstream. The city drew, for its first century, on springs and wells in the wadis around it and on the Zarqa river, which now carries its sewage; then on wells in the plateau's aquifers, which have been pumped below their recharge since the 1970s and fall by a metre or more a year; then on the King Abdullah canal, which takes the Yarmouk and the side wadis down the Jordan valley for the farms and is pumped, in part, a thousand metres up the escarpment to the city's treatment works at Zai, at a cost in electricity that is a large share of the utility's budget. And since 2013, on Disi.

The rain that does fall on the city, in a wet winter, runs off the plateau into the wadis and to the Dead Sea, which the Israel article on this site describes as a lake that has lost a third of its area because the rivers that fed it were taken first.

Disi

Under the southern desert of Jordan and the north of Saudi Arabia lies a sandstone aquifer that filled with rain in the wetter climate of the last ice age and has received almost nothing since, and in 2013 the government finished a project to tap it: fifty five wells in the desert near the Saudi border, a pipeline of three hundred and twenty five kilometres north to the capital, and a delivery of a hundred million cubic metres a year, about a third of Amman's supply. The water is old, tens of thousands of years, and it carries, as fossil sandstone waters do, a level of natural radium above the drinking limit, which the utility dilutes by blending with other sources. It is the groundwater article's fossil water case in its purest form: a stock, not a flow, spent at a rate the engineers put at fifty years to the end of the useful part, with Saudi Arabia pumping the same aquifer from its side for irrigation at several times Jordan's rate.

Disi was the answer for a generation, and the generation is a short one. The pipeline gave the city its first increase in supply in decades and it did not change the rota, because the demand had grown to meet it before it arrived.

The refugees

Jordan has taken in the refugees of every war around it, and since 2011 the war in Syria has added, by the government's count, well over a million people to a country of ten million, most of them in Amman, Irbid, Zarqa and Mafraq, in the north where the water was already shortest. The demand rose by about a fifth in three years. The rota in the northern cities went from two days a week to one, the wells in the north fell faster, and the refugee camps, at Zaatari above all, were supplied by tanker until boreholes and a network were built for them, at a cost the aid agencies paid and the aquifer bore. The refugees did not create Jordan's water problem. They arrived at the point on the curve where the arithmetic had no margin, and the margin was taken from everyone.

Half the water

Of the water that enters Amman's pipes, about half is not paid for. Some of it leaks from a network that was built fast in the 1970s and 1980s and is pressurised and depressurised every week by the rota, which the leaks article on this site describes as the surest way to break a pipe; some of it is taken through illegal connections, by households, by farms on the city's edge and, in the cases that reach the courts, by organised groups tapping the mains for tankers; and some is delivered to customers whose meters are broken or read wrong. The utility's programme to cut it has been running for twenty years, with results in the districts where the pipes have been replaced and the meters fitted, and the national figure has barely moved, because the network grows as fast as it is fixed. Half of a supply that is a tenth of scarcity is a large number, and it is the cheapest water Jordan could find.

The rota and the tank

The rota has consequences of its own. A pipe that is empty six days a week draws in, through its cracks, whatever is in the ground around it, and the first water of delivery day carries it; a tank on a roof in an Amman summer is warm, and the water in it, chlorinated a week ago at the works, has lost its residual by the fourth day, which is the Legionella and filter jug articles' conditions on every building in the city. Households that can afford it filter, and most drink bottled water, so that the utility's water is for washing and the water that is drunk comes from the shops. The tanks are also the equaliser: a wealthy household has a large tank and a pump and does not notice the rota, a poor one has a small tank and runs dry, and the water that the rota rations fairly on paper is rationed by tank size in practice.

The sea

The last source is the sea. The plan for a canal from the Red Sea to the Dead Sea, which would have desalinated on the way and refilled the Dead Sea with the brine, was studied for two decades with Israel and the Palestinians and abandoned in 2021. Its replacement is Jordan's alone: a desalination plant at Aqaba on the Red Sea, the country's only coast, and a pipeline of four hundred and fifty kilometres north and a thousand metres up to Amman, delivering three hundred million cubic metres a year, which is more than the city uses now, at a cost of several billion dollars and the electricity of a small country. It is under contract, financed by the development banks and the Gulf, and due in the early 2030s. It is the desalination article's last resort, in the country that has run out of every other kind, and it will supply, when it comes, a city that has grown to need it.

The farms in the valley

The Jordan valley below the city is the country's farmland, a strip of irrigated land along the river a few hundred metres below sea level that grows the tomatoes, bananas and citrus of the winter export trade on the water of the King Abdullah canal, and it takes about half of the country's water for a few percent of its economy. The argument in every water strategy since the 1990s has been whether that water should go to the city, and the answer has been the political one: the farms are the valley's people and the tribes' land, the water rights are old, and the treated sewage of Amman, pumped back down the escarpment from the works at As Samra, has been the compromise, so that the city's wastewater irrigates the valley's crops and the valley's canal water goes up to the city. As Samra now treats most of the capital's sewage to a standard that the farms accept, and the arrangement is the largest reuse of wastewater for agriculture in the region, which is the greywater and the Israel articles' pattern on this site at a country's scale.

What it teaches

Amman is the city on this site where scarcity is the design rather than the crisis. The rota is permanent, the tanks are the architecture, the aquifers are falling, the fossil water is being spent on a fifty year clock, half the supply is lost, and the next source is the sea at the far end of the country. The city works, day by day, because four million people have organised their lives around a delivery day, and because Jordan has been living on a tenth of scarcity for so long that the arithmetic is simply how things are. The pipe from Aqaba will double the supply. The tanks will stay on the roofs.

One day a week, a tank on every roof, and water that fell as rain thirty thousand years ago in a pipe from the desert.

Sources

1. Ministry of Water and Irrigation, Jordan, National Water Strategy 2023 to 2040: about 100 cubic metres per person per year; supply rota; non revenue water about 50 percent.
2. Miyahuna (Jordan Water Company, Amman), annual reports: supply schedule by district; rooftop storage; sources by share.
3. Disi Water Conveyance Project documents (2013): 55 wells, 325 kilometres of pipeline, 100 million cubic metres a year; Disi aquifer fossil age and radium content.
4. Aqaba Amman Water Desalination and Conveyance Project (AAWDC), Government of Jordan and financiers (2023 onward): 300 million cubic metres a year, about 450 kilometres of pipeline.
5. UNHCR and Government of Jordan, Syrian refugee population and its effect on water demand in Amman, Irbid, Zarqa and Mafraq.
6. Photographs: opener: Dusk- Amman Cityscape by andryn2006 (CC BY-SA) via Wikimedia Commons; inline: King Abudulla I canal JO by Freedom's Falcon (Public domain) via Wikimedia Commons; inline: Amman Cityscape by Fares Nimri (CC BY-SA) via Wikimedia Commons.

A kilogram of dry pasta carries about 1,850 litres of water, and about 1,800 of them fell as rain on a field of durum wheat before the grain was harvested. The bread article on this site does this arithmetic for soft wheat; pasta's is the same on a harder grain, grown in drier places, with a factory that adds almost nothing and a kitchen that adds, per kilogram, the largest share of any food on this site. This article is about where pasta's water goes, why the crop is a dry country's and Italy imports it from Canada, what a factory uses, what the pot uses and why that became an argument in Italy, and what a plate of spaghetti carries against the sauce on top of it.

The grain

Durum is a hard, amber wheat with a high protein content and a gluten that makes a dough which holds its shape when extruded and dried, which is why pasta is made from it and bread, for the most part, from softer wheats. Durum is a crop of hot dry summers: the south of Italy, Puglia and Sicily above all, where it has been grown since the Romans; the Canadian prairies, which grow more of it than any country, on the summer rain of Saskatchewan; the northern plains of the United States; Kazakhstan; and the dry parts of Greece, Turkey, Syria and North Africa, where it is the wheat of couscous and bulgur. It is sown in autumn or spring, harvested in high summer, and grown almost entirely on rain, so that the footprint is green in the way the wool and sunflower articles use the word: water that fell on farmland that would have grown something else, or nothing.

A hectare yields two to four tonnes, less than soft wheat, which is why the per kilogram figure is a little higher than bread's, and the grey share, for the fertiliser that runs off, is a few percent.

Italy and Canada

Italy makes about a third of the world's pasta, eats more of it per person than anywhere, about twenty three kilograms a year, and grows about half the durum it needs, on the dry plains of the south where the yields are modest and the drought years, which the olive oil article on this site describes for the same fields, take a share. The rest is imported, and the largest supplier for most of the last thirty years has been Canada, whose prairies grow durum at a scale and a protein content the south of Italy cannot match, with the United States, Greece, Kazakhstan and, in some years, Australia behind. The argument in Italy about foreign wheat in Italian pasta, which produced a labelling law in 2017 requiring the origin of the grain on the packet, was about farmers' prices and about a herbicide that Canadian growers use before harvest and Italian ones may not. The water was never in it. A bowl of spaghetti in Rome carries, for the share that came from Canada, the rain of Saskatchewan, which is water nobody in Saskatchewan was short of.

The pattern is the virtual water article's on this site: a dry country importing the crop of a wet one's rain, in the form that travels best, which is grain, and the trade flows from the places with rain to spare to the places without.

The factory

The pasta factory is a mill and an extruder. The durum is milled to semolina, the coarse yellow flour, mixed with about a third of its weight in water into a stiff dough, forced through bronze or teflon dies into the shape, and dried, slowly, in long warm tunnels that take the moisture back out to about twelve percent so that the pasta keeps for years. The water that went in comes out again as vapour in the drier, and the factory's net use is small: a few litres per kilogram for the mixing, the cleaning of the lines and the cooling, which is a fraction of a percent of the grain's. The effluent is starch water from the cleaning and is mild. Pasta is one of the few foods on this site whose processing is close to nothing in water and modest in energy, because the product is dry and the shelf life is long, which is why it has been the food of dry countries and long voyages for five hundred years.

The pot

The pot is where the cook's share is. The instruction on every packet, a litre of water for every hundred grams of pasta, means ten litres for a kilogram, which is more water per kilogram than any other food on this site uses in the kitchen, and all of it is brought to the boil and, at the end, poured down the sink. The ten litres are about half a percent of the pasta's footprint and, in the terms the dishwasher and hot tap articles use, they are the only part of it the cook can change. They are also most of the energy: the life cycle studies of a plate of pasta put the cooking at a third or more of its energy footprint, because heating ten litres of water to the boil and holding it there for ten minutes is more energy than the milling, the extrusion, the drying and the transport together.

In 2022 the Italian physicist Giorgio Parisi, who had just won a Nobel prize, said in a post that pasta could be cooked with the heat turned off after two minutes and the lid on, saving most of the gas, and Italy had an argument. The chefs said the texture suffered; the industry's association published a guide agreeing that the heat could be turned off and the lid put on, and that less water could be used than the packet said; and the physicists pointed out that the pasta does not know how much water it is in, only how hot. The argument was about gas and tradition, and it was, underneath, the one this site keeps finding in the kitchen: the water that runs while nobody is using it, or in this case boils.

The sauce

The pasta is the smaller half of the plate. A hundred grams of dry pasta carries about 185 litres of water and a litre in the pot; the tomato sauce on it, at a hundred and fifty grams of tomatoes, carries about thirty, from the tomato article on this site; the olive oil, a tablespoon, carries about two hundred, from the olive oil article; the cheese, twenty grams of Parmesan, carries about a hundred, from the cheese article; and a meat sauce with a hundred grams of minced beef carries fifteen hundred. A plate of spaghetti with tomato and cheese is about five hundred litres and a plate of bolognese is two thousand, and the difference is the beef, which is the finding every meal on this site produces. The pasta itself, the thing the plate is named for, carries rain on a dry field, and less of it per calorie than almost any staple except the potato.

The dry countries' food

Pasta, couscous and bulgur are the durum foods, and they are the foods of the countries with the least water around the Mediterranean, because durum grows on the winter rain of a dry climate and its products keep without refrigeration through a hot summer. A kilogram of couscous in Morocco or bulgur in Turkey carries the same water as a kilogram of spaghetti, on the same grain, from the same kind of field, and the countries that grow the most of it are among the driest on this site. The trade runs the same way as Italy's: North Africa imports durum from Canada and France in the years its own rain fails, which is more years than it was, and the water in a plate of couscous in Algiers is, in those years, the rain of the Canadian prairie shipped as grain. It is the virtual water pattern that this site finds in every staple, and durum is the clearest case of a crop bred for the dry and now bought from the wet.

What it teaches

Pasta is the food on this site whose water is almost entirely rain on a dry field, whose factory adds nothing, and whose kitchen adds the one share the cook can see: ten litres in a pot, boiled and poured away, which is the argument Italy had with a physicist and the argument this site has with every tap that runs while nobody is using it. The plate's water is in the sauce. The pasta's is in Puglia and Saskatchewan, and it fell from the sky.

1,850 litres a kilogram, ten of them in the pot, and a Nobel laureate who said to put the lid on.

Sources

1. Mekonnen, M.M. and Hoekstra, A.Y. (2011). The green, blue and grey water footprint of crops and derived crop products. Durum wheat and pasta about 1,850 litres per kilogram; wheat global average about 1,800.
2. International Pasta Organisation and UN.A.FOOD Italia, world pasta production about 17 million tonnes; Italy about 3.6 million tonnes.
3. Italmopa and Coldiretti, Italian durum supply: domestic production about 4 million tonnes; imports about 2 million, mainly Canada, the United States, Greece and Kazakhstan.
4. Barilla Center for Food and Nutrition, Double Pyramid and pasta life cycle assessment: cooking about 30 to 40 percent of the energy footprint of a plate of pasta.
5. Parisi, G. (2022). Statement on passive cooking of pasta, and the Italian pasta industry's response; Unione Italiana Food guidance on cooking water ratios.
6. Photographs: opener: (Pasta) by David Adam Kess (pic.2) by David Adam Kess (CC BY-SA) via Wikimedia Commons; inline: Triticum durum sl3 by Stefan.lefnaer (CC BY-SA) via Wikimedia Commons; inline: Cooked dry pasta by MarkTraceur (CC BY-SA) via Wikimedia Commons.

The snow on a piste in the Alps in January was, more likely than not, made the night before from water pumped up the mountain and blown through a fan into cold air. Snowmaking began as insurance for a poor week and became, over thirty years of warmer winters, the foundation of the industry: ninety percent of Italy's pistes, about seventy percent of Austria's and more than half of Switzerland's are covered by artificial snow for part of every season, and a resort that does not make snow at the altitudes where most resorts were built does not open reliably in December. The water is the part of this that the skier does not see, and it is the reason this industry is on this site.

This article is about what snowmaking is and what it uses, where the water comes from and goes, what the reservoirs above the treeline are for, what an Olympic Games in a dry mountain range showed, and what a warmer winter does to the arithmetic.

What a snow cannon does

A snow cannon, or a lance, which is the same thing on a pole, takes water at a few degrees, pushes it through nozzles at pressure with compressed air, and throws the resulting fine droplets twenty or thirty metres into air that is below freezing, where they freeze on the way down into grains of ice that pack into a dense, durable snow that lasts longer on a piste than the natural kind. It works when the wet bulb temperature, which is the temperature with the cooling of evaporation counted in, is below about minus two, and it works best on a cold dry night, which is why the cannons run from dusk to dawn in November and December to build the base before the holidays. A cubic metre of water makes about two cubic metres of snow, roughly; the rest of the volume is air, and a fifth to a third of the water is lost on the way, to evaporation and to wind that carries the plume off the piste.

A kilometre of piste, thirty metres wide, needs a base of thirty to fifty centimetres before it can be skied, which is several thousand cubic metres of water, and it needs topping up through the season as the base wears and melts. A large resort with a hundred kilometres of piste uses between half a million and a million cubic metres of water a season, and the electricity of a small town to pump it up the mountain and drive the fans.

Where the water comes from

The water comes from the mountain, and the timing is the problem. Snowmaking happens in November and December, when the streams of the Alps are at their lowest, because the summer's melt has ended and the winter's has not begun, and a resort that pumped straight from the stream in those weeks would take most of what it carries. So the resorts store: reservoirs dug into the high pasture above the treeline, lined with plastic, filled from the streams through the summer and autumn and drawn down in the snowmaking weeks, holding fifty thousand to several hundred thousand cubic metres each. There are hundreds of them in the Alps now, most built since 2000, in landscapes that had no standing water before, and each is a small dam above a valley, of the kind the tailings article on this site describes in another industry, with a village below it.

Where the resort's own catchment is too small, the water comes from the valley: pumped up from a river a thousand metres below, at a cost in electricity that is the larger part of the snowmaking bill, or from the village's drinking supply, which in the resorts of the Tyrol and the Savoie has been the subject of the argument the hotel article on this site describes about tourism and the towns it sits in. A resort's snowmaking demand, in a dry December, can exceed the drinking water use of its own village, and the village's spring is the same water.

Where it goes

The water is not consumed, mostly. It sits on the mountain as snow until spring, melts, and runs down the same streams into the same rivers, minus the share lost in the making, and a resort's water footprint, in the terms the rest of this site uses, is that share: a fifth to a third of what it pumped, which is tens or hundreds of thousands of cubic metres a season. The rest is moved in time, from autumn to spring, and the moving is what the streams notice. A stream drawn down in November, when it is lowest, loses the flow its trout and its invertebrates depend on in the months they are most stressed; the reservoir that fills it in summer takes the summer's flow; and the melt in April, coming off dense artificial snow that melts later than the natural kind, comes in a pulse that the stream's ecology did not evolve for. The additives that some resorts use to help the water freeze, a protein from a bacterium, go down the stream with it.

The other change is the mountain. The pistes under artificial snow are graded flat, the soil compacted by the machines, the vegetation under a dense, late melting snowpack changed to the species that tolerate it, and the ground under a piste that has been snowmade for twenty years drains differently from the meadow it was.

Beijing

The Winter Olympics of 2022 were held in mountains north of Beijing that get little snow and have, in February, air dry enough to make it, and the organisers made all of it: the alpine venues at Yanqing and the Nordic ones at Zhangjiakou ran on artificial snow entirely, from water piped from reservoirs and a river in a region that the Beijing article on this site describes as among the most water short in China, at a volume the organisers put at a few hundred thousand cubic metres and the critics at more. It was the first Games with no natural snow, it will not be the last, because the candidate cities of the next decades are warmer than the ones of the last, and it showed the industry's future to the industry's customers: a sport that runs on pumped water, in mountains chosen for their politics, in a climate that no longer provides the material.

The warmer winter

The reason for all of it is the temperature. The Alps have warmed by about two degrees since the late nineteenth century, twice the global rate, the snow line has risen by a few hundred metres, and the resorts below 1,500 metres, which are most of the older ones, have lost the natural snow reliability they were built on. Snowmaking is the adaptation, and it has a floor: it needs cold nights, and a warm December that stops natural snow also stops the cannons, so that the resorts that depend on it most are the ones for which it works least. The studies of the Alps under the climate scenarios find snowmaking keeping most resorts above 1,800 metres viable through the century and the lower ones closing one by one, having spent their last decades pumping water up a mountain that was no longer cold enough to hold it.

The village below

The reservoir above the treeline is a dam, and the village at the bottom of the valley is what a dam is above. The Alpine reservoirs for snowmaking are earth embankments lined with plastic, holding tens or hundreds of thousands of cubic metres at two thousand metres in ground that freezes and thaws, and the safety rules that govern them were written, in most of the Alpine countries, only after the first ones had been built. The failures have been few and small so far. The argument about them is the one the tailings article on this site describes for a different industry: a structure that holds water above people, built by a company for its own use, inspected by whoever the canton assigns, in a landscape that the tourists came to see without a lake in it.

The lakes are also the resort's summer product now, marketed for walks and mountain biking around a body of water that did not exist a generation ago, which is the part of the arithmetic that keeps the local councils voting for them.

What it teaches

Ski resorts are the industry on this site whose water use is almost entirely borrowed and whose borrowing changes the lender. The water goes up the mountain in November and comes down in April, a fifth of it evaporated, the streams drawn down at their lowest and flushed at their highest, in reservoirs built on pastures that had none. It keeps the resorts open in winters that would otherwise close them, and it works only while the nights stay cold enough, which is the one thing the industry cannot pump.

Ninety percent of Italy's pistes, a cubic metre of water for two of snow, and a reservoir above the treeline that was a meadow in 1999.

Sources

1. Vanat, L. (2024). International Report on Snow and Mountain Tourism. Share of pistes with snowmaking by country: Italy about 90 percent, Austria about 70, Switzerland about 54, France about 39.
2. François, H. et al. (2023). Distribution of ski resort water and energy demand for snowmaking under climate change. Nature Climate Change 13. Water use per hectare and per resort in the Alps.
3. Steiger, R. and Scott, D. (2020). Ski tourism in a warmer world: increased adaptation and regional economic impacts in Austria. Tourism Management 77.
4. Beijing 2022 Organising Committee and IOC, snowmaking water use at Yanqing and Zhangjiakou: about 100 percent artificial snow; water from reservoirs and the Baihe river.
5. Rixen, C. et al. (2011). Winter tourism and climate change in the Alps: an assessment of resource consumption, snow reliability, and future snowmaking potential. Mountain Research and Development 31.
6. Photographs: opener: Thunder Ridge Ski Area – DemacLenko Snow Fan 01 by Coldstreamer20 (CC BY-SA) via Wikimedia Commons; inline: Speichersee Nähe Mittelstation zur Beschneiung by Franzi-45 (CC BY-SA) via Wikimedia Commons; inline: Espace Villard-Corrençon (38) - Piste Violette - 4 by Floppy36 (CC BY-SA) via Wikimedia Commons.

A car washed with a hose on the drive on a Saturday morning uses between three and five hundred litres of water, most of it running off the bonnet and down the gutter while the sponge is in the bucket, and every litre of it carries the road's grime, the brake dust, the oil film, the phosphate in the soap and the copper from the pads into a storm drain that, in most cities, leads directly to the nearest river with no treatment between. A car washed in a modern commercial tunnel uses about a hundred and fifty litres of fresh water, because most of what touches the car in the tunnel has touched another car first and been filtered in between, and what leaves the tunnel goes to the sewer. The car wash is on this site because it is one of the few industries where the professional version beats the domestic one on water and on effluent both, and because it is the first thing a city bans in a drought anyway.

Where the water goes

A car passing through a tunnel wash is wetted, soaped, brushed or blasted, rinsed, waxed if it is paying for that, rinsed again with softened or deionised water so that it dries without spots, and blown dry, and each stage sprays water at the car from arches and nozzles at a few litres a second for a few seconds. An old tunnel with no reclaim used four to six hundred litres of fresh water per car, all of it once. A modern one collects everything that falls into the pit under the conveyor, settles the grit, skims the oil, filters the rest, and sends it back to the early stages, the wetting, the soap and the first rinse, where a little cleanliness does not matter; fresh water is used only for the final rinse and the spot free rinse, and the fresh water per car falls to a hundred and fifty or below, with the best tunnels under a hundred. An in bay automatic, the roll over kind at a petrol station, is similar. A hand wash business with buckets and a pressure washer in a car park is a driveway wash with more cars.

What goes down the drain

The water is the smaller half of the case. What washes off a car is the residue of the road: fine particles of tyre and brake, which carry copper, zinc and the microplastics the article of that name on this site describes; the film of oil and fuel that every car sheds; the soap, which in the cheaper brands carries phosphate; and the salt in winter. From a drive, all of it runs down the gutter into a storm drain, which in most cities is a separate system from the sewer that carries rain straight to a stream, and the storm overflows article describes what the combined kind does with it. A single car is a small dose. A city's worth of Saturday washes is a measurable one, and the studies of urban streams find the metals of brake dust at levels that the fish notice.

A commercial wash is connected to the sewer, its pit traps the grit and the oil, and what it discharges is treated at the works with everything else. The industry's argument that it is the cleaner option is the industry's argument, and it is also, on this point, the regulator's.

The reclaim system

The reclaim system is a small treatment works under the floor. Water from the wash runs into a series of settling tanks in which the sand and grit drop out and the oil floats and is skimmed; it passes through a cyclone or a filter to take out the finer solids; in the better systems it is treated with ozone or a biological stage to stop it going septic and smelling, which is the reason the older reclaim systems were hated by their operators and their customers; and it is pumped back to the arches. The systems recover sixty to eighty percent of the water, they cost tens of thousands of dollars, and the operators install them because water and sewer charges are a large share of a wash's costs and because, in California, Arizona, Australia and South Africa, a permit or a drought exemption depends on having one. The wash with a reclaim system is the version of every industry on this site that has learned to use its water twice, and it learned for the usual reason.

The drought

The car wash is the first business a city closes in a drought, and it is closed for the reason every city on this site rations by what people can see. A city that has banned lawn sprinklers and asked its residents to take four minute showers cannot leave a tunnel wash running on the main road, whatever its reclaim rate, and the bans that followed the droughts in Sydney in 2007, Cape Town in 2018 and California in 2015 and 2022 closed or restricted the commercial washes along with the hoses. The better drafted rules exempted washes with reclaim and required the rest to close, which is the right distinction; the worse ones closed everything, and the cars were washed on drives with buckets, which used more water per car and sent it to the river. Cape Town in its Day Zero year allowed washes only with recycled or non drinking water, and the washes that survived were the ones that had trucked in borehole water or built a reclaim system in a hurry.

The industry's response has been to make its case in advance, with the numbers in the table above, and to build the reclaim systems that the case depends on, so that the wash on the main road in the next drought has the permit that says it uses a hundred litres and returns them to the sewer. It is a small industry making the argument that every large one on this site makes, and it is easier to see the water in a car wash than in a steelworks, which is the whole of the industry's problem.

The water in the car

The car wash's water is a footnote to the car's. The car article on this site puts the water in the manufacture of a car in the tens of thousands of litres, most of it in the steel, the paint shop and the battery, and the fuel article adds a few litres for every litre in the tank. A car washed once a fortnight for ten years in a modern tunnel has used about forty thousand litres, roughly the water of its own manufacture, and about a third of that if the wash reclaims well. The same car washed on the drive has used three times that and sent it to the river. It is a modest number either way, and the choice between the tunnel and the hose is the largest single water decision a car owner makes after the decision to own the car.

The hand wash in the car park

The fastest growing part of the trade in Europe is the hand wash: a dozen workers in a supermarket car park or a disused petrol station with pressure washers, buckets and sponges, washing a car for a few pounds in twenty minutes. It uses the water of the driveway, two to three hundred litres a car, and sends it where the driveway sends it, down the car park's drains to the storm sewer, with the soap, the wax and the grime. Most are unconnected to the sewer, few have a permit, and the regulators in Britain have found, when they have looked, that a share were discharging illegally and a share were employing people who were not free to leave. The tunnel wash on the main road, with its reclaim system and its sewer connection, is the industry's better half, and the price war with the hand wash in the car park is the reason it has been losing, which is the argument every regulated industry on this site has with the unregulated version next door.

What it teaches

The car wash is the smallest industry on this site and the one that shows most clearly that the visible use is the one that gets banned. Its water is a hundred and fifty litres a car and falling, most of it used twice, and the alternative on the drive uses more and pollutes a river with it. The drought bans that close it are about what a city can see rather than what a city loses, and the washes that survive them are the ones that built the treatment works under the floor before the ban came.

A hundred and fifty litres in the tunnel, five hundred on the drive, and the difference running down a storm drain to the river.

Sources

1. International Carwash Association, Water Use in the Professional Car Wash Industry (2018 and 2023 updates): in bay automatic about 130 to 170 litres fresh water per car; tunnel about 110 to 190; driveway hose 300 to 500.
2. US EPA, Vehicle Washing and Stormwater: pollutants in driveway wash runoff; storm drain versus sanitary sewer discharge.
3. California State Water Resources Control Board, Emergency Regulation 2015 and 2022: restrictions on vehicle washing without a shut off nozzle; commercial washes with recirculation exempted.
4. City of Cape Town, Level 6B water restrictions 2018: commercial car washes permitted only with recycled or non potable water.
5. Brown, C. (2002). Water use in the professional car wash industry. International Carwash Association research report; reclaim system performance 60 to 80 percent.
6. Photographs: opener: Auto wash interior by CurranH (Public domain) via Wikimedia Commons; inline: 2015 Kłodzko, ul. Dusznicka, myjnia samochodowa 02 by Jacek Halicki (CC BY-SA) via Wikimedia Commons; inline: Mister Car Wash, 9516 Snow Heights Circle NE, Albuquerque, New Mexico - May 2026 by Christian David (CC BY-SA) via Wikimedia Commons.

A house with a hundred square metres of roof, in a place with six hundred millimetres of rain a year, which is London, Melbourne, Delhi or Chicago, has about sixty thousand litres of water land on it every year, and in almost every city in the world that water runs into a gutter, down a pipe, into a drain and away. Caught, after the losses, it is about forty eight thousand litres, which is close to the whole indoor use of a careful household of two or three, and it is the one source of water that a city has never been charged for and rarely thinks to keep. This article is about what rainwater harvesting is and what it can supply, why a roof is a clean catchment and what still has to come out, what a tank costs and how large it should be, why some places made it the law and others a habit, and where it fits in the plain water pillar of this site.

What a roof catches

The arithmetic is a multiplication. Roof area in square metres, times rainfall in metres, times a runoff coefficient of about eight tenths for a tiled or metal roof to allow for evaporation, splash and the water that stays on the surface, gives litres in thousands: a hundred square metres, six tenths of a metre, eight tenths, is forty eight cubic metres, or forty eight thousand litres. A larger roof or a wetter place gives more; a dry place gives less, though even Phoenix, with two hundred millimetres, gives sixteen thousand from the same roof, and a city where rain falls in a short season, like Chennai or Mumbai, gives its whole year's yield in a few months and needs a tank sized to hold it.

The catch is the timing. Rain falls when it falls, use is steady, and the tank is the buffer between the two; a tank that is too small overflows in the wet season and runs dry in the dry one, and the sizing calculation, which every rainwater manual gives, balances the tank's cost against the share of the year's use it can cover. For a household using its rain on the garden, the toilet and the laundry, which is about half of indoor and outdoor use together, a tank of a few thousand litres in a climate with year round rain covers most of it; in a monsoon climate the tank has to be much larger or the rain has to go into the ground instead.

What has to come out

A roof is among the cleanest catchments there is, and it is not clean. The water that runs off it carries what landed on it since the last rain: dust, pollen, leaves, the droppings of birds, and, from some roofs, what the roof is made of, which for old lead flashing and some treated timbers and paints is worth knowing. The first flush device, a chamber that fills with the first few litres of each storm and diverts them to the drain before the cleaner water goes to the tank, deals with most of it; a leaf screen on the gutter and a mesh on the tank inlet deal with the rest, and a tank kept dark and closed grows little. Water from a well kept tank is fit, without treatment, for the garden, the toilet and the washing machine, which is how most of the world's household rainwater is used, and it is fit for drinking with the household version of the plain water pillar: a fine filter and an ultraviolet lamp, or a boil. The surveys of tank water in Australia, where a share of rural households drink it, find bacteria in a large minority of tanks and illness attributable to them rarely, and the public health advice is to treat before drinking and to use it untreated for everything else.

The water is soft, which the hard water article on this site explains, and it carries no chlorine, no fluoride and, from a roof in a city, a little of what the city's air carries, which the rain is pure myth article describes.

The tank

The tank is the cost and the space. A polyethylene tank of three to five thousand litres, above ground beside the house, costs a few hundred to a couple of thousand dollars installed with the gutters, the first flush and a pump to feed the toilets and the laundry; a larger underground tank costs several times that; and the payback, at the water tariffs of most rich cities, is long, because mains water is cheap and the tank saves a few tens of cubic metres a year. The economics work where water is expensive or scarce, where the mains are unreliable, where a city subsidises the tank because it is cheaper than a dam, and in the countryside where there is no main at all. Australia's Millennium Drought produced all four at once, and the tank became the ordinary fitting of a new house.

The other tank is the ground. Where the soil takes water, the rain from a roof can be sent into a soakaway, a recharge pit or a well and into the aquifer, which the groundwater article on this site describes as the storage that a city has under it and rarely refills on purpose. Chennai's rule, the Bangalore article's and the Delhi one's are about that: the roof feeds the well.

Bermuda, Chennai, Adelaide

Bermuda has no rivers, no lakes and a small brackish aquifer, and since the seventeenth century its houses have caught their own water: the stepped white limestone roofs that are the island's architecture are catchments, whitewashed to keep the water clean, feeding tanks under the house, and the law requires eighty percent of every roof to be so connected and a tank sized to the roof. The island's water is its rain, house by house, and it has been for four hundred years.

Chennai made it the law in 2003. The city had run dry in the drought of the year before, the aquifer under it had fallen and salted, and the state government required a rainwater harvesting structure on every building, existing and new, with inspections and a deadline, and the aquifer rose by metres in the wet years that followed, by the utility's measurements, and the salt line moved back toward the sea. The Chennai article on this site describes the drought of 2019 that followed anyway, when the rains failed for three years and no roof could catch what did not fall; the rule was the reason the city's wells had anything in them when it came.

Adelaide is the driest capital in Australia, and about half of its households have a tank, most of them fitted in the Millennium Drought under a state rule that required one on every new house and a rebate that paid for the rest. Across Australia about a quarter of households have one, and in the country towns it is the water supply. The tanks cut the cities' mains demand by a few percent, which is a small dam's worth, and they gave a generation of households a gauge on the side of the house that shows how much rain fell, which the water meters article on this site suggests is worth a few percent on its own.

Where it fits

Rainwater harvesting is the plain water pillar with the works removed: catchment, storage, a screen and, for drinking, a filter and a lamp, at the scale of a house. It cannot supply a city; a city's roofs catch a fraction of what a city uses, and in the dry season the tanks are as empty as the reservoirs. It can supply the half of a household's use that does not need drinking water, take the garden and the toilet off the mains in the months that matter, and, sent into the ground, refill the aquifer that the wells are emptying. It is the cheapest new source most cities have, and it is on the roof of every building in them.

What it teaches

Rainwater harvesting is the plain water article about the water a city already has. Forty eight thousand litres a year off an ordinary roof in an ordinary climate, clean enough for half of what a household does with water and, with a lamp, for the rest, running down a drainpipe in every city on this site while the city builds a pipeline. Bermuda has caught it for four centuries, Chennai since 2003, Adelaide since the drought, and the rest have the roof and the drain and the tariff that makes the tank not worth it, until the reservoir is low.

Forty eight thousand litres a year, off a roof that is already there, into a drain that leads to the sea.

Sources

1. Australian Bureau of Statistics, Environmental Issues: Water Use and Conservation (2013 and later): about 26 percent of households with a rainwater tank; 45 percent in Adelaide; the main sources of water for use.
2. Government of Bermuda, Public Health (Water Storage) Regulations: roof catchment and tank storage required, 80 percent of roof area, tank capacity per square foot.
3. Chennai Metropolitan Water Supply and Sewerage Board, Rainwater Harvesting: mandatory since 2003 under the Tamil Nadu Municipal Laws (Second Amendment) Act; groundwater level rise in the following years.
4. Texas Water Development Board (2005). The Texas Manual on Rainwater Harvesting, 3rd edition. Yield calculation, first flush, tank sizing and treatment for potable use.
5. WHO (2022). Guidelines for Drinking Water Quality, chapter on rainwater harvesting: quality of roof runoff and the treatment needed for drinking.
6. Photographs: opener: Bermuda roof by Acroterion (CC BY-SA) via Wikimedia Commons; inline: Rainwater system at nursery house - Nakuru London (ROSA project) (3449001593) by SuSanA Secretariat (CC BY) via Wikimedia Commons; inline: Descente d'eaux gargouille by Jebulon (CC0) via Wikimedia Commons.

A cargo ship that has unloaded is too light to be safe at sea: it rides high, its propeller and rudder are half out of the water, and a wind or a wave can roll it. So it takes on ballast, which for the last century has meant seawater pumped into tanks built into the double bottom and the sides of the hull, tens of thousands of tonnes for a large ship, and carries the water to the port where it will load, where it pumps the water out to make room for the cargo. The world's fleet does this continuously and the sum is about five billion tonnes of seawater a year moved from one sea to another, each tonne carrying, alive, whatever was in the water where it was pumped aboard. This article is about what ballast water is and what it moved, what a mussel and a jellyfish did to two seas, what the convention of 2004 required and why it took thirteen years, and how a ship now treats its water before it lets it go.

Why ships carry water

Ships have always needed ballast. The sailing ships carried stone and sand, loaded by hand at one port and dumped at the next, which is why the harbours of the old trading cities have beaches of foreign rock; the steamships carried water in tanks, which could be pumped, and by the early twentieth century water was the ballast of every large vessel. A bulk carrier that unloads iron ore in China takes on forty or fifty thousand tonnes of water in the harbour there and sails empty to Australia, where it pumps the water out into the harbour at Port Hedland and loads ore; a tanker does the same between the Gulf and Rotterdam; a container ship trims its ballast tanks constantly to balance the boxes. The water is a structural necessity, and there is no substitute for it that a ship could carry.

The pumping is the problem. The water taken on in a harbour is harbour water, full of the plankton, larvae, eggs and spores of everything that lives there, and in the dark tanks of a ship for two weeks a share of them survive, and are released into a harbour on the other side of the world that has never met them.

The mussel and the jellyfish

The zebra mussel is a thumbnail sized freshwater mussel from the rivers of the Caspian and Black Sea basins, and in 1988 it was found in Lake St Clair between Ontario and Michigan, arrived in the ballast of a ship that had taken on water in a Black Sea port and discharged it in the Great Lakes. Within five years it had spread through all five lakes and down the Mississippi, in numbers of tens of thousands to the square metre, and it settles on every hard surface it finds: the intakes of water works and power stations, which it clogs and which have to be scraped and chlorinated, the hulls of boats, the shells of the native mussels it smothers, and the pipes of the cities of the lakes, which the Chicago and Detroit utilities now treat at their intakes for the mussel before they treat for anything else. The cost has been put at hundreds of millions of dollars a year and the mussel is in most of the rivers of the eastern United States.

The comb jelly went the other way. Mnemiopsis is a small transparent jellyfish of the American Atlantic coast, harmless there because it has predators, and in the early 1980s it arrived in the Black Sea in ballast from an American port. It had no predators, it ate the plankton and the eggs and larvae of the fish, and by 1989 it made up most of the living mass in the sea and the anchovy fishery, which had fed Turkey, Bulgaria and the Soviet coast, had collapsed. It was checked, in the end, by a second comb jelly, which arrived in ballast in 1997 and eats the first, and the Black Sea's fish have partly returned; the same jelly has since reached the Caspian, the Baltic and the North Sea. The list of ballast introductions runs to hundreds: the Chinese mitten crab in the Thames and the Elbe, the North Pacific seastar in Tasmania, the cholera bacterium itself, which was found in the ballast of ships arriving in the American Gulf from Latin America in the epidemic of 1991.

The convention

The International Maritime Organization, which writes the rules for shipping, adopted a convention on ballast water in 2004, after fifteen years of drafting, and it took another thirteen years to come into force, in 2017, because a convention needs the ratification of countries holding a share of the world's tonnage and the shipowners argued, with some reason, that the treatment systems it required did not yet exist at the scale of a fleet. The convention has two standards. The first, exchange, required a ship to pump out its coastal ballast in mid ocean and replace it with open ocean water, which carries far fewer organisms and ones that will not survive in a harbour; it is cheap, it is dangerous in a storm, and it was the interim measure. The second, the performance standard, sets a limit on what may be discharged: fewer than ten living organisms larger than fifty micrometres in a cubic metre, fewer than ten per millilitre in the size class below, and limits on the cholera bacterium, E. coli and enterococci, which is a drinking water standard for the sea. Since 2024 every ship in international trade has been required to meet it.

The works below the waterline

Meeting it means a treatment plant on the ship. The systems that have been approved, several dozen of them, follow the plain water pillar of this site in miniature: a filter, usually a screen of fifty micrometres that takes out the larger organisms and the sediment as the water is pumped aboard, and then a disinfection stage, which is ultraviolet light in about half of the systems and electrochlorination in most of the rest, in which the seawater's own salt is turned into chlorine by an electric current, dosed into the tanks, and neutralised again before discharge. Ozone, heat and deoxygenation are used on a few. The systems cost a few hundred thousand to a few million dollars a ship, the world's fleet of some sixty thousand ships has fitted them over a decade, and the port inspectors now sample the discharge and count what is alive in it, which is the first time the sea has had an inspector at the outfall.

The systems work, on the tests, and the arguments now are about the ones that do not, in cold or turbid water where the ultraviolet cannot reach, and about the sediment in the bottom of the tanks, which the convention also covers and which carries the cysts of the algae that survive everything. The mussels and the jellies that have already moved will not move back.

The sediment and the hull

Two things the convention reaches for and does not fully hold. The first is the sediment: the mud that settles in the bottom of a ballast tank over years of pumping harbour water in and out, which holds the resting cysts of the algae that cause the toxic blooms, and which the treatment systems, which treat the water as it flows, do not touch. The convention requires the tanks to be cleaned and the sediment landed at a port facility, and the ports with such a facility are few. The second is the hull itself, on which the barnacles, the mussels and the weed of one harbour grow and travel to the next, and which the convention on ballast does not cover; the organisms moved on hulls are, by the surveys, at least as many as those moved in tanks, and the rules for them are guidelines rather than law. The ship, in other words, is treated inside and carries the sea on its outside, and the mussel that the Great Lakes' intakes are scraped for could have come either way.

What it teaches

Ballast water is the plain water article about the water nobody meant to move. Five billion tonnes a year, pumped from one harbour into another for a reason that has nothing to do with water, carried the living contents of every port to every other, and the mussel in the Great Lakes' pipes and the jelly in the Black Sea's plankton are the result. The fix, agreed in 2004 and fitted by 2024, is a water treatment works in every ship's hull, filtering and disinfecting seawater to a standard written like a drinking water rule, because that is what it took to stop the sea being poured, alive, from one side of the world into the other.

Five billion tonnes a year, seven thousand species in transit, and a treatment works below the waterline of every ship.

Sources

1. International Maritime Organization, International Convention for the Control and Management of Ships' Ballast Water and Sediments (BWM Convention), adopted 2004, in force 8 September 2017; the D-1 exchange and D-2 performance standards.
2. Carlton, J.T. (1999). The scale and ecological consequences of biological invasions in the world's oceans. In Invasive Species and Biodiversity Management. Kluwer. Estimate of species in transit.
3. Pimentel, D. et al. (2005). Update on the environmental and economic costs associated with alien invasive species in the United States. Ecological Economics 52. Zebra mussel costs.
4. Kideys, A.E. (2002). Fall and rise of the Black Sea ecosystem. Science 297. Mnemiopsis leidyi and the anchovy fishery.
5. Lloyd's Register and IMO GloBallast, ballast water treatment technologies: filtration with ultraviolet or electrochlorination; type approval and the 2024 status of the world fleet.
6. Photographs: opener: 2024-09-11 HAPPINESS BULKER - IMO 9919515 – Port Angeles WA USA by Gordon Leggett (CC BY) via Wikimedia Commons; inline: Zebra Mussels on Boat Propeller by US Fish and Wildlife Service (Public domain) via Wikimedia Commons; inline: Bulk carrier ship view from top by Alet123 (CC BY-SA) via Wikimedia Commons.

In a drought the swimming pool is the thing the neighbours point at. It is a tank of drinking water in a back garden, it evaporates all summer, the children splash it out, and it looks like exactly what a city short of water should not have. The lawn beside it, which is usually what the pool replaced, looks like nothing at all, and in the dry cities of the world where pools and lawns are both common, it uses more water. The arithmetic is plain, it has been measured in the cities where it matters, and the myth survives because a pool can be seen and a sprinkler runs at dawn.

This article is about what a pool loses and what a lawn loses to the same sun, why the fill is a one time cost and the sprinkler is not, what a cover changes, what the studies in Phoenix and Perth found, what a pool sends to the drain, and what the honest comparison is.

What a pool loses

A pool loses water three ways. Evaporation is the largest: an open surface of water in a hot dry climate loses, to the sun and the wind, about a metre and a half of depth a year, more in Phoenix and less in a cooler or more humid place, and a pool of fifty square metres therefore loses about seventy five cubic metres a year, roughly two hundred litres a day in summer and a fraction of that in winter. Splash and carry out, on the swimmers, is a smaller share. And backwashing, the cleaning of the sand filter by running water backwards through it, sends a few hundred litres to the drain each time it is done, which is once a week or two in the season. The fill, which is the number that looks largest, is fifty to seventy cubic metres for a pool of that size, once, when it is built, and a pool that is looked after is not emptied again for a decade.

A cover changes the first number. A floating cover, the bubble wrap kind or a rigid one, cuts evaporation by half to seventy percent, and a covered pool in Phoenix loses about half a metre a year rather than a metre and a half. It also keeps the heat in and the leaves out, which is why the energy agencies recommend it before the water agencies do.

What a lawn loses

A lawn in the same climate loses water through the grass, by the evaporation and transpiration that the lawn article on this site describes, at the rate the sun and the wind set, which in Phoenix, Perth or Las Vegas is about a metre and a half to two metres a year for a lawn kept green, and it loses it every year, because the sprinkler that replaces it is the household's largest single use of water and runs from spring to autumn. A lawn of fifty square metres therefore takes about seventy five to a hundred cubic metres of sprinkler water a year, and the studies of how people actually water find that most lawns get more than they need, because the timer is set for the hottest week and left, and because the sprinkler waters the pavement.

The comparison is direct. A pool and a lawn of the same size, in the same garden, in the same summer, lose water to the same sun at about the same rate, because an open water surface and a well watered grass surface evaporate at nearly the same rate per square metre. The pool loses a little less, because it has no leaves pumping water into the air, and it can be covered, which a lawn cannot.

What the cities measured

The two cities that have metered this most carefully are the two with the most pools per household in the dry world. Perth's water utility, in its residential use studies, found pools at about four percent of a household's water and irrigated lawns and gardens at more than forty, and found that a household that built a pool on its lawn used, on average, no more water than before, because the grass had gone. Phoenix's studies found pool evaporation at about a metre and a half a year and turf irrigation at nearly two, and the analysis of the city's household data found lot size and turf area, not pool ownership, to be the drivers of outdoor use; a pool on a small paved lot used less than a lawn on a large one. Los Angeles, in the drought rules of 2015 and 2022, restricted lawn sprinklers to two days a week and then one and left pools alone, on the arithmetic above, and paid households to replace turf with gravel and native plants at a rate per square metre, and it did not pay them to fill in pools.

The myth persists in those cities regardless, because the pool is visible and the sprinkler runs at five in the morning, and because a pool is a luxury and a lawn is a default, and a city rationing water rations by what looks like a luxury.

What a pool sends to the drain

The pool's other cost is the chemistry. A pool is chlorinated, at levels the chlorine byproducts article on this site describes as making the compounds that give an indoor pool its smell, and the backwash and the occasional drain send that water, with its chlorine, its salt if the pool is a saltwater one, and the sweat and sunscreen of the summer, to the sewer or, in the worse cases, the storm drain and the creek. It is a small volume and the utilities' rules say where it should go. The lawn's cost is the fertiliser and the weedkiller, which the lawns article describes running off into the same creek in larger quantities from more gardens, and it is the sprinkler that carries them.

The honest comparison

The honest comparison is neither of them against the other. A pool uses the water of a lawn, and a covered pool uses less than half of it; a lawn uses the water of a pool, and a lawn in a dry city is the largest thing a household could stop watering. The garden that uses no water is the one with neither: gravel, native planting, a tree for shade and a patio, which is what Las Vegas and Los Angeles are paying households to build and what Perth's newer suburbs are built with. A household that wants a pool in a dry city should cover it, keep it full rather than refilling, and run the backwash to the sewer, and it will have a pool that uses less water than the lawn its neighbour is running a sprinkler on while pointing at it.

The public pool

The pool that uses least water per swimmer is the one with the most swimmers. A municipal pool of five hundred square metres loses to evaporation what ten back garden pools do and serves a thousand people a day, and the water per swim, counted that way, is a few litres, which is the dishwasher article's arithmetic on this site: the shared machine beats the private one because it is full. The cities that have restricted private pool filling in droughts, Cape Town and the towns of southern France in 2023 among them, have mostly kept the public pools open for the same reason, and the argument in France, where the mayors of a few dry communes banned new private pools outright, was about the two hundred thousand private pools built in a decade in a region that has one long dry summer a year. A pool per household is a lawn per household by another name. A pool per town is a different thing. The lido that a town fills in May and drains in September, with a thousand people in it on a hot Saturday, is the most water efficient swimming there is, and it is the one that the drought rules never need to close.

What it teaches

The pool myth is the one on this site that is about what water looks like. A tank of water in the sun looks like waste and a lawn looks like a garden, and the meter says they are the same, or that the lawn is worse, in every dry city that has measured them. The pool can be covered and the lawn can be removed, and the city that rations by what it can see bans the pool fill and lets the sprinkler run. The water was never in the pool. It was in the air above both.

A metre and a half a year off the pool, two metres onto the lawn, and a cover that halves the one thing the neighbours can see.

Sources

1. Water Corporation of Western Australia, Perth Residential Water Use Study: swimming pools about 4 percent of household use; irrigated lawn the largest outdoor use; pool evaporation about 1.5 metres a year uncovered.
2. Arizona Municipal Water Users Association, Residential Pool Water Use studies (Phoenix): about 1.5 metres a year evaporation; comparison with turf at about 1.7 to 2 metres of irrigation a year.
3. Wentz, E.A. and Gober, P. (2007). Determinants of small area water consumption for the city of Phoenix. Water Resources Management 21. Pools, turf and lot size as drivers of household use.
4. Los Angeles Department of Water and Power, turf replacement rebates and outdoor water use estimates: irrigation about half of single family household use.
5. US Department of Energy, Swimming Pool Covers: evaporation reduction of 50 to 70 percent with a cover.
6. Photographs: opener: Upton House and Gardens - swimming pool (27370979195) by Elliott Brown (CC BY-SA) via Wikimedia Commons; inline: Swimming pool cover by Sikander Iqbal (CC BY-SA) via Wikimedia Commons; inline: Travelling irrigation sprinkler 2 2014-05-29 by Slaunger (CC BY-SA) via Wikimedia Commons.

A kilogram of soybeans carries about two thousand litres of water, which puts it with wheat and above rice, and most of that kilogram is never eaten as soy. The bean is crushed for its oil, which goes into cooking oil and margarine, and the protein rich meal that remains, which is most of the bean by weight, goes into the feed of pigs, chickens and farmed fish, so that the soy in the world's diet arrives, for the most part, as meat. Tofu, soy milk, tempeh and sauce, the foods that the word calls to mind, take about seven percent of the crop.

This article is about where a soybean's water goes, why the crop is the least irrigated of the great crops and the most argued over, what the foods made directly from it carry against the foods they replace, and what the spread of the crop across the savanna of Brazil has to do with the water it carries.

The crop

Soy is a legume, a plant that fixes its own nitrogen from the air with bacteria in its roots, which is why it needs little fertiliser and why it is grown in rotation with maize on the same fields. It is a summer crop, sown in spring and harvested in autumn, and it grows on rain: the corn belt of the American midwest, the plateau of central Brazil and the pampas of Argentina, which between them grow four fifths of the world's crop, all have wet summers, and irrigated soy is rare outside a few dry districts. The global footprint of about 2,100 litres a kilogram is ninety five percent green, with a small blue share from those districts and a small grey share from the fertiliser and the herbicide that a crop bred to tolerate herbicide receives.

A hectare yields about three tonnes, and the bean is about twenty percent oil and forty percent protein, which is the highest protein content of any major crop and the reason the world grows so much of it. The world's crop has quadrupled since 1990, faster than any other major crop, and almost all of the growth has been in South America and almost all of it for feed.

Where the soy goes

About three quarters of the world's crop is crushed, and the crushing gives about a fifth oil and four fifths meal. The oil is the second vegetable oil in the world after palm, and it goes into cooking oil, margarine, processed food and, increasingly, biodiesel. The meal is the world's largest source of feed protein, and it goes into the rations of chickens above all, then pigs, then dairy cattle and farmed fish, so that the chicken and pork articles on this site, which count the water in the feed, are counting soy. About seven percent of the crop is eaten directly by people, mostly in East Asia as tofu, soy milk, natto, tempeh and sauce, and a few percent goes to whole beans and edamame.

The consequence for the water arithmetic is that the water in a kilogram of soy is mostly in a kilogram of chicken, at a conversion of about two kilograms of feed per kilogram of bird, and the way to reduce the world's soy footprint is to eat less meat rather than less tofu.

Tofu and soy milk

The foods made directly from soy carry the bean's water and a little more. Tofu is soy milk curdled with a salt and pressed, and it takes about half a kilogram of beans to make a kilogram, plus the water of the processing, so that it carries about 2,500 litres; soy milk is beans ground with water and strained, at a tenth of a kilogram of beans per litre, and carries about 300. Against the foods they replace the comparison is one sided: tofu against chicken by a factor of nearly two, against beef by six, and soy milk against cow's milk by three, on water, and by larger factors on land and on carbon. The processing plant that makes them is a small food factory with a mild effluent, the whey of the tofu press, which is high in sugar and goes to a digester or to pigs.

The soy in a European or American carton of soy milk is, by the brands' own accounts, almost always grown in Europe or North America rather than Brazil, because the buyers of tofu and soy milk are the buyers who ask. The soy in the chicken is not asked about.

The Cerrado

The argument about soy is about land, and the land is in Brazil. The crop spread across the Cerrado, the savanna of central Brazil, from the 1970s, when Brazilian scientists found how to lime its acid soils and breed varieties for its latitude, and it has since covered an area the size of Germany, most of it converted from native savanna and some of it, further north, from the Amazon forest. A moratorium agreed by the traders in 2006 stopped the purchase of soy from land cleared in the Amazon after that date, and the clearing of forest for soy fell sharply; the Cerrado, which is less famous, less protected and holds a third of Brazil's species and the headwaters of most of its rivers, was not covered, and the clearing moved there. The water in a Brazilian soybean is the rain that used to fall on a savanna whose deep roots held it and fed the rivers of the São Francisco and the Paraná, and the São Paulo article on this site is, at one remove, about that.

The Cerrado's rain is also changing. The clearing of the savanna and the forest north of it has reduced the moisture that the forest recycles into the air and that falls as rain on the fields, and the droughts of the 2010s and 2020s in central Brazil, which have cut soy yields and stopped the barges on the rivers that carry the crop to port, are, by the models, in part the crop's own doing.

The bean in the tank

A growing share of the oil is fuel. Brazil, Argentina and the United States blend biodiesel made from soy oil into their diesel by law, at ten to fifteen percent, and the water of the soy in a litre of that diesel is about two thousand litres, for a fuel whose climate benefit over the diesel it replaces is, once the land clearing is counted, small or negative. The hydrogen article on this site makes the same point about fuel made from water; the fuel made from a crop carries the crop's water and the crop's land, and the argument for it is that the crop was being grown anyway, which in the Cerrado is exactly the question.

The bean in Asia

The crop's home is East Asia, where it was domesticated three thousand years ago and where it is still eaten as itself. China was the world's largest producer until the 1990s and is now the largest importer, buying more than half of all the soy that crosses a border, almost all of it from Brazil and the United States, to feed the pigs and chickens that a richer diet demands; its own crop, which is smaller than its imports, goes to tofu, soy milk and sauce, on which the country does not want to depend on foreign beans. Japan and Korea, which eat more soy per person than anywhere, import nearly all of it. The water arithmetic of that trade is the virtual water article's: China imports, in its soybeans, more water than any country imports in any commodity, and the water is rain on the Cerrado and the corn belt, shipped as protein at a fraction of the cost of the water it would take to grow it on the North China Plain, which the groundwater article on this site describes as running out of water for the crops it already has. The bean crosses the Pacific at about a fortieth of the cost of the water it carries, which is why it moves and the water does not, and why a drought on the Brazilian plateau is now a matter for the pig farmers of Sichuan.

What it teaches

Soy is the crop on this site that carries a modest amount of water and a great deal of argument, and the argument is mostly about the crop's other consumers. Its water is rain on farmland; three quarters of it goes into animals, so that the water in the world's soy is mostly in the world's chicken; and the foods made from it directly carry less water, land and carbon than any animal food they replace. The Cerrado, which the crop has covered, is where the rain that grows it used to feed the rivers, and that is where the footprint of a soybean actually lands.

2,100 litres a kilogram, three quarters of it in a chicken, and a savanna the size of Germany under the crop.

Sources

1. Mekonnen, M.M. and Hoekstra, A.Y. (2011). The green, blue and grey water footprint of crops and derived crop products. Soybeans about 2,145 litres per kilogram; tofu about 2,500; soy milk about 300 per litre.
2. Ritchie, H. (2021). Is our appetite for soy driving deforestation in the Amazon? Our World in Data. About 77 percent of soy to animal feed; about 7 percent to direct human food.
3. USDA Foreign Agricultural Service, Oilseeds: World Markets and Trade. Brazil, the United States and Argentina about 80 percent of production.
4. Soterroni, A.C. et al. (2019). Expanding the Soy Moratorium to Brazil's Cerrado. Science Advances 5.
5. Poore, J. and Nemecek, T. (2018). Reducing food's environmental impacts through producers and consumers. Science 360. Footprints of tofu, soy milk, and the animal products they substitute.
6. Photographs: opener: Mike Starkey Soybean Harvest (52490323767) by US Department of Agriculture (Public domain) via Wikimedia Commons; inline: Making Tofu - 20170206 - Spliting by DragonSamYU (CC BY-SA) via Wikimedia Commons; inline: PLANTAÇÃO DE SOJA SAFRA 2013 - DOM OSORIO - panoramio by Paulo Roberto (CC BY-SA) via Wikimedia Commons.

Santiago sits in a bowl at the foot of the Andes, seven million people between the mountains and the coast range, and its water comes down from the snow. The Maipo river, which drains the high Andes to the south east of the city, supplies about eighty percent of what the city drinks, from a reservoir at El Yeso and from the river itself, and the Maipo runs on the melt of the winter snowpack through the long dry summer. Since 2010 central Chile has had a run of dry winters longer than anything in its record, the snowpack has shrunk to a fraction of its former depth, and the river that the city was built beside has carried, in the worst years, less than half of what it did.

This article is about what thirteen years of drought have done to the city's water, how Chile came to treat water as private property and what that has meant when there is less of it, what happened to a lake that dried up with its water rights intact, and what it means when a city of seven million writes a plan to ration.

The snow

Santiago's rain falls in winter, from May to August, on a Mediterranean pattern like California's, and most of it falls as snow in the Andes above the city, where it lies until spring and melts through the summer into the rivers. The Maipo and the Mapocho, which runs through the city itself, are snowmelt rivers, and the city's summer, when its gardens and its fruit farms need water most, is when the snow supplies it. The water company draws from the Maipo above the city, holds a reserve at El Yeso, a reservoir at 2,500 metres in the mountains, and pumps the rest from wells in the valley floor, which are fed, in the end, by the same rivers.

The drought that began in 2010 has cut the winter rain by a third on average and, in the worst years, by more than half, and the snow by more, because the winters have been warmer as well as drier and the snow line has risen. The river's summer flow has fallen with it, El Yeso has been drawn down to the levels the company keeps as a last reserve, and the wells in the valley have deepened.

Water as property

Chile's water law is the most market based in the world, and it was written in 1981, under the military government, by economists who held that water, like land, would be best used if it were owned. The code separated rights to water from rights to land, granted them in perpetuity and free of charge to whoever applied, and made them tradable, so that a farmer could sell his river allocation to a mine, a mine to a utility, and a utility to a fruit exporter, at whatever price they agreed. The state's water agency registered the rights and allocated new ones until the rivers were fully allocated, and in several basins, the Maipo among them, beyond that, on the flows of wet decades.

The theory was that the market would move water to its most valuable use. In practice it moved it to the users who could pay, which in central Chile has meant the avocado and fruit exporters, the mines and the city, and away from the small farmers and rural villages who sold or were never granted rights; and it left no mechanism for reducing everyone's draw when the river fell, because a right is a right, and the last user on the river is the one who finds it empty. The reform of 2022, passed after a decade of argument and the protests of 2019, made new rights temporary, put drinking water first in a shortage, and gave the state the power to cut allocations in a declared scarcity. The rights already granted, which are most of the water, were left as they were.

Aculeo

Laguna de Aculeo was a lake of twelve square kilometres in the coast range an hour south of the city, ringed by weekend houses, with boats on it and a fishery in it, and between 2010 and 2018 it dried completely. The drought took the rain that filled it; the streams that fed it were drawn down by the farms and the houses upstream, each exercising rights granted when the streams ran full; and the lake, which had no right of its own, went last. In the summer of 2018 the lakebed was dry mud with the boats sitting on it, and it has held water only briefly since. The lake is the image of the drought in Chile, and it is also the image of the code: every user upstream of it was within their rights.

The same arithmetic has run in Petorca, the valley north of the city where the avocado article on this site describes the orchards taking the river and the villages drinking from tankers, and in the Aconcagua, where the mines, the fruit farms and the towns hold, between them, rights to more water than the river has carried in any year since 2010.

The plan to ration

In April 2022, after the driest winter in the record, the governor of the Santiago region announced a rationing plan, the first in the city's history: four stages, from a public campaign to cuts in pressure to rotational cuts by district of up to twenty four hours, to be triggered by the Maipo's flow. The plan was published, the winter of 2022 was better than the ones before it, the river recovered enough to keep the plan at its first stage, and the plan stayed on the shelf, which is where a rationing plan should be. The company has since spent on the things the plan assumed: more storage in the valley, deeper wells, a reduction of leakage that runs at a third of the supply, and, in the long run, a study of desalination on the coast a hundred kilometres away, which is where Antofagasta and the mining cities of the north already drink from.

The city's use, per person, has fallen by a fifth through the drought, with campaigns and prices, and it is still among the highest in Latin America, because Santiago's wealthier districts have gardens and pools in a climate that has decided it is a desert.

The avocado and the mine

Two users of the Maipo's water are worth naming because they hold the rights. The fruit exporters, whose avocados and grapes go to Europe and China from orchards in the valleys around the city, hold rights bought over three decades from the small farms that could not compete, and they use them through the summer when the river is lowest; the avocado article on this site describes the same arithmetic in the Petorca valley to the north, where the river ran dry and the villages went onto tankers while the orchards stayed green. The mines, above the city in the Andes, hold rights to the headwaters, and the copper article on this site describes what a mine does with water and why the largest in Chile have moved to desalinated water piped up from the coast at a cost that only copper can pay. The city's water company holds the third block of rights and buys more when it can, and in a drought, the rights it holds are worth what the river carries, which is the point.

The coast

The answer that the mining cities of the north found twenty years ago is being studied for the capital. Antofagasta, in the Atacama, drinks desalinated water and has since 2003; the mines around it have built plants that pump seawater a hundred kilometres inland and three thousand metres up. Santiago is a hundred kilometres from the sea over a coast range, and a plant on the coast at Valparaíso with a pipe to the city would cost several billion dollars and the electricity of a small city, which is the price the desalination article on this site puts on the last resort. The company's plan puts it in the 2030s, after the wells, the leaks and the storage, and after the winters have shown whether the drought is a run of dry years or the climate.

What it teaches

Santiago is the city on this site where the water was allocated by a market and the river was allocated by the weather, and the two have not agreed for thirteen years. The snow that fed the city is shrinking, the rights to the rivers add up to more than the rivers carry, and the lake that had no rights dried up first. The reform of 2022 put drinking water at the head of the queue and left the queue as it was, and the city's first rationing plan waits on the next dry winter, which the climate models say is the ordinary kind now.

Thirteen years of drought, a river owned on paper by more people than it can supply, and a lake that dried while every right on it was exercised.

Sources

1. Garreaud, R.D. et al. (2020). The Central Chile Mega Drought (2010 to 2018): a climate dynamics perspective. International Journal of Climatology 40. Precipitation deficits of 20 to 40 percent; snowpack decline.
2. Aguas Andinas, annual reports: the Maipo river about 80 percent of Santiago's supply; El Yeso reservoir; the Mapocho and wells the rest.
3. Dirección General de Aguas (DGA), Chile: water rights registry; over allocation of the Maipo and Aconcagua basins; the 2022 reform of the Water Code (Ley 21.435).
4. Muñoz, A.A. et al. (2020). Water crisis in Petorca basin, Chile: the combined effects of a mega drought and water management. Water 12.
5. Valdés Pineda, R. et al. (2021). Drying of Laguna de Aculeo. Regional Environmental Change; Gobierno de Santiago, Plan de racionamiento, April 2022.
6. Photographs: opener: Sanhattan desde el Cerro San Cristóbal, Santiago de Chile by AlexVanHeusen (CC BY-SA) via Wikimedia Commons; inline: El Yeso Dam, Chile (Unsplash) by Pablo Acevedo (CC0) via Wikimedia Commons; inline: Aculeo oli2 2023250 lrg by NASA Earth Observatory (Public domain) via Wikimedia Commons.

Metro Manila is fourteen million people on a bay at the mouth of a river, in a country that gets two and a half metres of rain a year and a typhoon every few weeks in season, and it drinks almost entirely from one reservoir in the mountains. The Angat dam, sixty kilometres north east of the city, supplies about ninety seven percent of the raw water that the city's two water companies treat and pipe, and when the rains fail and the reservoir falls below the level the regulator sets, the companies turn districts off by rota. The city has known this for forty years, and the second source has been planned for most of them.

This article is about where Manila's water comes from, what the largest water privatisation in the world did to the pipes, what happened in March 2019, why the next dam is contested, and what it means to have too much rain and too little water in the same year.

One dam

Angat was built in 1967 across a river in the Sierra Madre foothills, to supply the city, to irrigate the rice of the plain below, and to make electricity, in that order of priority when the reservoir is low and the reverse when it is full. The water is drawn off through a tunnel to a smaller reservoir at Ipo, then by aqueduct to the La Mesa reservoir on the city's edge, where the two companies' treatment works take it, and the arrangement has not changed since the city had three million people. The allocation is four billion litres a day, and the city's demand has grown to the allocation and past it, so that the reserve the dam provides in a dry year is what the rice farmers give up.

The reservoir is filled by the typhoons of the second half of the year and drawn down through the dry season, and in a year when the typhoons pass elsewhere, as in the El Niño years of 1998, 2010, 2016 and 2019, it goes into the dry season low and reaches, by March or April, the level at which the regulator cuts the city's allocation. The city has been rationed by rota in each of those years.

The concessions

In 1997 the state utility served about two thirds of the city, for a few hours a day in most of it, and lost about two thirds of its water to leaks and illegal connections, and the government, at the urging of the development banks, split it into two zones and sold twenty five year concessions to private companies. The east zone went to a consortium led by a Filipino conglomerate, and it became, over the following decade, the case study that the industry cites: leakage cut from sixty three percent to under ten, supply extended to twenty four hours across the zone, and, most cited of all, the connection of the informal settlements, through community schemes in which a single meter served a cluster of houses and the residents ran the distribution, at a price below the vendors' and above the utility's standard tariff. The west zone, which had the older pipes and the poorer districts, went to a consortium that struggled, went through a bankruptcy and a change of owners, and reached similar results a decade later.

The concessions fixed the pipes. They did not, and could not, build a dam, because the raw water was the state's, and the state's second source, planned from the 2000s under one name and another, was not built. The companies and the regulator argued for fifteen years about who should pay for it, and in the meantime the city's demand grew into the allocation and the margin went.

March 2019

In the first week of March 2019 the La Mesa reservoir, the small lake at the end of the aqueduct from which the east zone's works draw, fell below the level of the works' intake, and the east zone company began turning off districts, at first for hours and then for most of the day, across the eastern half of the city. It went on for weeks. Queues formed at the fire hydrants the company opened; hospitals ran on tankers; the malls, which have their own wells, became the places people went to use a toilet; and the Senate held hearings at which the company, the regulator and the dam's operator blamed each other. The immediate cause was a dry year and a demand that had reached the allocation with no margin, and the deeper one was the second source: a treatment plant on Laguna lake, which the company had been building to add a tenth to its supply, was a year late, and the dam that would have added a sixth had been in the planning stage for fifteen years.

The company was fined and its concession renegotiated. The rain came in June. The reservoir fell again in the dry season of 2020, and the city was rationed again, more quietly, during the pandemic.

The dam in the forest

The second source is the Kaliwa dam, on a river in the Sierra Madre east of the city, and it has been on the plans since the 1970s under one name or another. It was finally contracted in 2018, with a loan from China and a Chinese contractor, to deliver six hundred million litres a day, a seventh of the city's use, through a tunnel to the existing aqueducts, and the tunnelling began in 2022. The dam floods land that the Dumagat and Remontado peoples hold under ancestral title, and their opposition, and that of the environmental groups who argue that the Sierra Madre's forest is the barrier that protects the city from the typhoons, has slowed it and taken it to the human rights commission and the courts. The government's answer is the one every government in this position gives: the city needs the water, and the alternative, a larger dam further out, floods more.

The dam will add a seventh. The city grows by that much in a decade, and the plan beyond Kaliwa is a larger dam beyond it, in the same mountains, with the same objections.

Too much water

Manila's other water problem is the opposite one. The city floods every year, in the typhoons and the monsoon, and in September 2009 a typhoon called Ondoy dropped a month's rain in a day and put a third of the metropolis under water, killing hundreds. The drainage was built for a smaller city and is choked with the settlements built along the creeks and the plastic in the channels, the bay is rising, and the city is sinking, as the Jakarta and Bangkok articles on this site describe, where the groundwater has been pumped. The wells that the malls and the factories run, outside the concessions' control, are the reason the city's northern districts have subsided by several centimetres a year. A city that floods in September and is rationed in March has the water. What it lacks is the storage and the pipes between the two, which is the story of every monsoon city on this site.

The wells and the malls

Outside the two concessions there is a third supply that nobody regulates well: the deep wells that the malls, the factories, the hospitals and the subdivisions sank through the decades when the utility could not be relied on, and that they kept when it could. The water agency licenses them and has tried, since the 2000s, to close them, because the aquifer under the city is falling and, near the bay, filling with seawater, and because a mall on its own well pays the utility nothing. The malls resist, the licences are renewed, and in the rationing of 2019 the malls were where the city went to wash. The pattern is the Jakarta and Bangkok one, where the ground sinks under the pumping, and Manila's northern districts, on the soft sediment of the delta, are subsiding by several centimetres a year, which the tide, on the days of a high water and a typhoon surge, has begun to exploit.

What it teaches

Manila is the city on this site that fixed its pipes and not its source. The privatisation of 1997, whatever its politics, cut the losses of the east zone from two thirds to a tenth and connected the districts the state had written off, and it could not build a dam, because a dam is a decision that only a state can make and this one took forty years to make it. The reservoir that supplies ninety seven percent of the city falls in every dry year, the rota follows, and the second source is being built in a forest whose people did not agree to it. The rain is not the problem. It never is.

Ninety seven percent from one dam, leaks cut from two thirds to a tenth, and a city turned off by rota in the month the reservoir ran low.

Sources

1. Metropolitan Waterworks and Sewerage System (MWSS), annual reports: Angat dam about 97 percent of raw water; allocation 4,000 million litres a day; Kaliwa dam project 600 million litres a day.
2. Manila Water and Maynilad, annual and sustainability reports: non revenue water from 63 percent in 1997 to under 10 percent (Manila Water, east zone); Maynilad from 66 percent to about 25 percent.
3. Asian Development Bank (2008). Manila Water concession: a case study; and World Bank Water and Sanitation Program reviews of the 1997 privatisation.
4. Philippine Senate and House hearings on the March 2019 water shortage; Manila Water reports on the La Mesa reservoir level and the Cardona treatment plant delay.
5. Commission on Human Rights of the Philippines and the Dumagat Remontado communities, statements on the Kaliwa dam; New Centennial Water Source project documents.
6. Photographs: opener: Makati Guadalupe-Poblacion skyline with Paco, Quirino (Manila; 12-23-2023) by Patrickroque01 (CC BY-SA) via Wikimedia Commons; inline: Angat Dam reservoir 05-2010 - Flickr by Albert Lozada (CC BY) via Wikimedia Commons; inline: 9848Effects floods of Typhoon Goni Telacsan Tacasan Macabebe 67 by Judgefloro (CC0) via Wikimedia Commons.

Lahore is a city of thirteen million people built on the bank of a river that no longer reaches it as a river, and it drinks entirely from the ground. The utility that supplies it runs more than five hundred deep wells sunk into the sand and silt of the Punjab plain, and private wells by the tens of thousands supplement them, and the water table under the city, which was a few metres below the surface when the wells were first sunk in the 1960s, has been falling by half a metre to a metre a year for three decades and is now forty to fifty metres down under the centre. The city sits on the largest irrigation system in the world, the canals of the Punjab, and has no surface water it can drink.

This article is about why, which is a story about a treaty, a river turned into a drain, an aquifer being spent, and a plan to build a new city where the river was.

The river that was given away

The Ravi rises in the Indian Himalaya and flows south west into Pakistan past Lahore, and until 1960 it was the city's river: the Mughal gardens were laid out on its bank, the old city's walls faced it, and its floods and its fish were part of the city's year. The Indus Waters Treaty, signed that year to settle the partition of the Punjab's rivers, gave the three eastern rivers, the Ravi, the Beas and the Sutlej, to India, and the three western ones to Pakistan, and India has since built the dams and barrages that take the Ravi's flow for its own canals before the border. What crosses into Pakistan, for most of the year, is a trickle, and what reaches Lahore is that trickle plus the city's own discharge.

The treaty is the most durable agreement the two countries have, and it has survived three wars. It also left Lahore, whose water had never depended on the river directly, with a channel beside it that carries, in the dry season, almost nothing but what the city puts in.

The river as a drain

The Ravi below Lahore is, by the surveys that have measured it, among the most polluted stretches of river in Asia. The city has sewers for most of its centre and treatment works for almost none of it, so that the sewage of thirteen million people goes to the river through a dozen outfalls untreated, along with the effluent of the tanneries, the textile mills, the paper plants and the chemical works of the industrial districts along the road to Sheikhupura, and the river in the dry season has no oxygen in it, the black colour and the smell that the Dhaka article on this site describes for the Buriganga, and a fish population of nothing. In the monsoon the floods from India flush it, and the sewage and the floods together spread over the plain.

A city that had a river to treat would build a works on it. Lahore has a drain to treat, and the utility's calculation for fifty years has been that the wells were cleaner and cheaper. That calculation is now changing at both ends.

What is in the wells

The aquifer under Lahore is the alluvium of the Indus plain, hundreds of metres of sand and silt laid down by the rivers, and it is a large and good aquifer that the city has been drawing on faster than the plain's rain and the leakage of the canals refill it. The fall of the water table is the plain arithmetic of the groundwater article on this site, and it has two consequences beyond the pumping cost. The first is the Ravi itself, which seeps into the ground along its bed, so that the wells nearest the river draw water that was, a few years earlier, the city's sewage, filtered through sand but carrying what sand does not stop. The second is arsenic. The sediments of the Indus plain, like those of the Bengal delta, carry arsenic that dissolves into groundwater under the conditions that heavy pumping creates, and surveys since the 2000s have found a share of Lahore's wells above the ten microgram limit that the arsenic article on this site describes, with the highest levels in the older, shallower wells near the river. The utility has been sinking its new wells deeper, to two hundred metres and more, to get below both, and the deep water is the water that does not refill.

The salt is the third. The Punjab's canals, spread over the plain for a century and a half, raised the water table under the fields and left salt in the soil as the water evaporated, and the shallow groundwater across much of the plain is brackish; the city's aquifer is fresher, and it is fresher because the city sits on a strip that the rivers recharged, which the treaty's diversions and the pumping are both reducing.

The plan

The city's plan, in two parts, is the plan of every groundwater city on this site. The first part is surface water: a canal offtake from the Indus system to the east of the city, and a treatment works, the city's first, to take a share of the supply off the wells. It has been in the utility's master plan since the 1990s, and the first plant is now in design, at a scale that would cover a fifth of the city's use. The second part is recharge: wells that take the monsoon floods, which flood the streets every summer, and send them into the aquifer instead of the Ravi, at a few sites so far and in the plan at hundreds. The plan's third part, which is the one that was announced first, is the Ravi Riverfront: a new city for six million people, to be built along forty kilometres of the riverbed and its floodplain, with the river itself to be turned into a chain of lakes fed by treated sewage and barrages. The courts stopped it in 2022 on the grounds that the land had been taken from farmers without due process, and the government has since restarted it in parts. The new city would need water, and the plan's answer is the same aquifer.

The politics of the river

Lahore's water is also a border question. The treaty that gave the Ravi to India is under strain: India has announced projects to use every drop of the eastern rivers before they cross, which the treaty permits, and Pakistan has objected to Indian dams on the western rivers, which it may not, and in 2025 India said it was holding the treaty in abeyance. The Ravi's flow into Pakistan, already a trickle, is the least of the rivers at stake and the one that matters most to one city. A river that was a city's for a thousand years is now a line in an agreement between two governments who are not speaking, and the city drinks from the ground while they decide.

The canal through the city

Lahore has one open water it can see, and it is the canal. The Bambanwala Ravi Bedian canal, dug in the 1860s to carry Ravi water to the fields east of the city, runs through the middle of Lahore's newer districts under a double row of trees, and it is the city's river in every sense but the one that matters: the boys swim in it in the summer, the roads run beside it, the housing estates are named for it, and the water in it is irrigation water, silt laden and allocated to farms, that the city has never treated for drinking and that the treaty's arithmetic does not spare for it. The plan for the city's first treatment works takes its water from this canal, at the eastern edge where it enters, and it will need an allocation that the farms downstream currently hold. The Punjab's irrigation system is the largest on Earth and every drop of it is spoken for, which is the other reason a city on a plain full of canals has drunk from the ground for sixty years.

What it teaches

Lahore is the city on this site whose water was decided by a treaty and spent by a pump. The river was given away in 1960 and turned into a drain by the city that lost it, the wells that replaced it have been falling for thirty years and are reaching arsenic and the river's own seepage, and the plan to fix both has been in a drawer since the 1990s while a plan to build a new city on the riverbed went ahead. The Punjab has more canals than any place on Earth. Lahore has none it can drink from.

A hundred percent from the ground, a metre a year, and a river that carries the city's sewage past the gardens that were built to face it.

Sources

1. Water and Sanitation Agency (WASA) Lahore, annual reports: about 550 tube wells; supply about 2 billion litres a day; groundwater the sole source; water table decline of 0.5 to 1 metre a year.
2. Indus Waters Treaty (1960), Annexure: the Ravi, Beas and Sutlej allocated to India; the Indus, Jhelum and Chenab to Pakistan.
3. Basharat, M. and Rizvi, S.A. (2011). Groundwater extraction and waste water disposal regulation: is Lahore aquifer at stake with as usual approach. Pakistan Engineering Congress.
4. Punjab Environmental Protection Department and WWF Pakistan, Ravi river water quality surveys: dissolved oxygen near zero below the city; arsenic in groundwater above 10 micrograms per litre in a share of wells.
5. Ravi Urban Development Authority, Ravi Riverfront Urban Development Project documents (2020 onward); Lahore High Court rulings 2022.
6. Photographs: opener: Badshahi Mosque in Lahore ,Pakistan 07 by Harvinder Chandigarh (CC BY) via Wikimedia Commons; inline: Ravi Lahore LRM by Hamzazahid04 (CC BY-SA) via Wikimedia Commons; inline: Canal of Lahore, Pakistan by S zillayali (CC BY-SA) via Wikimedia Commons.

A litre of olive oil carries, in the global averages this site uses, about fourteen thousand litres of water, which is the largest number for any cooking oil and more than a kilogram of beef. It is a number that, for most of the three thousand years the tree has been cultivated, described rain on a hillside where nothing else would grow, and it is a number that is becoming, in the country that makes half the world's oil, a draw on a river that has other claimants. The olive is the crop on this site that is moving from the wool article's kind of water to the almond article's, and this is about how.

The tree

The olive is an evergreen that grows where summers are hot and dry and winters are mild, on soils too thin and stony for a cereal, and it survives drought by shedding fruit rather than leaves and by roots that go deep. A traditional grove has a hundred trees to the hectare, planted by hand and left for a century, and yields two or three tonnes of fruit in a good year and much less in an off year, because the tree bears heavily in alternate years. A kilogram of fruit gives about two hundred grams of oil, and the water the tree took up over the year, which is the rain that fell on the hectare, is divided across those two hundred grams, which is why the oil's number is so much larger than the fruit's and why both are large: the yield of oil per hectare is a few hundred kilograms, against a tonne of sunflower oil or four of palm.

The water is also, for the old groves, almost all green. The rain of Andalusia, Puglia, Crete or the Peloponnese fell on the hillside, the tree used what it could and let the rest go, and nobody else could have had it.

Spain

Spain makes about forty five percent of the world's olive oil in a normal year, and most of that comes from Andalusia, and most of Andalusia's from the province of Jaén, where the olive covers more of the land than any crop covers any province in Europe: sixty million trees, in groves that run to every horizon. The old groves are rain fed and yield what the year gives. The groves planted since the 1990s, which are now about a third of the Spanish area and most of the new area in Portugal, Morocco, Tunisia and California, are different. They are hedgerows: dwarf trees at a thousand to two thousand to the hectare, drip irrigated, mechanically harvested by a machine that straddles the row, bearing in their third year and giving several times the oil per hectare of a traditional grove, at a cost of two to four thousand cubic metres of water per hectare per year, drawn from the Guadalquivir's reservoirs, from the canals of the basin authority, and from wells.

The water makes the yield. An irrigated grove gives ten to twelve tonnes of fruit a hectare against the old grove's two or three, and it gives them every year, and the oil is cheaper. It also gives the Guadalquivir basin, which the rice, the cotton, the strawberries and the cities of Seville and Córdoba already draw on, a new user, and the basin has been over allocated since the 1990s.

2022 and 2023

In 2022 Andalusia had the driest year in its record, and in 2023 a hot spring killed the flowers on the trees, and the Spanish harvest halved two years running, from about 1.4 million tonnes to about 660,000. The world price of olive oil tripled, the supermarkets of Spain put security tags on the bottles and locked the shelves, the theft of fruit from the groves at night became a police matter, and the Guadalquivir's reservoirs fell to a fifth, so that the basin authority cut the irrigation allocation to the groves by half and then by more. The irrigated groves, which had been the answer to the drought, had less water; the rain fed ones, which had never had any, had less fruit; and the price of the oil, which had never before been a matter of national politics, was on the front pages.

The rain returned in 2024, the harvest recovered, and the price fell back. The reservoirs did not refill, and the allocation to the groves remains a share of what they were planted for. The pattern is the one the almond article on this site describes for California: a crop planted on the assumption of water that a drought withdraws.

The mill

The mill's water is small and its effluent is not. Olives are washed, crushed to a paste, mixed, and spun in a centrifuge that separates oil from water and solids, and the traditional three phase mill added water to the paste and sent out, for every tonne of fruit, about a cubic metre of alpechín, the black vegetation water of the olive, with an oxygen demand of forty to a hundred grams a litre, a hundred times sewage, and polyphenols that kill the bacteria a treatment plant depends on. It was, for a century, poured into the rivers of every olive region in the Mediterranean in the months after the harvest, and the rivers of Jaén ran black. The two phase decanter, adopted across Spain from the 1990s, adds no water and sends out a wet pomace instead, which is dried and burned or extracted for its last oil, and the black rivers of Andalusia are largely a memory, though the mills of the eastern Mediterranean still run the old way.

Where the rest is grown

Italy, Greece, Tunisia, Turkey, Morocco and Portugal make most of the rest, and the pattern repeats: old groves on rain in the hills, new hedgerows on irrigation in the plains, and a rising blue share in the driest countries. Tunisia, the largest exporter after Spain, has planted hedgerows on the water of a coastal aquifer that the sea is entering. Morocco has planted a million hectares under a national plan and irrigates them from dams that its cities also drink from. California's groves, which supply the American market for oil that is not Spanish, are in the Central Valley, on the water the almond article describes. The tree of the dry hillside is being planted in the plains, on pipes, everywhere the oil is sold.

The oil in the bottle

A litre of extra virgin oil is the juice of about five to seven kilograms of olives, pressed cold, and the water in it is the water of those kilograms, which is the arithmetic in the table. The cheaper grades carry the same water and less of the tree's care: refined oil is what remains when the flawed oils are treated with heat and chemicals to remove the flavour, pomace oil is what a solvent takes from the pressed pulp, and the blends on the supermarket shelf are, on the tests that the regulators run, mislabelled at a rate that has made olive oil the most adulterated food in Europe. None of that changes the litres. The Tunisian oil sold in an Italian bottle carries Tunisian rain, and the Spanish hedgerow oil sold as a supermarket's own brand carries the Guadalquivir, whatever the label says about where it was bottled.

The tree in a warming climate

The olive is being planted in the plains because the plains have pipes, and it is also, in its old range, being stressed by a climate that has moved past what the hillsides gave it. The flower needs a cool winter to set and a mild spring to hold, and the springs of 2023 in Andalusia, with a week over thirty five degrees in May, killed the flowers on trees that had survived four hundred summers. The fly that ruins the fruit is spreading north, the drought years are more frequent, and the growers of Jaén, whose groves are too old to move and too steep to irrigate, are the ones the climate is testing first. The hedgerows in the plains, on a pipe from a reservoir, can be watered through a heatwave. The hillside, which was the point of the tree, cannot.

What it teaches

Olive oil is the product on this site whose water is changing kind. The old number, fourteen thousand litres of rain on a hillside that grew nothing else, was large and harmless; the new groves of Spain and the countries copying them turn a share of it into water from a river that has run dry twice in three years, to make an oil cheaper. The mill's black water has been fixed. The grove's blue water is the argument the Guadalquivir has not settled.

14,500 litres a litre, rain on a hillside for three thousand years, and a river in it now.

Sources

1. Mekonnen, M.M. and Hoekstra, A.Y. (2011). The green, blue and grey water footprint of crops and derived crop products. Olives about 3,000 litres per kilogram; olive oil about 14,400.
2. International Olive Council, world olive oil production and prices 2021 to 2024: Spain about 45 percent of world output in a normal year; Spanish production about 660,000 tonnes in 2022 and 2023 against 1.4 million typical.
3. Junta de Andalucía and Spanish Ministry of Agriculture, olive grove surveys: irrigated area, superintensive plantings, water allocations in the Guadalquivir basin.
4. Fernández, J.E. (2014). Understanding olive adaptation to abiotic stresses as a tool to increase crop performance. Environmental and Experimental Botany 103. Olive water needs and deficit irrigation.
5. European Commission, Best Available Techniques for the food, drink and milk industries: olive mill wastewater and the two phase decanter.
6. Photographs: opener: Roads crossing the Andalusian olive groves (49463059508) by Andrés Martín Rodríguez (CC BY-SA) via Wikimedia Commons; inline: Olivares en Archidona by Tyk (CC BY-SA) via Wikimedia Commons; inline: Some of my olives ready for oil at the mill. Drosia. Greece by Jebulon (CC0) via Wikimedia Commons.

The beer article on this site counts the water in a glass and finds it in the barley field: about three hundred litres for a litre of beer, almost all of it rain on a cereal crop in Europe, Canada or Australia. This article is about the building where the beer is made, which uses about three litres of water for every litre it sends out, and which is on this site for the reason every industry article is: the three litres are the water a town notices, and the water that leaves the plant is the kind a river notices most.

A brewery is a food factory that makes a sugar solution from grain, ferments it, and cleans everything, and its water goes into those three things in roughly equal measure.

Where the three litres go

About one litre of the three is in the beer, or was. Malted barley is mashed in hot water to dissolve its sugars, the sweet liquid is drained off, boiled with hops, cooled and fermented, and the water that went into the mash is the water in the glass, plus what evaporated in the kettle and what stayed in the spent grain. Another litre, roughly, is cleaning: every tank, pipe, filler and floor in the plant is washed between batches with hot water, caustic and acid, in the cleaning in place cycles that the pharmaceutical and food articles on this site describe, because beer is a growth medium for anything that reaches it and a brewery's enemy is a bacterium. The third litre is cooling and the rest: the wort has to be cooled from boiling to fermenting temperature in minutes, the fermenters and the cellars are chilled, the bottles and cans are pasteurised in hot water sprays and rinsed, and the boilers that make the plant's steam need feed water and blow down.

A generation ago the total was six or seven litres, and the fall to three came from the plain measures every industry article on this site lists: reusing the last rinse of one cycle as the first rinse of the next, recovering the cooling water, catching the pasteuriser's overflow, and metering every line so that the losses could be seen. The best plants in the world, in Europe, Latin America and Africa, are at two and a half or below, and the arithmetic below that gets hard, because a litre of beer needs a litre of water and the cleaning cannot be skipped.

What leaves the plant

The effluent of a brewery is sweet. It carries the spent yeast, the last of the wort from every tank, the beer lost in filling, the caustic and acid of the cleaning, and the sugar and protein dissolved from the grain, and its oxygen demand, by the BOD measure the article of that name explains, is about a thousand to fifteen hundred milligrams a litre, three to four times raw sewage, in a warm stream whose pH swings from the caustic rinse to the acid one within the hour. A sewage works can take it, and charges by strength to do so, and a brewery that sends its effluent to the sewer pays for its BOD at a rate that makes a treatment plant of its own worth building above a certain size. The larger breweries run anaerobic digesters, which turn the sugar into methane that fires the boilers, and the largest run the full sequence the effluent article describes and discharge to a river.

The solids are worth more than the water. Spent grain, the husks and protein left after the mash, is about twenty kilograms for every hundred litres of beer, and it goes to cattle feed wet or is dried for the same; the spent yeast goes to yeast extract, which is the flavour in a jar of savoury spread, and to animal feed; and the hop residue goes to compost. A brewery that sends its grain and yeast out as product rather than down the drain has removed most of the oxygen demand from its water before treating any of it.

Where breweries are

A brewery is built near its market, because beer is mostly water and water is heavy to ship, and the largest brewers have plants in every large city they sell in. That puts breweries in every dry city on this site: in Mexico's northern deserts, in Johannesburg, Nairobi and Lagos, in Chennai and Bangalore, in Phoenix and Las Vegas. A large plant uses a few million litres of water a day, which is the water of a town of twenty or thirty thousand people, and it draws it, in most of the world, from the town's supply or from its own wells in the same aquifer. In an ordinary year that is a large customer. In a drought it is a brewery making beer while the suburbs are rationed, which is a story a local newspaper writes in an afternoon.

The brewers know this, which is why the three largest in the world publish their water ratios plant by plant and set lower targets for the plants in stressed basins, and why they have paid, in Mexico, India and South Africa, for the watershed work, the leak repairs and the rainwater schemes that keep the town's supply and their licence intact. It is the cheapest insurance in the industry.

Mexicali

In 2016 an American drinks company began building a brewery at Mexicali, in the desert of Baja California on the United States border, to make beer for export north, with a permitted draw of about twenty million litres a day from the aquifer that the city and the irrigated farms of the Mexicali valley also use. The valley's water comes from the Colorado, which the Las Vegas and Phoenix articles on this site describe as a river already spoken for, and the farmers and the city's activists argued for four years that the brewery's water would come out of theirs. In March 2020, with the plant about seventy percent built, the federal government held a public vote, and seventy six percent of those who voted said no. The permits were cancelled, the company wrote off the plant and moved the project to a wetter state in the south, and the tanks stood in the desert.

The lesson the industry drew was the one this article is about. The brewery's water was three litres a litre, a modest figure by the standards of this site, and it was the figure that decided whether the plant could exist, because it was the town's water and the town could see it.

The craft brewery

The small brewery, which has multiplied across every rich country in the last twenty years, uses more water per litre than the large one, often five to ten, because the equipment is small, the cleaning is the same, the batches are short and the losses between them are a larger share. A brewery making a few thousand litres a week is a small user in a town and a large one in a village, and its effluent goes to the sewer at a strength the works notices, which is why the craft brewers' associations publish water guides and why the local utility, in a town with a dozen small breweries, has begun to charge them by strength. The best small brewers have learned the large ones' habits: the last rinse reused, the spent grain to a farmer, the yeast to a baker, and a meter on the incoming main so that the ratio is known. The ones that have not are at eight litres a litre and do not know it.

The water in the beer

The water that goes into the beer is chosen, and brewers have always known that the town's water made the town's beer: the hard, sulphate rich water of Burton made pale ale, the soft water of Pilsen made the lager that copied it, and the carbonate water of Dublin made stout. A modern brewery takes the town's water, strips it by reverse osmosis to almost nothing, and adds back the salts that the style needs, so that a lager brewed in Lagos can match the one brewed in Amsterdam, and the reverse osmosis rejects a quarter of what it takes in, which is part of the three litres. The water article that the beer is famous for, in other words, is now a recipe rather than a place.

What it teaches

Breweries are the industry on this site where the plant's water is a hundredth of the product's and the whole of the argument. The barley's three hundred litres are rain on a field somewhere else; the plant's three are the town's, drawn from its supply in the middle of it, and what leaves is a sweet warm stream that the town's works has to treat. The brewers have cut the three from six, count it plant by plant, and pay for the town's leaks, because Mexicali showed what happens to a plant that does not.

Three litres a litre, three hundred in the barley, and a vote in the desert that stopped a brewery three quarters built.

Sources

1. Beverage Industry Environmental Roundtable (BIER), Water Use Benchmarking in the Beverage Industry, 2023: brewery water use ratio median about 3.0 litres per litre; best in class about 2.5.
2. AB InBev, Heineken and Carlsberg sustainability reports 2023: water use ratios of 2.5 to 3.2 litres per litre; targets at water stressed sites.
3. European Commission, Best Available Techniques for the food, drink and milk industries: brewery water use 3 to 10 hectolitres per hectolitre; effluent BOD 1,000 to 1,500 mg/L.
4. Government of Mexico, Consulta ciudadana de Mexicali, March 2020: 76.1 percent against the Constellation Brands brewery; the plant's permitted draw about 20 million litres a day.
5. Mekonnen, M.M. and Hoekstra, A.Y. (2011). Water footprint of crops and derived crop products: beer about 300 litres per litre, almost all in the barley.
6. Photographs: opener: Tanks at Green Leaf brewery by Ruth Hartnup (CC BY) via Wikimedia Commons; inline: Budweiser Budvar brewery, tanks 03 by Czeva (CC BY-SA) via Wikimedia Commons; inline: Amanecer en Mexicali by B.jars (CC BY-SA) via Wikimedia Commons.

The bottled water article on this site is about the myth that the water in the bottle is better than the water in the tap, and about what the bottle costs. This article is about the building where the bottle is filled, which is a well with a filling line on top of it, and which is, by volume, one of the smallest industrial water users on this site and, by the fights it has caused, one of the largest. A bottling plant pumps a spring or an aquifer, filters it, fills it, and trucks it away, and the trucking is what makes it different from every other user in the town: the water leaves the watershed and does not come back.

What the plant does

Water arrives at the plant from a borehole, a spring that has been tapped with a pipe, or, for the largest brands in the world, the municipal main, which is treated again and sold as purified water. It is filtered, usually through sand and carbon; it is passed under ultraviolet light or dosed with ozone to kill whatever it carries, because a sealed bottle of untreated water on a warm shelf grows things; and in the case of spring and mineral waters it is otherwise left as it came, because the label says so and the law in most countries requires it. The bottles are blown on site from small plastic preforms, filled, capped, labelled and stacked on pallets, and a large line does tens of thousands of bottles an hour.

The plant's own water use, beyond the product, is the rinsing of the bottles, the cleaning of the line and the cooling of the blow moulders, and the industry's own benchmark puts the total at about 1.3 litres of water for every litre sold. It is among the most efficient ratios of any beverage, because the product is the water and the process is short. The plastic is the other footprint, and the bottle's several litres of water and its carbon are counted in the bottled water article.

Why the fights are in small towns

A few hundred million litres a year is the water of a town of a few thousand people, and it is drawn, in the case of a spring water brand, from a spring in the countryside, because that is where springs are, and the countryside is where the fight happens. The plant is usually the largest single water user in a rural township, its permit is usually granted by a state or provincial agency far away, the water leaves and the trucks stay, and the people who live over the spring see a stream fall in a dry summer and a lorry every ten minutes. Whether the stream fell because of the plant is the question every one of these disputes turns on, and it is, hydrologically, a hard one, because a spring's flow varies with the rain and a well's effect on it is a matter of modelling. The company's model says little effect. The township's says otherwise. The courts, in most cases, have found for the permit.

The same pattern has run in Ontario, in Oregon, in Florida, in the south of France, in the mountains of Pakistan and in the hills of Maharashtra, and the arguments are the same. The plant pays little or nothing for the water, because water in the ground, in most of the world's law, belongs to the landowner or to nobody. It pays for the permit, the land, the plant and the trucks. What it sells is the water.

Michigan

The case that became the emblem ran in the state that sits on the Great Lakes, a fifth of the world's fresh surface water, and it ran for twenty years. A Swiss company's American arm bought a spring in the sandy country of central Michigan in 2000, built a plant, and pumped it for a brand of spring water sold across the midwest, at a rate that a state permit set at a few hundred gallons a minute. A citizens' group sued over the effect on the stream and the wetland below the spring and won a reduction. In 2016 the company applied to raise the rate to four hundred gallons a minute, about two million litres a day, and the state granted it in 2018, over the objections of most of the eighty thousand people who wrote in, for an annual paperwork fee of two hundred dollars. The township whose zoning the company needed for a booster station refused it, the company sued, and the case went through the state's courts until 2020, when the company sold its American spring brands to a private equity firm that took the permit with them.

Two hundred dollars a year for two million litres a day became the figure the argument was about, and it was the right figure to argue about, because it named the thing the permit had done: converted a public resource into a private product at a price of nothing. The state's answer, that a farmer or a factory pumping the same water would pay the same nothing, was true and was not an answer. The farmer's water comes back. The lorry's does not.

Fiji

The bottle of water with the square shape and the flower on the label is pumped from an aquifer on the largest island of Fiji, at a plant owned by an American couple, and shipped nine thousand kilometres to the shelves of Los Angeles and London, which makes it the product on this site with the longest journey for the least reason. It is also the plant that showed what a country can do. In 2010 the Fijian government imposed a levy on water extracted for export, of a few cents a litre, aimed at the one company that extracted it; the company closed the plant and said it would leave; the government said the aquifer would be there when it did; and the company reopened within days and paid. The levy became a line in the country's budget. The island's villages, near the plant, drank from wells and standpipes that the plant's water, which was too valuable to sell locally, passed on its way to the port.

California

The other long case is in the mountains above San Bernardino, where a brand has been drawing spring water from a national forest since the 1920s under a federal permit that expired in 1988 and was renewed, without review, by the annual payment of the fee. The drought of 2015 put the arrangement in the newspapers, the state's water board investigated, and in 2023 it ordered the company to stop taking most of the water on the grounds that it had no right to it under state law. The company is appealing. The spring is in a forest that burned in 2020, and the creek below it, which the forest's fish and the town's wells depend on, has been the subject of the same modelling argument as the Michigan stream, with the same result: nobody can prove what the spring would do without the pipe, and the pipe has been there a century.

The purified brands

The two largest bottled water brands in the world, and most of the volume in the developing world, have no spring behind them. They are municipal water, bought from the town's main at the town's tariff, put through reverse osmosis and ozone at a plant in an industrial estate, dosed with a pinch of minerals for taste, and bottled, and their footprint is the plant's 1.3 litres plus the reverse osmosis reject, which puts them nearer two litres of town water per litre sold. In India, Indonesia, Mexico and Nigeria, where the tap is not trusted and the Lagos article on this site describes the sachet as the poor's supply, these brands and their thousands of local imitators are a parallel water utility run for profit from the public one, and the argument about them is the bottled water article's argument in its sharpest form: a city that cannot make its tap trusted pays a bottler to make the same water trusted at a hundred times the price.

The plant's other water

The plant's second product is the bottle, and the bottle is water too. A polyethylene terephthalate bottle is blown on site from a preform, and the preform was made from resin, and the resin from oil and gas at a petrochemical plant that used, for cooling and process, several litres of water for every kilogram of resin, so that the bottle carries a few litres of water of its own before it is filled. The bottled water article counts that. The plant that fills it counts only what passes through its meter, which is the industry's 1.3, and both numbers are true.

What it teaches

The bottling plant is the industry on this site with the least water and the largest argument, and the argument is the right one. Its volume is a town's, its ratio is among the best of any beverage, and its product leaves the watershed by road, paid for at the rate of the permit paperwork, which in Michigan was two hundred dollars a year. The bottled water article says the water in the bottle was never the point. The plant is where that becomes visible, one spring at a time.

Two hundred dollars a year for two million litres a day, and a lorry every ten minutes taking it away.

Sources

1. Michigan Department of Environment, Great Lakes and Energy, Permit 1701 (2018): Nestlé Waters North America, White Pine Springs well, 400 gallons per minute; annual fee $200; Osceola Township litigation 2017 to 2020.
2. International Bottled Water Association, Water and Energy Use Benchmarking Study (2013 onward): about 1.32 litres of water used per litre of bottled water, including the product.
3. Beverage Marketing Corporation, global bottled water volumes: about 500 billion litres a year.
4. Fiji Water and the Government of Fiji, 2010 export levy dispute; Yaqara aquifer, Viti Levu.
5. California State Water Resources Control Board, Arrowhead (BlueTriton) Strawberry Canyon cease and desist order, 2023; US Forest Service permit review, San Bernardino National Forest.
6. Photographs: opener: Kangso Yaksu Mineral Water Bottling Plant (11360552806) by Clay Gilliland (CC BY-SA) via Wikimedia Commons; inline: USMC-090428-M-2360J-002 by US Marine Corps (Public domain) via Wikimedia Commons; inline: Fiji Waters bottles at an airport by Jon Roig (CC BY) via Wikimedia Commons.

Half the households in England pay for water the way most of the world's cities did in 1900: a fixed charge, set by the size or the value of the house, whatever comes out of the tap. The other half have a meter, and the difference between the two halves, measured in every trial since the 1980s, is about fifteen percent: the metered house uses less, without being asked to and without noticing much, because a bill that moves with the tap is a reason the fixed charge never gave. A water meter is the cheapest water a utility can buy, and it is on this site because the arithmetic of it is plain and the politics are not.

This article is about what a meter does, why the number falls, how a tariff can be shaped once there is one, what a smart meter finds that a plain one cannot, and what happened in the one country that tried to introduce charges and then abolished them.

Why the number falls

The trials that England ran on the Isle of Wight and in eleven other districts in the early 1990s, fitting meters to whole towns and watching, found the fall almost at once and found it persisted: about eleven percent in the first years, and in the universal metering programme that one southern company ran across its area between 2010 and 2015, about sixteen. The fall comes from the things a household does when it can see the number: a dripping tap fixed, a hose used less, a cistern's leak noticed, and, above all, a garden watered less, because the garden is the use that varies most and the one the meter charges for in the driest month. It also comes from a second thing the trials found, which is that a share of the houses had been leaking, on their own side of the boundary, for years, and the meter's first bill was the first anyone knew.

A metered household in England now uses about a hundred and thirty litres a person a day, and an unmetered one about a hundred and seventy, and the difference is what the country's water strategy counts on to get through the droughts the Sydney and Melbourne articles on this site describe.

The bill without a meter

An unmetered bill in England is based on the rateable value of the house, a figure fixed in 1990 for a tax that no longer exists, so that the household of one in a large old house pays more than the household of six in a small new one, and neither has any reason to turn off a tap. It is the arrangement most of the world's cities began with, because the meter was a cost and the water was cheap, and it is the arrangement that many of the poorest still have, in the form of a flat monthly charge or no charge at all, with the consequence the leaks article on this site describes: a utility that cannot see where its water goes cannot find where it is lost. The cities of this site with the highest losses, Nairobi, Karachi, Lagos and Johannesburg, are the ones with the fewest working meters, and the utilities that turned their losses round, in Manila, Phnom Penh and Singapore, began by metering everything.

The tariff

Once there is a meter, the price can be shaped. The rising block tariff, which most of the dry cities of this site use, charges the first block of water, enough for drinking, cooking and washing, at a low price or none, and each block above it at a higher one, so that a household's basic needs are cheap and its pool is not. Cape Town's tariff in the Day Zero year rose by a factor of ten between the first block and the last; Singapore's charges a conservation tax that rises with use; Las Vegas pays people to take out lawns because a tariff alone did not. The tariff's difficulty is the large family in the small house, whose basic block is too small, and the answer is a block sized by the number of people, which needs the utility to know how many there are, which is a different argument.

The other difficulty is the fixed cost. A water network costs almost the same to run whether the customers use a lot or a little, and a utility whose revenue follows the meter finds, in a wet year or after a campaign, that its income has fallen with its sales, and raises the price per litre to cover it, which the customers who saved water read as a punishment. Most tariffs therefore have a standing charge for the pipe and a volume charge for the water, and the argument is about the split.

The smart meter

A plain meter is read twice a year, by a person with a torch, and it tells the utility what a household used in six months. A smart meter reads itself every hour and sends the reading over the air, and it tells the utility three things the plain one cannot. It shows a household whose use has never fallen to zero, at four in the morning, for a week, which is a leak on the customer's pipe under the floor or the drive, and the utility can write to the customer before the bill does. It shows the district's total, hour by hour, so that a burst in the main is found the same night rather than when the road subsides. And it shows the customer, on a screen or a phone, what the tap is doing, which is the fifteen percent effect made continuous. The English programmes that have fitted smart meters have found customer side leaks in a share of houses that surprised the utilities and account, across the country, for about a quarter of all the leakage the leaks article counts.

The objections are the ones every smart meter raises: the data, which shows when a house is empty; the cost, which is the customer's in the end; and the bill, which finds the leak and charges for the water it lost.

Ireland

Ireland had no domestic water charges from 1997, paid for its water from general taxation, and lost about half of its supply to leaks. In 2014 the government created a national utility, fitted meters to most of the country's houses, and began to charge, at rates that were low by European standards and were, in the aftermath of the financial crisis and the austerity that followed it, the last straw for a large share of the public. The protests were the largest in the country in a generation, tens of thousands in the streets of Dublin, the meters were pulled out of the ground in some estates, the charges were suspended in 2016 and abolished for ordinary use in 2017, and the meters that had been fitted, a million of them, were left in the ground, reading. The utility uses them to find leaks. It cannot charge for the water they measure, except above a threshold that catches only the largest users. Ireland's per capita use is among the highest in Europe.

The lesson is the one the tariff section makes and the politics of this site repeat: a meter is a measuring device, a charge is a decision, and the second needs the consent that the first does not.

The meter in a dry city

The meter is also what a rota is made of. The cities on this site that have rationed by district, Nairobi, Johannesburg, Manila and Cape Town, could do so only where the meters and the valves let them see and control what each district used, and Cape Town's Day Zero campaign, which cut the city's use by half, ran on a weekly map of every household's metered use, published street by street, with the heavy users named. Bangalore and Karachi, with meters on a minority of connections, ration by tanker and by pressure and cannot tell who is using what. The meter is the difference between a city that can ask a district to save a tenth and a city that can only turn the district off.

The tanker economy that the Nairobi and Lagos articles describe is, in the end, a metering failure: the vendor sells by the jerry can, at a price per litre that is the truest meter in the city, to people the utility never fitted with one.

What it teaches

The water meter is the plain water article about the simplest piece of equipment on this site and the one that saves the most water for the least money. Fitted to a house, it cuts the use by a seventh without a word; read hourly, it finds the leaks that the utility could not see; shaped into a tariff, it makes the pool pay and the kettle not. Half of England has one. Ireland has a million in the ground that cannot send a bill. The water a meter saves is the cheapest water there is, and whether anyone is allowed to charge for it is the question the meter cannot answer.

Fifteen percent, without asking, and a million meters in Ireland reading water nobody pays for.

Sources

1. Ofwat and Environment Agency, Metering and household consumption: the Isle of Wight and National Metering Trials (1989 to 1993), about 11 percent reduction; Southern Water universal metering programme (2010 to 2015), about 16 percent.
2. Water UK and Discover Water, household metering in England and Wales: about 60 percent of households metered in 2024; per capita consumption metered about 130 litres a day against unmetered about 170.
3. Thames Water and Anglian Water smart metering programmes: customer side leakage detection; about a quarter of total leakage on the customer's pipe.
4. Irish Water and Government of Ireland, domestic water charges 2015 to 2016 and their suspension; the Water Services Act 2017.
5. OECD (2010). Pricing Water Resources and Water and Sanitation Services. Rising block tariffs and their effects on consumption and equity.
6. Photographs: opener: Two blue water meters on a matte red painted brick facade with pipes in Seoul by Basile Morin (CC BY-SA) via Wikimedia Commons; inline: Water meters in india by Nizil Shah (CC BY-SA) via Wikimedia Commons; inline: CCWW Water Meter by Axxter99 (CC BY-SA) via Wikimedia Commons.

The plain water pillar of this site is a description of what stands between a river and a glass: the works, the chemistry, the tests, the pipes, the residual chlorine and the inspector. About forty three million Americans, one in eight, and a share of the population of every country, drink from a well in the yard to which none of it applies. The drinking water law of the United States, and of most countries, regulates the public supply, and a well serving one house is the owner's business: nobody treats it, nobody tests it unless the owner pays for a test, and nobody tells the owner what the test found unless the owner asks. The water tastes fine, as it has for years, which is what most of the things that are wrong with it taste like.

This article is about what a private well is and why it sits outside the rules, what the national surveys found when they looked, what the owner is expected to do and how rarely it is done, and why the water people trust most is the water nobody checks.

What a private well is

A private well is a hole drilled, dug or driven into an aquifer under a house, with a pump and a pipe to the taps, and it serves one household or a few. It is the water supply of the countryside everywhere: the farms and villages that the mains never reached, the houses on the edge of towns that were built before the pipe arrived, and, in the developing world, most rural households, with a hand pump on a tube well of the kind the arsenic and Dhaka articles on this site describe. The water is groundwater, and the groundwater article explains why that is, in most places, the cleanest water there is, filtered through metres of sand and needing no treatment, and also why it carries what the rock and the fields above it put in.

The rules stop at the property line. The American drinking water law of 1974 applies to systems serving more than twenty five people, and the rest, some fifteen million wells, are regulated by nobody at the federal level and by a patchwork at the state one, which usually amounts to a permit to drill and a test at the time of drilling, and nothing after. Britain regulates its private supplies through the councils, which are required to sample them every few years and mostly do; most of the world does not regulate them at all.

What the surveys found

The national geological survey of the United States sampled about two thousand private wells across the country between 1991 and 2004 and published the results in 2009, and they are the best picture there is. About one well in five had at least one contaminant above a health limit. Nitrate, from the fertiliser and the manure of the fields above, was the commonest in the farming regions of the midwest and the plains; arsenic, from the rock, in the west and the northeast; uranium and radon, from granite and its cousins, in New England, the Rockies and the Piedmont; fluoride in a few districts; and manganese, which is a nuisance in most places and a concern in a few. Bacteria, which the survey tested separately, were in about a third of the wells at some level, and E. coli in a small share, most often where a septic tank drained too close to the well or the well's cap and casing had failed. The pesticides and the solvents that the survey looked for were found often and above a limit rarely.

The owners, when asked, had mostly never tested their water, and of those who had, most had tested once, at the time of drilling or the sale of the house, for bacteria alone. The survey's authors made the point that this site keeps making about groundwater: the water was clear, cold, and tasted of nothing, and had been drunk for decades, and arsenic at three times the limit tastes of nothing too.

The owner as inspector

The system depends on the owner, and the owner, in the studies of what owners do, tests the well when it is drilled, when the house is sold, when the water changes colour or smell, or when a neighbour finds something. The state health departments' advice is to test for bacteria and nitrate every year and for arsenic, radon and the rest at least once, at a cost of a few tens to a few hundred dollars, and the surveys of behaviour find that a small minority follow it. A state that tried to change this, New Jersey, passed a law in 2002 requiring a well to be tested at every sale of a house, and the results, in the study that followed, were both the discovery of arsenic and nitrate in tens of thousands of wells whose owners had not known and the finding that a share of those owners, told their water was over the limit, installed no treatment, because the treatment cost money and the water had always been fine.

The treatment, when installed, is the household version of the plain water pillar: a chlorinator or an ultraviolet lamp for bacteria, a reverse osmosis unit under the sink for arsenic and nitrate, an ion exchange bed for uranium, an aeration system for radon, and a softener for the hardness that makes most owners think about their water in the first place. The filter jug article on this site describes what does not work. A well owner with a test result has, for the first time, the information to buy the thing that does.

Rural Britain, rural India

England's private supplies, about a percent of the population on farms and in the remote villages, are sampled by the councils, and the failure rate of the samples is several times that of the public supply: around five to ten percent of samples fail, most for coliform bacteria and E. coli from a spring or a shallow well with livestock above it, some for nitrate, a few for lead from the house's own pipes. The council writes to the owner, and the owner may or may not act. In rural India and Bangladesh, where the arsenic article on this site began, the tube wells were dug by the million in the 1970s and 1980s to get people off the surface water that carried cholera, and it worked, and the wells carried arsenic that nobody tested for until the 1990s. The testing and marking of wells, red for unsafe and green for safe, has since covered most of Bangladesh, and the story is the private well's story at national scale: the water that was cleaner than the river, drunk for twenty years, and never checked.

The well and the septic tank

The private well's commonest companion is the private sewer, and the two are usually a few metres apart in the same garden. A septic tank takes the house's sewage, settles the solids and drains the liquid into a soakaway, from which it moves through the soil toward the nearest groundwater, which is the water the well draws. The rules that exist say the well must be a certain distance uphill of the soakaway, and the wells that fail for bacteria are, in most surveys, the ones where the rule was not followed or the soil was too thin or too cracked to filter what the tank sent. The nitrate in a well under a suburb of large gardens is, in the surveys that have traced it, as often from the neighbours' septic tanks as from any farm. A house on a well and a tank is a small water utility and a small sewage works with nothing between them but the ground, which is the arrangement that the plain water pillar of this site exists to replace, and it serves a hundred million people in the rich world alone.

What a test costs

A basic test, for bacteria and nitrate, costs a few tens of dollars at a state laboratory, and the full panel that the survey ran, with arsenic, uranium, radon, fluoride, metals, pesticides and solvents, costs a few hundred. The treatment that a bad result calls for costs from a few hundred for an ultraviolet lamp to a few thousand for a reverse osmosis system with its own maintenance, and the studies of what owners do with a bad result find that the second cost, more than the first, is where the process stops. A state that wanted to change the outcome would pay for the test, as some now do, and for the treatment in the households that cannot, as almost none do, and would find, as New Jersey did, that the map of its private wells has more red on it than anyone had expected.

What it teaches

The private well is the plain water article about water with nothing between it and the glass. It is, most of the time and in most places, the cleanest water on this site, and one well in five, when a survey finally looked, was over a limit for something the owner could not taste. The rules stop at the property line, the owner is the inspector, and the inspector tests once, when the house is sold, for the one thing that can be smelled. The whole of the plain water pillar exists because a public supply is tested every day. Forty three million people are drinking from a supply that is tested when they remember.

Forty three million people, one in five wells over a limit, and water that tasted fine for twenty years.

Sources

1. US Geological Survey (2009). Quality of water from domestic wells in principal aquifers of the United States, 1991 to 2004. Circular 1332. About 23 percent of 2,100 wells with at least one contaminant above a health benchmark.
2. US EPA and CDC, private drinking water wells: about 43 million people; not regulated under the Safe Drinking Water Act; testing guidance.
3. UK Drinking Water Inspectorate, Private water supplies in England, annual reports: about 1 percent of the population; failure rate of samples around 5 to 10 percent, mostly E. coli and coliforms.
4. Ward, M.H. et al. (2018). Drinking water nitrate and human health: an updated review. International Journal of Environmental Research and Public Health 15. Private wells in agricultural areas.
5. Flanagan, S.V. et al. (2016). Arsenic in private well water part 1 of 3: impact of the New Jersey Private Well Testing Act on household testing and mitigation behavior. Science of the Total Environment 562.
6. Photographs: opener: Rural Hand Pump by Prasoon Panthayi (CC BY) via Wikimedia Commons; inline: Drilling for clean water (7534901370) by Oxfam East Africa (CC BY) via Wikimedia Commons; inline: EPA GULF BREEZE LABORATORY, CHEMISTRY LAB. THE CHEMIST IS TESTING WATER SAMPLES FOR PESTICIDES - NARA - 546277 by William C. Shrout (Public domain) via Wikimedia Commons.

The claim runs like this: boiling water drives off some of it as steam, so whatever was dissolved in the water becomes more concentrated, so a kettle that is boiled twice holds water with twice the arsenic, twice the nitrate and twice the fluoride, and those things are harmful, so never reboil a kettle. Each step in the chain is true in the way a myth needs it to be, and the arithmetic of the whole is wrong by a factor of about twenty. This article is about what boiling actually does to water, what a kettle loses to steam, why reboiled water tastes different and why that is a different question, and what the small true part of the warning is.

What a boil takes away

A kettle brought to the boil and switched off has turned a small share of its water into steam. The measured figure, from the energy tests that look at how much people overfill kettles, is between two and six percent of the volume per boil, more for a kettle left rolling, less for one that clicks off at the first boil. Take five percent. The water that remains holds everything that was dissolved in the original volume, so the concentration of every dissolved thing has risen by about five percent: arsenic at ten micrograms a litre becomes ten and a half, nitrate at twenty milligrams becomes twenty one. A second boil does it again, to about eleven and twenty two. To double the concentration of anything in the kettle, the kettle would have to be boiled until half of it was gone, which takes an hour on a hob with the lid off and does not happen in a kitchen by accident.

The limits, meanwhile, are set with margins. The guideline for arsenic is ten micrograms a litre, set at a level where the lifetime risk is judged acceptable and well below the level at which the arsenic article on this site describes harm; the limit for nitrate, fifty milligrams a litre, was set from the blue baby cases at levels several times higher; and fluoride at one and a half milligrams a litre is a tenth of the level that causes the bone disease. A tap water that meets the standard, boiled twice, or ten times, is a tap water that meets the standard.

Water that does not meet the standard

The one case in which the warning has a point is water that was over the limit to begin with. A well with arsenic at fifty micrograms a litre, which the wells of Bangladesh and West Bengal in the arsenic article on this site carry, is dangerous cold, boiled once and boiled twice, and boiling makes it very slightly worse and, more to the point, does nothing to make it better, because boiling removes nothing that is dissolved. The same is true of nitrate in a farm well and fluoride in a Rift Valley borehole. The boil water notice that a utility issues after a burst main is about bacteria, which a minute of boiling kills, and it is not about chemistry, which a minute of boiling concentrates by a few percent and otherwise leaves alone. The myth's error is to take a warning that belongs to contaminated water and apply it to a kettle full of tap water that passed fifty tests.

Why the tea is flat

Reboiled water does taste different, and the reason is the gas. Cold tap water holds dissolved oxygen and carbon dioxide, a boil drives most of them off, and a second boil drives off the rest, and tea made with water that has no dissolved gas in it tastes flat, in a way that tea drinkers notice and that the chemists who have looked at it attribute to the gases' effect on how the tea's compounds extract and on the sensation in the mouth. The advice to use freshly drawn water for tea is old, and it is right, and it is about taste. The tea trade's own guidance, which has run tests on it, puts the difference down to the oxygen and to the water's temperature on the leaf rather than to anything that would show up in a laboratory's analysis of what the water contains, because nothing in that analysis changes with the boil except the gas. It has been borrowed by the myth as evidence of harm, and flat tea is not a harm.

Boiling also drives off the chlorine, which is why boiled and cooled water tastes better to most people than the tap, and why a jug of it in the fridge is the answer the filter jug article on this site gives to the household that dislikes the taste. The chlorine leaving is a loss of the disinfectant, which matters in a bottle kept for days and not in a cup drunk now.

What does build up

Two things do build up in a kettle, and neither is dissolved. The first is scale: the calcium and magnesium of hard water, which the hard water article on this site describes, come out of solution when the water is heated and coat the element and the walls in a white crust that reduces the kettle's efficiency and flakes into the cup. It is chalk, it is harmless to drink, and a kettle in a hard water district that is never descaled boils slowly rather than dangerously. The second is the kettle left with a little water in the bottom for days, in a warm kitchen, with the lid closed, which is a small version of the still warm water that the Legionella and filter jug articles describe: a place where the bacteria that live in any water grow to numbers that the next boil will kill anyway. Empty the kettle between uses, and the question of reboiling does not arise. The kettle that is worth worrying about is the one nobody has descaled in a year, in a hard water district, that takes four minutes to boil a cup: it is wasting electricity rather than concentrating anything, and a bag of citric acid fixes it in an afternoon.

Where the myth came from

The claim appeared on lifestyle websites around 2015, in articles that named arsenic, nitrate and fluoride because those are the three chemical contaminants that the reader has heard of, and it spread because it had the shape of a warning and cost nothing to pass on. It has since been repeated by television doctors, by appliance makers selling kettles with a fresh water function, and by well meaning relatives. The fact checkers who have looked at it found no study behind it and no case of harm, and the water agencies' own advice on boiling, which is that it kills organisms and does not remove chemicals, has been on their websites since before the claim was made. The myth persists because it is easy to believe that doing something twice makes it more, which is the same mistake that the bath and shower myth and the dishwasher myth on this site turn on, in the other direction.

The energy in the kettle

The one thing reboiling reliably wastes is electricity. A kettle boiled twice has used the energy of two boils, and the second was for water that was already hot enough to make tea a minute ago, and the household that fills the kettle to the top for one cup and boils it again for the next is spending, over a year, the energy that the dishwasher article on this site counts as the larger half of a hot water appliance's footprint. The advice that saves energy and the advice that makes better tea are the same: fill the kettle with what you need, from the cold tap, and boil it once. The myth, which tells people to pour out the water and start again for their health, sends them to the same place for the wrong reason, and a myth that happens to give good advice is still a myth.

What it teaches

The reboiling myth is the smallest on this site and it is a lesson in arithmetic. A boil loses five percent of the water and raises whatever is dissolved by five percent, against limits set with margins of tens and hundreds, and the water that would be dangerous reboiled was dangerous cold. Flat tea is real and is about gas. Scale is real and is chalk. A kettle emptied between uses, descaled when it slows, and filled from the cold tap, is safe to boil as many times as the tea requires. The energy is the only thing the second boil costs, and the taste is the only thing it changes.

Five percent to steam, five percent more of everything in the rest, and a limit that a hundred boils would not reach.

Sources

1. WHO (2022). Guidelines for Drinking Water Quality: arsenic 10 micrograms per litre; nitrate 50 milligrams per litre; fluoride 1.5 milligrams per litre; derivation of the guideline values and their margins.
2. US EPA and UK DWI, boil water advisories: one minute at a rolling boil kills bacteria, viruses and protozoa; boiling does not remove chemical contaminants.
3. Evaporative loss in domestic kettles: measured losses of 2 to 6 percent of volume per boil in energy efficiency testing (Energy Saving Trust, kettle overfilling studies).
4. Dissolved oxygen and carbon dioxide loss on boiling and the effect on tea flavour: UK Tea Academy and Royal Society of Chemistry, How to make a perfect cup of tea (2003).
5. Snopes and Full Fact, reviews of the reboiling claim (2015 onward).
6. Photographs: opener: Kettle (4141300920) by Lee Haywood (CC BY-SA) via Wikimedia Commons; inline: Waterboiler-internal by Henna (CC BY-SA) via Wikimedia Commons; inline: Pouring Tea. by Mizzyjo (CC BY-SA) via Wikimedia Commons.

A salmon lives in water and its water footprint is on land. The fish in a cage in a Norwegian fjord swims in seawater that costs nothing and is used up by nobody, and the two thousand or so litres that the footprint method assigns to a kilogram of it are the rain and irrigation that grew the soy, the wheat and the rapeseed in its feed, on fields in Brazil, Europe and Canada that the fish has never seen. Salmon is the animal on this site whose water is furthest from the animal, and the reason it is here is that the distance is instructive.

This article is about what a salmon farm does with fresh water and sea water, where the footprint comes from, what the cages leave behind, and why an industry built on the sea is moving onto land, where the fresh water number will change.

The two lives of a farmed salmon

A farmed Atlantic salmon is hatched in fresh water and grown, in tanks at a hatchery fed from a river or a lake, to a smolt of about a hundred grams, which takes most of a year. Then it is put to sea, in a net cage in a sheltered fjord or sea loch, and grown for another year and a half or two to a fish of four or five kilograms. The fresh water phase is where the farm's own water is used: a hatchery of the older, flow through kind passes a river through its tanks, tens of thousands of litres for every kilogram of smolt, and returns it a little warmer and carrying the fish's waste; the newer recirculating hatcheries, which are most of the new ones in Norway, filter and reuse almost all of it and take in a few hundred litres per kilogram to make up what is lost.

At sea, the water is the fjord's, and the farm consumes none of it. What the cages do to the fjord is another question.

The feed

The salmon was, until the 1990s, fed on fish: meal and oil from anchovy, sardine and sand eel caught for the purpose, which had no freshwater footprint and a large one in the sea. As the industry grew past what the forage fisheries could supply, the pellet was reformulated, and a Norwegian salmon's feed is now about seventy percent plant: soy protein concentrate from Brazil, wheat and wheat gluten from Europe, rapeseed oil, sunflower and pea protein, with fishmeal and fish oil at a quarter or less. The plants carry the water of the crop articles on this site, and at a little over a kilogram of feed per kilogram of fish, which is the most efficient conversion of any farmed animal, the fish carries about two thousand litres, most of it rain on the soy fields of Mato Grosso and the wheat fields of the Baltic.

The soy is the part the industry is asked about, and it has answered with certification: the concentrate the Norwegian farms buy is, by their own accounting, from fields that were not forest after 2008. The water in it is green, and it is, as the sunflower article argues, rain that fell on farmland that could grow something else.

What the cages leave

The fjord is not consumed, and it is changed. A cage holding two hundred thousand fish drops feed and faeces onto the seabed below it, and the organic load under a farm is the BOD article's story in salt water: a patch of black, oxygen free sediment that recovers, when the cage is moved, over a year or two. The nitrogen and phosphorus that dissolve feed the algae of the fjord. Sea lice, a parasite that lives on wild salmon at low numbers, breed in the cages at densities the wild fish never offered, and the treatments, which have included hydrogen peroxide baths, pesticides in the feed and, latterly, warm water and lasers, go into the fjord with the fish. Escapes put farmed genes into wild rivers. The Norwegian regulator's traffic light system now limits the industry's growth, fjord by fjord, on the lice count, and the fjords where the farms are densest are red.

None of this is water use. All of it is what the water carries, which is the theme of every industry article on this site.

Onto land

The industry's answer to lice, escapes and the limit on fjord sites is to leave the sea. Land based salmon farms, which grow the fish for its whole life in tanks in a building, have been built in Norway, Denmark, Florida, the American northeast and Japan since the late 2010s, and they promise no lice, no escapes, no seabed and a farm next to the market rather than at the end of a fjord. They also need water and power. A recirculating farm reuses more than ninety five percent of its water, and the few percent it takes in, fresh or salt, is multiplied by a fish that lives for three years in the building; a large plant takes in thousands of cubic metres a day and discharges a concentrated stream that needs the treatment the effluent articles describe. And the pumps, filters and chillers that keep the water moving and clean use electricity at a rate that makes the land based salmon's carbon footprint several times the fjord's.

It is the data centre trade again, and the semiconductor one: keep the water in the building and pay in electricity, or use the open water and pay in what it carries. The land based farms have, so far, struggled to make money, and the fjords still hold nearly all of the fish.

Chile

The second salmon country is Chile, which has no native salmon and grows a million tonnes a year in the fjords of Patagonia, and its story is the Norwegian one with weaker regulation. The industry arrived in the 1980s, grew faster than Norway's, and in 2007 was nearly destroyed by a virus that spread between cages placed too densely in fjords with too little current, killing most of the stock and putting tens of thousands out of work in a region that had no other industry. It rebuilt with rules borrowed from Norway, on spacing, fallowing and the movement of fish, and it has since had the largest algal bloom losses in the industry's history, in 2016, when warm water and the nutrients of the farms and the ocean together killed a fifth of the year's fish, which were dumped at sea. Chile uses more antibiotics per tonne of fish than any other salmon producer, by a wide margin, because the bacterial disease that Norway vaccinates against has no effective vaccine in Chilean waters, and the antibiotics go into the fjord with the feed. The fresh water at the hatcheries is drawn from the lakes of the Lake District, and the lakes carry the hatcheries' phosphorus.

The wild fish

The salmon in the cage is the descendant of the salmon in the river, and the river's fish are the reason the fjord is contested. Norway's wild salmon rivers, which draw anglers who pay a great deal, have lost half their runs since the 1980s, and the government's scientists attribute the largest share of the loss to the lice and the escapes of the farms the fish pass on their way to sea. In Scotland, on the west coast where the farms are, the wild runs have collapsed while the east coast's, with no farms, have held. In British Columbia the government has ordered the open cages out of the wild salmon's migration routes by 2029. The water the farmed fish is made of is on a soybean field; the water it lives in is the wild fish's, and that is where the industry's licence to operate is decided.

What it teaches

Salmon is the animal on this site that separates, most cleanly, the water an animal lives in from the water it is made of. The fjord is free and clean and full of lice; the footprint is on a soybean field a hemisphere away; and the farm's own fresh water is at the hatchery, where it is now recycled, and will be in the tank farms, where it will be recycled harder at a cost in power. The number is close to chicken's and it is honest. The fjord, which the number leaves out, is where the argument about salmon actually happens.

Two thousand litres a kilogram, none of it in the fjord, and most of it on a field the fish never saw.

Sources

1. Pahlow, M. et al. (2015). Increasing pressure on freshwater resources due to terrestrial feed ingredients for aquaculture production. Science of the Total Environment 536. Water footprint of aquafeed crops; salmon feed around 1,500 to 2,500 litres per kilogram of fish.
2. Norwegian Directorate of Fisheries and SalMar, Mowi annual reports: production about 1.5 million tonnes in Norway; feed conversion ratio about 1.1 to 1.3.
3. Ytrestøyl, T., Aas, T.S. and Åsgård, T. (2015). Utilisation of feed resources in production of Atlantic salmon in Norway. Aquaculture 448. Feed composition: about 70 percent plant ingredients.
4. Norwegian Institute of Marine Research, risk assessment of Norwegian fish farming: sea lice, escapes, and organic loading under cages.
5. Badiola, M. et al. (2018). Energy use in recirculating aquaculture systems. Aquacultural Engineering 81. Water and energy use of land based salmon farms.
6. Photographs: opener: Jaulas flotantes de salmón, Svolvær, Lofoten, Noruega, 2019-09-05, DD 54 by Diego Delso (CC BY-SA) via Wikimedia Commons; inline: Atlantic salmon at the Craig Brook National Fish Hatchery (23133898042) by US Fish and Wildlife Service (Public domain) via Wikimedia Commons; inline: Salmon fillet by BrokenSphere (CC BY-SA) via Wikimedia Commons.

Johannesburg is the largest city in the world that has no river. It was founded on a ridge, the Witwatersrand, because that is where the gold reef broke the surface in 1886, and the ridge is a watershed: the rain that falls on its north side drains to the Indian Ocean through the Limpopo and the rain on its south side to the Atlantic through the Vaal and the Orange, and neither leaves a river a city could drink from. Six million people live on it now, sixteen million in the province around it, and every litre they use is pumped about three hundred metres uphill from the Vaal, seventy kilometres to the south, by a utility that runs the largest pumping operation in Africa.

This article is about where that water comes from, why the mountains of another country are in it, what the mines left under the ridge, and why a city with a full dam went without water in 2024.

A city with no river

The Vaal is a modest river, and by the 1930s the mines and the city on the ridge had outgrown it. The Vaal dam was built in 1938 and enlarged since, and the utility that draws from it, Rand Water, treats the water at two enormous works on the river and pumps it north through pipelines up the escarpment to a ring of reservoirs on the high ground, from which it flows by gravity to the city. The lift is about three hundred metres, and the electricity to do it is the largest cost of the water. About 4,800 million litres a day are moved this way, to Johannesburg, Pretoria and the mining and industrial towns of the Rand, which together are the economic heart of the country.

The Vaal's own catchment could not supply that, and from the 1950s the engineers looked further: to the Tugela, pumped over the Drakensberg escarpment from the east, and, in the largest scheme of all, to the mountains of Lesotho.

The mountains of Lesotho

Lesotho is a small, poor, mountainous kingdom entirely surrounded by South Africa, and its mountains catch the rain that the plateau below does not. The Lesotho Highlands Water Project, agreed by treaty in 1986 and built through the 1990s and 2000s, dammed the headwaters of the Orange river in the Maluti mountains, at Katse and Mohale, and drove tunnels through the range to deliver the water north into the Vaal's catchment, about 780 million cubic metres a year, close to half of what the Vaal system now supplies. Lesotho is paid for the water and gets electricity from the tunnels' turbines; South Africa gets water that flows to the Vaal by gravity and needs only the last lift to the ridge. The second phase, a dam at Polihali that will add nearly as much again, has been under construction since the late 2010s and is years behind, and the Vaal system has been running close to its limit while it waits.

The arrangement is the largest transfer of water between two countries in Africa, and it is a dependence the treaty was written to manage. In 1998 South African troops entered Lesotho during a political crisis, and among the places they went first was the Katse dam.

The water under the ridge

The gold mines that made Johannesburg went down four kilometres into the reef and, for a century, pumped out the water that seeped into their tunnels. As the mines closed, from the 1990s, the pumping stopped, and the tunnels of the Witwatersrand basins have been filling since. The rock is rich in pyrite, and water on pyrite in air makes sulphuric acid, which leaches iron, manganese, uranium and other metals out of the reef, so that what is rising in the old workings is acid mine drainage of the kind the tailings article on this site describes, at a scale of millions of litres a day. In the Western basin it reached the surface in 2002 and ran into a stream that feeds a nature reserve. In the Central basin, under the city itself, it was rising toward the level at which it would reach the foundations and the groundwater, and the government built pumping and treatment plants in the 2010s to hold it down, neutralised with lime and discharged, still salty, into the rivers.

The plants will have to run for as long as the water rises, which is forever, and their cost falls on a state that did not make the profit. It is the mine tailings story in another form, and it is under the city's feet.

2024

In the southern summer of 2023 and 2024 the Vaal dam was full, and Johannesburg went without water. The failures were in the last twenty kilometres. A heatwave raised demand past what the pumps from Rand Water's reservoirs could lift; the city's own reservoirs, on the high ground, emptied faster than they filled; and the reservoirs could not hold what they had because the pipes below them, many of them a century old, leak or feed illegal connections at a rate the utility puts at more than forty percent of everything it receives. Suburbs in the north and south went for days, in some cases more than a week, with dry taps and tankers in the street. Hospitals trucked water in. The city's utility said it needed tens of billions of rand to replace the pipes and had a fraction of that, because the tariff had been held down, the collections were poor, and the money that was collected had gone elsewhere.

The cause of the crisis, in other words, was the crisis of every article on this site about a city that stopped maintaining its network: a supply that was adequate, delivered through a system that had been run to failure. Rand Water can pump the Vaal to the ridge. It cannot fix the pipes on the other side of the meter, and the city that owns them had not.

Water shifting

The city's answer, for now, is what it calls water shifting: closing valves at night to let the reservoirs recover, and rationing by pressure across suburbs in turn, which is the Nairobi rota under another name. The longer answer is the pipe replacement programme that every audit has recommended and no budget has funded, the fixing of the leaks that would, on their own, give the city a third more water without a litre more from Lesotho, and the recovery of a utility whose skilled engineers left through the 2010s. Polihali, when it is finished, will add water to the Vaal. The reservoirs on the ridge will still leak.

The townships and the tankers

The failures of 2024 fell hardest where they always do. The northern suburbs with their boreholes, rainwater tanks and the money for a delivery ran their gardens through the outage; Soweto, Alexandra and the settlements on the city's edge, where a standpipe serves a street and a household's storage is a bucket, waited for the municipal tanker, which came late and left early, and paid the informal sellers who filled the gap at prices the Nairobi article on this site describes. The province's own audit found that the same suburbs had been on intermittent supply for years before the crisis reached the north and the newspapers, and that the complaints had been logged and closed. A network run to failure fails first at its ends, and the ends of Johannesburg's network are where the people without a borehole live.

The other cost was to the hospitals and the schools. Several of the city's largest hospitals ran on tankers and boreholes for days, postponed surgery and moved patients, and the government's answer was to drill for them, which is the answer of a state that no longer trusts its own pipes. A city that pumps three hundred metres uphill from another country's mountains and then tells its hospitals to sink wells has lost the argument with itself about what the utility is for.

What it teaches

Johannesburg is the city on this site that shows what happens when the supply is engineered and the network is not. The water is brought from the mountains of another country, tunnelled under a range, pumped three hundred metres up a ridge, and lost, close to half of it, in the last few kilometres of pipe. The dam was full and the taps were dry, and the difference between the two was maintenance. Under the city, meanwhile, the water that the gold mines left is rising, and it will have to be pumped for as long as the city stands.

Three hundred metres uphill, half of it lost in the pipes, and acid rising in the tunnels below.

Sources

1. Rand Water, annual reports and Integrated Annual Report 2023: about 4,800 million litres a day supplied to Gauteng; pumping from the Vaal river system to reservoirs on the Witwatersrand.
2. Lesotho Highlands Water Project, Treaty of 1986 and Phase 2 documents: Katse dam (1998), Mohale (2003), Polihali (under construction); 780 million cubic metres a year to South Africa.
3. Johannesburg Water, annual reports and Water Services Development Plan: non revenue water about 40 to 46 percent; pipe replacement backlog.
4. Department of Water and Sanitation, South Africa, Blue Drop and No Drop reports 2023.
5. Council for Geoscience and Department of Water Affairs, acid mine drainage in the Witwatersrand goldfields: the Western, Central and Eastern basins.
6. Photographs: opener: Johannesburg-city-skyline-night by Frankbel (CC BY-SA) via Wikimedia Commons; inline: Katse Dam by SkyPixels (CC BY-SA) via Wikimedia Commons; inline: View Of The Vaal Dam with 14 Sluice Gates Opened 2010 by Michael Gundelfinger (CC BY-SA) via Wikimedia Commons.

Dhaka is the wettest city on this site and one of the most short of water. It gets about two metres of rain a year, most of it in the four months of the monsoon, and it sits at the centre of the Ganges and Brahmaputra delta, ringed by rivers and crossed by canals, on a floodplain that the monsoon covers most years. And it drinks, for two thirds of what it uses, from wells: about nine hundred deep tube wells drilled by the utility and thousands more drilled privately, pulling water from the sand under the city faster than the rain can put it back, so that the water table has fallen by two or three metres a year for four decades and is now sixty or seventy metres down under the centre.

This article is about how a city surrounded by water came to depend on the ground beneath it, what its rivers carry, what the utility is doing to turn back to them, and what it takes to treat a river that a city has used as a drain.

The wells

Dhaka had a million people in 1970 and has more than twenty million now, in a metropolitan area that has spread across the floodplain in every direction, and the utility that serves it, Dhaka WASA, grew its supply the fastest way available: by drilling. The sand under the city is a good aquifer, the wells were cheap, the water needed only chlorination, and each new well could be sunk where the demand was. By the 2000s the city was pumping about two billion litres a day from the ground, and the aquifer, which is recharged slowly from the rivers and the rain through a layer of clay, was being emptied faster than it filled.

The consequences are the ones every article on this site about groundwater describes. The wells go deeper each year and the pumps use more power; the older, shallower wells go dry; and a city whose ground is a soft delta deposit begins, in places, to subside. The arsenic that poisons the shallow wells of rural Bangladesh, which the arsenic article on this site describes, is largely absent from Dhaka's deep aquifer, which is one reason the city drilled deep. The deep aquifer is also the one that does not refill in a human lifetime.

The rivers

The city is ringed by rivers and it drinks from almost none of them, because of what it has put in them. The Buriganga, on the city's southern edge, is the oldest and the worst: a river that in the dry season, from November to May, has almost no oxygen in it, a black surface and a smell that reaches the streets, fed by the untreated sewage of a city of twenty million with a sewer network that reaches a fifth of it, the effluent of the garment industry's dye houses, which the textile article on this site describes, and, until the tanneries were moved in 2017, the chromium and sulphide of the leather district at Hazaribagh, which the leather article describes as one of the most polluted places on Earth. The tanneries were moved to an estate at Savar with a central treatment plant, which has not worked as designed, and the Dhaleshwari river there is now what the Buriganga was.

The Turag and Balu to the north and east carry the effluent of the industrial belts along them, and the Shitalakshya to the east, from which the city's largest surface water plant at Saidabad draws, has become so polluted in the dry season that the plant's intake has to be moved upstream. The rivers are the city's water, and the city has made them the one source it cannot use.

Turning back to the surface

The utility's plan since the 2010s has been to reverse the ratio and take two thirds of the city's water from rivers by the 2030s, and because the rivers at the city's edge are unusable, it is going further out. The Padma plant at Jashaldia, opened in 2019, draws from the Padma, the main channel of the Ganges, thirty kilometres away, and pipes 450 million litres a day into the city; a second plant on the Meghna to the east is planned to be larger; and the Saidabad works on the Shitalakshya is being expanded with a new intake further upstream, where the water is treatable for more of the year. Each is a conventional works of the kind the plain water pillar describes, with coagulation, settling, filtration and chlorine, and each is a large pipeline across a floodplain to a city that grows around it.

The plan is expensive and it is the only one there is. The wells are a stock being spent, the rivers at the edge are a treatment problem no plant can solve until what goes into them is treated first, and the rivers thirty kilometres out are, for now, clean enough. The utility is also spending on the other end: the sewers and treatment works that would take the sewage out of the Buriganga, which have been planned since the 1990s and built at a fraction of the pace of the city.

The garment industry

The dye houses are the part of Dhaka's water that the rest of the world buys. Bangladesh is the second largest exporter of clothing on Earth, most of it made in and around Dhaka, and the washing and dyeing of the fabric uses, as the textile article on this site sets out, between fifty and a hundred and fifty litres for every kilogram of cloth, drawn from the same wells that the city drinks from and discharged, in most plants, to the same rivers. The larger factories that supply the European and American brands have built effluent plants and, under pressure from the brands' auditors, run them, and a few have built the recycling and zero discharge systems the industry's best plants use. The thousands of smaller ones, subcontracting for the large, have not. The cloth is exported. The water stays, in the wells the industry emptied and the rivers it filled.

The monsoon and the flood

The rain that Dhaka cannot drink still arrives, four months a year, at a rate that floods the city. The monsoon delivers about three quarters of the year's two metres between June and September, the rivers rise with the snowmelt and the rain of the whole Himalayan catchment, and the low lying districts of the city, which were wetlands and paddy when the older districts were built and are housing now, go under water for weeks. The city's drains, built for a smaller city and clogged with the plastic and the silt of a larger one, back up; the canals that carried the flood out have been filled and built over; and the water that could, in a planned city, be sent into the ground to refill the aquifer is pumped out into the rivers instead. Dhaka floods in the wet season and drills in the dry one, and the two are the same water.

The planners have known this for decades, and the flood control embankment that rings the western half of the city, built after the flood of 1988, keeps the rivers out and, in the monsoon, keeps the rain in. The eastern half, which the embankment was meant to reach and did not, floods from the rivers directly. The recharge schemes that would use the flood to fill the aquifer exist on paper, at a scale a city of twenty million would need, and in practice as a few pilot wells.

What it costs a household

The utility's water reaches most of the city through pipes and much of the city through nothing. The older districts have connections; the newer ones and the informal settlements, which hold a third of the population, have a shared standpipe, a private well, or a vendor. The utility's water is cheap, subsidised below its cost, and the vendors' water, drawn from the same wells and sold by the jerry can, costs several times more, in the pattern of every city on this site with a network that stopped short of its people. Households that can afford it filter or boil, because the utility's own tests find bacteria in a share of the samples from the pipes, which are old, leaky and under intermittent pressure, so that what was chlorinated at the well is not always clean at the tap. The plants on the Padma will treat the river. The pipes between the plant and the tap are the other half of the work.

What it teaches

Dhaka is the city on this site that shows what a city can do to water it has too much of. It is built on a delta, watered by a monsoon and ringed by rivers, and it drinks from a falling aquifer because it has made every river at its edge untreatable. The plan to turn back to the surface is the right one and it depends on rivers thirty kilometres out staying clean, and on a city of twenty million building, at last, the sewers that would let the ones at its edge recover. The wells are the loan it took while it waited.

Three metres a year, two thirds from the ground, and rivers on every side that the city cannot drink.

Sources

1. Dhaka WASA, annual reports and Master Plan: about 2.6 billion litres a day supplied; about two thirds from around 900 deep tube wells; groundwater decline 2 to 3 metres a year.
2. Hoque, M.A., Hoque, M.M. and Ahmed, K.M. (2007). Declining groundwater level and aquifer dewatering in Dhaka metropolitan area. Hydrogeology Journal 15.
3. Department of Environment, Bangladesh, river water quality monitoring: dissolved oxygen near zero in the Buriganga in the dry season.
4. Padma (Jashaldia) Water Treatment Plant, 450 million litres a day, commissioned 2019; Saidabad Phase 3 and Gandharbpur plants.
5. World Bank (2018). Toward a Multisectoral Action Plan for Sustainable Water Management in Dhaka.
6. Photographs: opener: Small boats on the Buriganga River by Yahya (CC BY-SA) via Wikimedia Commons; inline: Rain filled Dhaka Street,2014 by Aashaa (CC BY-SA) via Wikimedia Commons; inline: Tube well installation project for safe drinking water supply by Needy Foundation 03 by Needy Foundation (CC BY-SA) via Wikimedia Commons.

Athens gets about four hundred millimetres of rain a year, less than most of the cities on this site that are called dry, and it has no river that a city could drink from. The Kifissos and the Ilissos, which the ancient writers mention, are streams that are dry for most of the year and buried under roads for most of their length, and the city that grew round the Acropolis has, for its entire history, drunk water brought from somewhere else. The Roman emperor Hadrian dug a tunnel twenty kilometres long under the city in the second century to bring it from the foothills, and the tunnel still carries water. The modern city, four million people on a plain between the mountains and the sea, drinks from a reservoir two hundred kilometres away.

This article is about where Athens's water has come from over two thousand years, what the longest aqueduct in Europe was built for, what the droughts of 1990 and 2024 took, and why the city is reopening a Roman tunnel.

Two thousand years of bringing water

The classical city drank from springs, wells and cisterns, and from a few small aqueducts from the hills to the north and east, and it was a small city. Hadrian's engineers, around the year 140, dug a tunnel from the foot of Mount Parnitha, twenty kilometres north, at a depth of ten to forty metres, collecting the groundwater of the plain along its length through hundreds of shafts and delivering it to a reservoir on Lycabettus hill, from which it was piped to the city below. It worked for fifteen centuries, was repaired and extended by the Ottomans and by the new Greek state in the 1840s, and was the city's main supply until the 1920s, when Athens, swollen by the refugees of the exchange of populations with Turkey, outgrew it.

The twentieth century built what the Romans could not. The Marathon dam, finished in 1929 and faced with the same marble as the Parthenon, collected the streams of the hills north east of the city; a pipeline in the 1950s brought water from Lake Yliki, a natural lake in Boeotia sixty kilometres away, pumped up over a ridge; and when the city passed three million in the 1970s, the state dammed the Mornos river in the mountains of central Greece, a hundred and ninety kilometres to the west, and built an aqueduct across the mountains and the plain to bring it. The Mornos aqueduct, finished in 1981, is the longest in Europe, and it carries most of what Athens drinks.

1990

The Mornos scheme was designed for a city that would use more water each year, and it was ten years old when the rain stopped. From 1988 to 1993 the catchments of central Greece had a run of dry winters unlike any in the record, the Mornos reservoir fell to a small fraction of its capacity, and by 1990 Athens was rationing: the pressure cut by district, the price of water tripled above a basic allowance, campaigns on every channel, and a government that studied, seriously, the cost of bringing water by tanker from the Rhine and by pipe from the Acheloos. The city's use fell by a third and stayed down, as Melbourne's and Cape Town's did after their droughts, and the rain returned in 1993.

The answer built in the drought was the Evinos: a second river, the next valley west of the Mornos, dammed and tunnelled twenty nine kilometres into the Mornos reservoir, so that two catchments feed one aqueduct. It opened in 2001 and it doubled the security of the supply, and it was, for twenty years, enough.

2024

The winters of 2022 and 2023 were the driest in the Mornos catchment since the 1990 drought, and by the summer of 2024 the total storage of the four reservoirs was down by about a third in two years, the Mornos was showing the drowned village of Kallio at its bottom, which appears in the newspapers whenever the level is low enough to be a warning, and the utility had started pumping from Yliki, which it keeps for dry years because pumping from it costs electricity that the gravity aqueduct does not, and from a set of wells in the Boeotian plain that it holds as a last reserve. The storage was the lowest since the early 1990s. The government declared the region in drought, the utility asked the city to use less, and the winter of 2024 to 2025, which was not wet, did not refill the reservoirs.

The plan, published in 2025, is the plan of every dry city on this site: cut the leakage, which is a fifth of what enters the pipes; take more from the wells and Yliki in dry years; build a pipeline from the Acheloos, the river the 1990 government studied, which has been argued over for thirty years; and, in the long run, desalinate on the coast, which Athens has so far avoided and its islands have not. The city that has always brought its water from further away is running out of further.

Hadrian's aqueduct again

The most Athenian part of the plan is the oldest. Hadrian's tunnel still carries groundwater from Parnitha, several thousand cubic metres a day, and for a century that water has run unused into the sewers, because it is not treated to drinking standard and the city stopped drinking from the tunnel in the 1920s. From 2023 a municipality on the tunnel's route has been drawing it off to water its parks and street trees, and the utility is extending the scheme along the twenty kilometres, so that the parks of northern Athens are watered from a Roman tunnel and the Mornos water they used is kept for the taps. It is a small volume, a fraction of a percent of the city's use, and the point is the one the greywater article on this site makes: water that is not fit to drink is fit for a great deal, and a city in a dry country should not be watering a lawn from a reservoir two hundred kilometres away.

The islands

Attica's water is the mainland's problem. The Greek islands, which host most of the country's tourists in the driest months, have a worse one, and it is the hotel industry's story on this site: a resident population that triples in August on islands that have a few hundred millimetres of rain, no rivers and aquifers that the sea has already entered. For decades the answer was tankers from the mainland, sailing water to Aegina, the Cyclades and the Dodecanese at a cost per cubic metre many times Athens's, and since the 2000s it has been desalination, plant by plant, on Syros, Mykonos, Santorini and dozens more, paid for by the state and the hotels and run at full output for three months a year. The islands are Athens's future in miniature: a dry place with a growing demand, out of rivers to dam, turning to the sea.

Leaks and the last reserve

The utility's own accounting puts the losses in the network at about a fifth of what enters it, which is average for a European city and, in a drought, is the difference between the reservoirs holding and not. The pipes of central Athens date from the 1950s and 1960s, laid fast for a city that doubled twice, and the programme to replace them has run at a pace the reservoirs have not waited for. The reserve wells in Boeotia, which the utility drew on in 2024, are the last resort in a drier sense: the aquifer under the Copais plain, which the ancient lake drained in the 1880s left behind, is what the farms of the plain also pump, and the utility's draw on it in a dry year is a draw on their water too. The city has one more source after that, and it is the sea.

What it teaches

Athens is the city on this site that has been bringing its water from somewhere else for the longest, and the distance has grown with the city: twenty kilometres under Hadrian, sixty in the 1950s, two hundred by 1981. Each scheme was the answer for a generation and each was found short in a drought, and the droughts, on a coast that is drying, are coming closer together. The reservoirs will refill in the wet winters. The city is reopening a Roman tunnel and studying the sea, which is the oldest answer and the last one.

Two hundred kilometres of aqueduct, a reservoir a third down, and a tunnel from the year 140 watering the parks.

Sources

1. EYDAP (Athens Water Supply and Sewerage Company), annual reports: about 1.1 to 1.3 million cubic metres a day; sources Mornos, Evinos, Yliki and Marathon; the Mornos aqueduct about 188 kilometres plus the Evinos tunnel.
2. Ministry of Environment and Energy, Greece, reservoir storage data 2022 to 2025: total storage down about a third from 2022; lowest since the early 1990s.
3. Koutsoyiannis, D. et al. (2003). A decision support system for the management of the water resource system of Athens. Physics and Chemistry of the Earth 28. The 1989 to 1993 drought and the Evinos scheme.
4. Hadrian's Aqueduct revival project, Municipality of Chalandri and EYDAP: about 20 kilometres of the ancient tunnel, non potable use since 2023.
5. Hellenic National Meteorological Service, rainfall records for Athens: about 400 millimetres a year on average.
6. Photographs: opener: View from the Acropolis in Athens with the Lycabettus Hill, 20240531 1314 9656 by Jakub Hałun (CC BY) via Wikimedia Commons; inline: Τα Βαρδούσια καθρεπτίζονται στην Λίμνη του Μόρνου by George Kypreos (CC BY-SA) via Wikimedia Commons; inline: The Marathon Dam on July 22, 2020 by George E. Koronaios (CC BY-SA) via Wikimedia Commons.

A glass of orange juice at breakfast carries about two hundred litres of water, and the odds are that the juice in it was squeezed in the state of São Paulo, frozen into a concentrate, shipped across the Atlantic in a refrigerated tanker, and reconstituted with local water in a carton plant a few hundred kilometres from the table. Orange juice is the most traded fruit product in the world and one of the most concentrated, in both senses: a kilogram of the frozen concentrate is six litres of juice with the water taken out, and three quarters of what is traded comes from one place.

This article is about where an orange's water goes, why the juice carries twice the water of the fruit, what happened to Florida, why the Brazilian groves and the Spanish ones carry different kinds of water, and what depends on a single state in Brazil.

The fruit

An orange tree is an evergreen that needs water all year and most of it in the hot months, when the fruit is sizing, and it needs about a thousand millimetres over a season, which is why it grows where the summers are wet or where the irrigation is. A grove gives twenty to forty tonnes of fruit a hectare, and the footprint per kilogram, in the global averages, is about 560 litres: roughly the same as an apple's, a third of a kilogram of rice, and a fraction of any animal product. It is a modest number for a crop that is mostly water itself, about eighty five percent.

The juice doubles it. A kilogram of oranges gives about half a kilogram of juice, and the rest, peel, pulp, membranes and seeds, is dried and sold as cattle feed, which carries a small share of the water, so that the juice carries the rest: about a thousand litres for a litre, and two hundred for a glass. The concentrate carries the same water per litre of reconstituted juice, and the tanker that ships it carries a sixth of the volume.

Brazil

The world's orange juice comes from the citrus belt of São Paulo and the western tip of Minas Gerais, a plateau of about two hundred million trees where the summers are wet, the winters are dry and cool, and the groves are, for the most part, rain fed, with irrigation added in the last decade as the dry seasons have grown longer. The industry is three companies, which own groves, crushing plants, tank farms at the port of Santos and fleets of refrigerated tankers, and it ships about three quarters of all the orange juice traded in the world, most of it to Europe and the United States, as frozen concentrate at minus ten degrees and, increasingly, as chilled juice that has never been concentrated and is shipped at a sixth of the density and six times the freight.

The water in a Brazilian glass is therefore mostly green, rain on a plateau, and the blue share is growing as the groves are irrigated against the droughts of the 2010s and 2020s, from rivers and wells in a region that also grows sugar cane and, as the São Paulo article on this site describes, came close to running the city dry in 2014. The grey share is the fertiliser and the pesticide, and citrus is a sprayed crop, against the insect that carries the disease of the next section.

Florida

In 2000 Florida grew about ten million tonnes of oranges, nearly all for juice, and was the other half of the world's supply. In 2005 an insect called the Asian citrus psyllid, which had arrived a few years earlier, was found carrying a bacterium that causes the disease called greening, which blocks the tree's vessels, makes the fruit small, bitter and green, and kills the tree in a few years. There is no cure. By 2024 Florida's crop was down about ninety percent, the groves of the centre of the state were being sold for housing, and the juice plants had closed or turned to processing Brazilian and Mexican fruit. The disease reached Brazil at the same time and the industry there has held it, so far, with a programme of inspection, spraying and the removal of infected trees on a scale Florida's smaller growers could not afford, and with new plantings on ground the insect has not yet reached.

The point for this site is that the water of a product follows its geography, and the geography moved. A glass of juice in New York in 2000 carried Florida's water, which was rain and the aquifer under the state, drawn down by the groves; a glass in 2024 carries São Paulo's, which is rain on a plateau and a growing share of wells. The number did not change much. Where it fell did.

Spain and California

The oranges that are sold as fruit, in Europe from Valencia and Andalusia and in the United States from California, are the irrigated ones. Valencia's groves are on the coastal plain that the Moors irrigated a thousand years ago, from the Turia and the Júcar and, since the 1970s, from wells that the desalination and brine articles on this site describe as pulling the sea into the aquifer; Andalusia's are in the driest part of Europe, on water from the Guadalquivir that the olive groves and the strawberry farms also claim. California's are in the Central Valley, on the state's water project and on the wells that the Central Valley article describes as sinking the ground. A kilogram of Spanish or Californian oranges carries about the same total as a Brazilian one and a far larger blue share, from rivers and aquifers that have other claimants, in the months when the claimants need them most. The juice is Brazil's story. The fruit bowl is Spain's.

The concentrate

The concentrate plant is a food factory of the kind the food and beverage article describes, and its water is unremarkable: a few litres per litre of juice for washing the fruit, cleaning the extractors and cooling the evaporators, with an effluent high in sugar and peel oil that a biological plant treats without difficulty. The evaporator that takes the water out of the juice is the interesting part, because it removes about five litres of water from every six, and that water, condensed, is clean enough to reuse in the plant. A litre of concentrate that lands in Rotterdam is reconstituted with five litres of Dutch tap water, so that a glass of juice from concentrate is, by volume, mostly the water of the country it is drunk in and, by footprint, almost entirely the rain of São Paulo.

The frost and the price

The orange is a tree that a single night can ruin. A frost below about minus two degrees for a few hours kills the fruit on the tree, and a harder one kills the tree, and the groves of Florida were moved south through the twentieth century by a series of freezes, in 1962, 1983, 1985 and 1989, each of which ended citrus in a band of the state where it had been grown for a century. The juice futures market in New York, which is the only market in the world for frozen orange juice and the reason the product exists in its frozen form, moves on the Florida weather forecast, and the Brazilian industry, on a plateau where frost is rare, built its share in the gaps the freezes opened. Water enters the story at the edges: a grove that is irrigated can be sprayed on a frost night, because water freezing on the fruit releases heat and holds it at zero, and the sprinklers that saved the Florida groves in the 1980s drew the aquifer down on the coldest nights of the year, which is the one use of irrigation water on this site that has nothing to do with growing.

The price, in the 2020s, has been the highest ever recorded, because Florida is gone, greening is in Brazil, and the drought of 2024 cut the Brazilian crop by a fifth. A glass of juice carries two hundred litres of rain and, for the first time in fifty years, a price that reflects the fact that almost all of it falls in one place.

What it teaches

Orange juice is the product on this site that shows most clearly how a footprint travels. The number is modest and mostly rain, and it is drunk in a glass that contains, by volume, mostly local water added at a carton plant; the rain that grew it fell on a plateau in Brazil that also lost its own city's water in a drought, and the half of the world's supply that used to be Florida's has gone, to an insect, in twenty years. The oranges in the fruit bowl carry a river. The juice carries a plateau's rain and a ship.

A thousand litres a litre, two hundred a glass, and three quarters of it from one state in Brazil.

Sources

1. Mekonnen, M.M. and Hoekstra, A.Y. (2011). The green, blue and grey water footprint of crops and derived crop products. Oranges about 560 litres per kilogram; orange juice about 1,018 litres per litre.
2. USDA Foreign Agricultural Service, Citrus: World Markets and Trade: Brazil about 75 percent of orange juice exports; Florida production down about 90 percent from 2000 to 2024.
3. Fundecitrus (Brazil), citrus belt surveys: about 200 million trees in São Paulo and Minas Gerais; greening incidence.
4. University of Florida IFAS, citrus greening (huanglongbing) and its effect on Florida production and irrigation.
5. Generalitat Valenciana and Spanish Ministry of Agriculture, citrus irrigation in Valencia and Andalusia.
6. Photographs: opener: Orange Grove, Prospect Park, 12-2011 (6808109605) by inkknife_2000 (CC BY-SA) via Wikimedia Commons; inline: Picking oranges at Tel Mond (FL62120709) by IPPA photographer (CC BY) via Wikimedia Commons; inline: Glass of orange juice with ice by Grandmaster Huon (CC0) via Wikimedia Commons.

The water in a steak, which the beef article on this site puts at about fifteen thousand litres a kilogram, is almost all on the farm: the rain and irrigation that grew the grass and the grain the animal ate over its life. The plant that turns the animal into the steak uses about ten litres for every kilogram, a fraction of a percent of the total, and it is on this site anyway, because what a slaughterhouse sends out is among the strongest effluents in any industry and it goes into a river in a single stream from a single building. A meat plant is a small water user with a large water effect, and the two ends of that are the subject of this article.

What the plant does with water

A slaughterhouse is a washing operation with a killing floor attached. Animals arrive, are held, are watered and often washed; they are stunned, killed and bled; the carcass is scalded or skinned, eviscerated, split, washed, chilled and cut, and at every step there is water: hot water to scald a pig or a chicken so the hair or feathers come off, water to wash the carcass after evisceration so that what was in the gut does not stay on the meat, water to rinse knives and saws between animals, water in the chill tanks where chickens are cooled, and water, above all, to wash the floors, walls, hooks, conveyors and tables continuously through the shift and thoroughly at the end of it. The rule of the plant, written into food safety law in every country, is that nothing touches meat without being clean, and clean means washed, and washed means water at sixty degrees or more with chlorine or a detergent in it.

The totals are two to fifteen cubic metres per tonne of carcass for cattle and pigs, with ten typical, and ten to twenty five litres for each chicken, which is a small bird, so that a chicken plant killing a million birds a week uses the water of a town of thirty thousand people.

What goes down the drain

The effluent of a meat plant carries what the animal was made of and what was in it. Blood is the worst: a litre of blood has a biochemical oxygen demand, the measure the BOD article on this site explains, of a hundred and fifty to two hundred grams, against half a gram for a litre of raw sewage, so that the blood of a single steer, about fifteen litres, has the oxygen demand of the sewage of several hundred people for a day. Then fat, which floats and clogs and takes oxygen slowly; paunch and gut contents, which are manure; manure from the holding pens; hair, feathers, and fragments of tissue; and the wash water itself, hot and carrying chlorine, detergent and salt. The raw effluent of a plant that does not separate its streams has an oxygen demand of several thousand milligrams a litre, five to ten times sewage, and a plant that has let its blood go to the drain can send out more oxygen demand in a day than the town it sits in.

The first rule of the trade, therefore, is to keep the blood out of the water. It is collected at the kill, in a trough under the bleeding rail, and rendered into blood meal or plasma, which have a value; the fat and the trimmings go to the rendering plant with the bones and are made into tallow and meat meal; the paunch contents go to composting or the fields. What is left, at a well run plant, is the wash water, still strong, and it goes to a treatment plant that is, at most large sites, the plant's own.

The treatment plant

A slaughterhouse effluent plant is the effluent article on this site with a grease trap at the front. Screens take out the solids; a flotation tank, in which fine bubbles lift the fat and blood protein to the surface where they are skimmed, takes out the grease and a large share of the oxygen demand; and the rest goes to biological treatment, often anaerobic first, because the load is high and the methane is worth burning, then aerobic, then a settling tank and a discharge to the sewer or the river. The ammonia, from the protein, needs the nitrogen removal the nitrogen article describes, and the salt and the chlorine pass through. A plant that runs all of that sends out water that a sewer can take. A plant that runs a grease trap and a pipe sends out what the table above shows.

The plants are large and in the countryside, near the animals, on rivers that are small, and the record of fish kills below meat plants in the American midwest, in Brazil and in eastern Europe is a long one. In the United States the rules for the industry were written in the 1970s and revised in 2004, and the regulators' own audits since have found a large share of the plants discharging above their limits into rivers that also carry the runoff of the feedlots upstream.

The chicken

The chicken plant is the modern version of the trade, and the fastest. A large plant kills a hundred and fifty birds a minute on a line, scalds them, plucks them, guts them, washes them inside and out, and then cools them in chill tanks, long troughs of near freezing water with chlorine or peracetic acid in it, through which the carcasses are dragged for an hour to bring their temperature down before packing. The chill tank is where much of the water goes and where much of the contamination is exchanged, because a bird carrying salmonella shares the tank with a thousand that were not, which is why the tank is chlorinated and why the European Union, which forbids the chlorine wash, uses air chilling instead, at a higher energy cost and a lower water one. The plucked feathers and the offal go to rendering and become the fishmeal substitute in the feed of the next flock, and, as the salmon article on this site describes, of the fish.

The rendering plant

Behind every slaughterhouse is a rendering plant, and it is the part of the industry that turns the water problem into a product. Everything the plant does not sell as meat, which is about half of a cow and a third of a chicken, goes to the renderer: bones, fat, offal that is not eaten, blood, feathers, hair, hooves and the carcasses condemned at inspection. The renderer cooks it, at high temperature under pressure, and separates it into fat, which becomes tallow for soap, biodiesel and feed, and a dry protein meal that goes into pet food, fish feed and the feed of the next flock. The water that leaves is the water that was in the tissue, condensed from the cookers, and the renderer's effluent is the slaughterhouse's without the blood: strong, fatty, and smelling of what it was, which is why rendering plants are the most complained about neighbours in the food industry. A meat plant that sends its blood and fat to the renderer has cut its effluent's strength by more than half and has been paid for it, and the plants that do not are the small, old and remote ones, which are also the ones on the smallest rivers.

The water in the packet

The packet on the shelf carries the farm's water and the plant's, and a third that neither counts: the water sprayed on the carcass as an antimicrobial. In the United States a carcass is washed with chlorine, peracetic acid or lactic acid at several points on the line, and the chicken is chilled in chlorinated water that it absorbs, so that a share of the packet's weight, up to several percent, is water added at the plant and sold at the price of meat. The European Union forbids the washes and the absorbed water, which is the substance of the long argument about chlorinated chicken in trade negotiations, and its plants rely on hygiene along the line instead, at a lower water use and a higher labour one. The packet, either way, carries the plant's ten litres a kilogram and the farm's thousands, and the difference between the two systems is a few percent of water in the meat and the treatment of the water that leaves the building.

What it teaches

The slaughterhouse is the industry on this site where the footprint number is misleading in the opposite direction from wool. A kilogram of meat carries thousands of litres, almost all of them on the farm, and the plant that made it into meat used ten; and the plant, not the farm, is where a river is killed in an afternoon, because a building that handles blood and fat by the tonne concentrates what a thousand fields spread out. The water use is small and hot and endless, for hygiene, and the water effect depends entirely on whether the blood went to the rendering plant or the drain.

Ten cubic metres a tonne, twenty litres a bird, and a litre of blood with the oxygen demand of three hundred litres of sewage.

Sources

1. European Commission, Best Available Techniques reference document for the slaughterhouses and animal by products industries (2005, revised 2023): water use 2 to 15 cubic metres per tonne of carcass; poultry 10 to 25 litres per bird.
2. US EPA (2004). Effluent limitations guidelines for the meat and poultry products point source category: effluent BOD, oil and grease, ammonia.
3. Bustillo Lecompte, C.F. and Mehrvar, M. (2015). Slaughterhouse wastewater characteristics, treatment and management. Journal of Environmental Management 161. BOD 1,000 to 8,000 mg/L; blood about 150,000 to 200,000 mg/L.
4. USDA Food Safety and Inspection Service, pathogen reduction and HACCP rules: carcass washing and antimicrobial interventions.
5. FAO, meat production statistics: about 360 million tonnes a year; about 75 billion chickens.
6. Photographs: opener: Slaughter House Showing Carcasses Hanging(GN04710) by State Government Photographer, State Library of South Australia (CC0) via Wikimedia Commons; inline: Meat and Poultry Processing Expansion Program Announcement (20230221-OSEC-TEW-6102) by US Department of Agriculture (Public domain) via Wikimedia Commons; inline: Meat and Poultry Processing Expansion Program Announcement (20230221-OSEC-TEW-3434) by US Department of Agriculture (Public domain) via Wikimedia Commons.

A hotel is a small town that has agreed to behave badly about water. Its guests take longer showers than they take at home, its sheets and towels are washed daily, its pool evaporates through the afternoon, its gardens are green in the driest month of the year, and its restaurants, spas and golf courses use water in the ways that the food, textile and lawn articles on this site describe, all at once and on holiday. A guest in an ordinary city hotel uses about three hundred litres a night, twice what a resident of the same city uses; a guest in a resort with pools, lawns and a spa uses a thousand or two. And the industry has built itself, by preference, on the coasts and islands with the least water on Earth, in the season when they have least.

This article is about what a hotel does with water, why the number varies by a factor of twenty, what tourism did to the islands of the Mediterranean, and what the industry has done since.

Where the water goes in a hotel

The largest use in most hotels is the guest room: the shower, which runs longer and hotter on holiday than at home, and the toilet, and in the older hotels the bath. The second is the laundry, which washes every sheet and towel in the building every day unless the guest hangs the card on the door, at fifteen to twenty litres per kilogram of linen. The third is the kitchen, and behind those the pool, the gardens, the cooling towers of the air conditioning in a hot climate, and the spa. A city business hotel with none of the outdoor uses sits at two to three hundred litres a night. A resort adds the pool, which loses a centimetre a day to evaporation in the sun, and the gardens, which in a dry climate are irrigated lawn and planting that would die in a week without it, and the number climbs past a thousand.

Where the hotels are

The industry is largest where the rain is least. The Mediterranean is the most visited region on Earth, and its islands and southern coasts, which have the driest summers in Europe, host a tourist population that doubles or triples the resident one from June to September, exactly when the aquifers that both depend on are lowest. Mallorca has a million residents and fourteen million visitors a year; Crete, Rhodes, Malta, Cyprus and the Canaries have similar ratios; and the same pattern holds in the Caribbean, in Bali, in Zanzibar and on every dry coast where the beach is the product.

The hotels' water in those places came, at first, from the same wells as the towns', and the wells were drawn down through every summer of the 1970s and 1980s until the sea came in behind the fresh water, which the desalination and brine articles on this site describe as the salting of a coastal aquifer, and which on Mallorca and the Costa Blanca made the town supply brackish by the 1990s. The islands answered with desalination plants, which the hotels' tariffs pay for and which run at full output only in the summer months, and with pipelines from the wetter interiors. Malta, which has no rivers, drinks desalinated water for most of its supply and its tourists with it.

The golf course

The extreme case is the golf course, which the lawn article on this site describes as a lawn the size of a farm. A course in a dry climate needs about a million cubic metres of water a year to stay green, which is the drinking water of a town of ten to fifteen thousand people, and it needs it in the summer, when the town needs its own. The coasts of southern Spain have several hundred courses, most built between 1990 and 2008 as the anchors of resort developments, and the region's water plans of that period counted them, with some embarrassment, as one of the larger agricultural users. The rule that has since been written into most of them, and enforced unevenly, is that a course must water itself with treated effluent rather than the town's supply, and the better resorts now run their own treatment plants and irrigate the fairways with what their guests flushed. The worse ones have wells.

What the industry has done

The hotel industry has been counting its water for two decades, because the water is a cost and because the brands are asked about it, and the fixes are known. Low flow showers and dual flush toilets cut the room's use by a third without the guest noticing. The card on the door that asks whether the towels need washing, which every hotel now has, cuts the laundry by a fifth where it is honoured. Pool covers at night halve the evaporation. Gardens planted with what grows on the coast without irrigation replace the lawns, slowly, because the lawn is what the brochure shows. And the resort's own effluent, treated on site, waters what remains, so that the resort becomes a small version of the water recycling that the greywater and Singapore articles describe. The best resorts in the dry regions are now at three to four hundred litres a night, down from over a thousand, and they advertise it.

The larger problem is the number of guests. A coast that has cut its water per guest by half while doubling its guests has gained nothing, and most of the Mediterranean's coasts have done exactly that. The islands' plans now talk of limits on beds, which the islands' economies, built on the beds, resist.

What the guest carries

The water a guest uses in the room is the smaller part. The food the hotel serves carries the water of the food articles on this site, and a resort's buffet, with its beef, its cheese and its fruit flown in, carries several thousand litres per guest per day before anyone has showered. The linen carries cotton's water, the pool its chemicals, and the flight nothing at all in water and a great deal in carbon. Counted that way, a week in a resort is a few tens of thousands of litres, most of them far from the resort, which is the point the water you cannot see article makes about every product and which is true of a holiday.

The cruise ship and the villa

Two forms of tourism sit outside the hotel's accounting and inside the coast's. A cruise ship is a resort that carries its own water, made by desalination on board at a rate of a few hundred litres per passenger per day, and discharges its treated sewage at sea beyond the limit, so that a ship in a port takes little from the town and leaves its several thousand passengers on the quay for the day, using the town's toilets, showers and restaurants without a bed to be counted against. The villa, which the rental platforms have multiplied across every Mediterranean coast in the last decade, is a house with a private pool and a garden, occupied for a few weeks a year by people who are not counted as hotel guests and whose pool evaporates through the eleven months the house is empty. On Mallorca and in the Algarve the villa pools now outnumber the hotel ones, each holding tens of thousands of litres and topped up from the mains, and the islands' water plans, written for hotels, have had to be rewritten for houses.

The plans have tools. A meter on every villa, a tariff that rises steeply above a household's use, a rule that a pool must be covered when not in use and filled outside the summer, and the same reuse of treated effluent on gardens that the resorts have adopted. The villa owner, who is usually not on the island, is the hardest customer the utility has.

What it teaches

Tourism is the industry on this site that puts the largest number of people, using the most water per head, in the places and the months with the least. The hotel's water is not large by the standards of a steelworks or a farm; it is large because of where the hotel is and who is outside the gate. The fixes are the household's fixes, applied to a building that has every reason to apply them and a brochure that resists, and the islands have paid for them with desalination plants and rules about golf. The guest, meanwhile, could take the four minute shower they take at home.

Three hundred litres a night, two thousand with the golf, and a coast that doubles its people in the month its wells are lowest.

Sources

1. Gössling, S. et al. (2012). Tourism and water use: supply, demand, and security. Tourism Management 33. Direct use 84 to 2,000 litres per guest night; indirect use higher.
2. UNWTO and UNEP (2019). Tourism and water: sustainability in a dry region. Mediterranean case studies.
3. Hof, A. and Schmitt, T. (2011). Urban and tourist land use patterns and water consumption: evidence from Mallorca. Land Use Policy 28.
4. Rico Amoros, A.M. et al. (2009). Tourist land use patterns and water demand: evidence from the Western Mediterranean. Land Use Policy 26. Golf course water use in Alicante and Murcia.
5. Hotel industry benchmarks: Cornell Hotel Sustainability Benchmarking Index, water per occupied room by climate and hotel type.
6. Photographs: opener: Swimming pool and main building of Amantaka luxury Resort & Hotel in Luang Prabang Laos by Basile Morin (CC BY-SA) via Wikimedia Commons; inline: 2026-06-01 09 01 59 A sprinkler on the first hole green of the Mountain View Golf Course in the Mountainview section of Ewing Township, Mercer County, New Jersey by Famartin (CC BY-SA) via Wikimedia Commons; inline: Hotels near the beach of Cala Mesquida in Mallorca, Spain (48001491916) by dronepicr (CC BY) via Wikimedia Commons.

Almost all of the fresh water on Earth that is liquid and not in a lake is underground. It fills the spaces between grains of sand and gravel, the cracks in sandstone and limestone and, in a few places, caverns, and it is called an aquifer when there is enough of it, close enough to the surface and in rock porous enough, for a well to draw on. The world pumps about a thousand cubic kilometres of it a year, which is about a third of all the fresh water it uses and about half of what it drinks, and in the dry regions where the pumping is heaviest it pumps faster than the rain refills. The difference is the water table falling, and it is falling, by the satellites' measure, across every major farming region on Earth.

This article is about what an aquifer is and how it fills, who pumps it and for what, what the satellites saw, where the ground has sunk, and why the water under a field is the hardest water on this site to govern.

What an aquifer is

Rain that falls on the ground either runs off into a river, evaporates, is taken up by plants, or soaks down through the soil until it reaches a depth where every pore is full. That depth is the water table, and everything below it is groundwater, moving slowly, a few metres a year, through the rock toward the nearest river, spring or sea. An aquifer is a layer where the rock is porous enough to hold a lot and permeable enough to give it up to a well: the sands and gravels of a river plain, the sandstones of a basin, the limestone of a karst. Under a river plain the water table may be a few metres down and fill every wet season; under a desert basin it may be hundreds of metres down and fill at a millimetre a year from rain that fell on distant hills.

The recharge is the number that matters. An aquifer that fills at a hundred millimetres a year across its area can be pumped at that rate forever, and pumped at twice that rate it is being emptied, however large the stock. Most of the world's heavily pumped aquifers are in dry places, because dry places are where the rivers are not enough, and dry places have the least recharge.

Who pumps it

The pump changed everything. Until the 1950s a well was a bucket or a hand pump and it drew what a family or a village needed; the electric and diesel pump, spread through India, China, the American plains and the Middle East from the 1960s, drew what a field needed, which is a thousand times more, and the green revolution of Asia was, to a large extent, a groundwater revolution. India has about twenty million irrigation wells and pumps about 250 cubic kilometres a year, more than any other country and more than the United States and China together, with electricity that the states subsidise to nothing, so that a farmer in Punjab or Gujarat pumps until the pump runs dry and then drills deeper. The North China Plain, which grows the country's wheat, has drawn its water table down by tens of metres. The Ogallala aquifer under the American plains, which was filled in the ice ages and recharges at a few millimetres a year, has lost, in parts of Kansas and Texas, most of the water it had in 1950, and the irrigated circles of the plains are being turned back to dry wheat as the wells fail. The Central Valley of California, which the Phoenix and almond articles on this site touch, pumps hardest in a drought and has sunk, in places, by nine metres.

The cities pump too. Mexico City, Jakarta, Tehran, Beijing, Bangalore and Dhaka, on this site, are all cities living on aquifers they are emptying, and their articles describe what follows: wells that deepen each year, ground that sinks, and a shift, late and expensive, to water brought from further away.

What the satellites saw

From 2002 a pair of satellites called GRACE, flying in formation, measured the Earth's gravity finely enough to detect the mass of water moving on and under its surface, and their maps of change over fifteen years are the clearest picture of groundwater the world has. They show the water table falling across northern India at a rate that adds up to the largest loss of water anywhere; across the North China Plain, the Middle East from Turkey to Iran, the Central Valley, the Ogallala, the Murray Darling in its drought, and the Caspian basin; and rising, in a few places, where the rain has increased or the pumping has been reined in. The satellites cannot see a well. They can see the sum of all the wells, and the sum is a loss of a few hundred cubic kilometres a year, most of which ends up, after passing through crops and rivers, in the sea.

A study in 2024 of a hundred and seventy thousand wells in forty countries found the water table falling in most of the world's aquifers and falling faster in the twenty first century than the twentieth, with the exceptions, a few dozen aquifers that had recovered, all in places where pumping had been limited by law or water brought from elsewhere. Recovery is possible. It has to be decided.

Fossil water

Some of the water underground fell as rain in a wetter climate ten thousand or more years ago and has not been replaced since, and it is called fossil water because, like oil, it is a stock and not a flow. The Nubian sandstone under Egypt, Libya, Sudan and Chad holds more water than the Nile carries in five hundred years, and Libya's Great Man Made River, built in the 1980s and 1990s, pumps it to the coast. The Arabian peninsula's aquifers grew Saudi Arabia's wheat in the 1980s and 1990s, at a cost of most of the water, until the programme was ended in 2016 and the wheat was imported instead. The Ogallala is mostly fossil. Water of that kind cannot be managed sustainably, because there is no sustainable rate; it can only be spent slowly or fast, on something worth it or not.

Why it is hard to govern

A river can be seen, measured at a gauge and divided by a treaty. An aquifer is invisible, crosses every boundary drawn on the surface, and is reached by a well on private land that, in most of the world's law, belongs to the landowner with the water under it. A thousand farmers over an aquifer are a thousand people drawing from one account with no statement, and each has every reason to pump before the neighbour does. The places that have slowed the fall have done it by metering wells, by charging for electricity, by capping pumping district by district, as California began in 2014 and India has barely started, and by putting water back: the managed recharge that Orange County and Perth practise, sending treated water or flood water into the aquifer on purpose. The rest are living on the loan.

What comes up with it

The water that has sat in rock for centuries carries the rock, and the arsenic, fluoride and nitrate articles on this site are all groundwater stories: the arsenic of the Bengal delta and the fluoride of the Rift Valley and the Indian plains are in the sediment and the rock and dissolve into the water that sits in them, and the nitrate is what the fields above put down. Groundwater is also, in most of the world, the cleanest water there is, filtered through tens of metres of sand, free of the bacteria and the organic matter that a river carries, and needing, at a well run works, little more than a chlorine dose. The deeper the well, the older and cleaner the water and the less of it there is to come. The pump, which made it possible to draw on the deep stock, made it possible to spend it, and the same drill that reaches clean water reaches water that will not be replaced.

What it teaches

Groundwater is the plain water article about the water nobody can see, and it is the largest stock on this site and the one being spent fastest. A third of what the world uses comes from the ground, half of what it drinks, and in the regions that feed most of its people it comes faster than the rain puts it back, from wells that no one meters, on land that no treaty covers. The satellites can see the total falling. The fix is the one every falling aquifer on this site has reached in the end: pump less, put some back, and decide what the stock is for.

A thousand cubic kilometres a year, a fifth of it never replaced, and the ground sinking where the wells go deepest.

Sources

1. Wada, Y. et al. (2010). Global depletion of groundwater resources. Geophysical Research Letters 37. Depletion about 280 cubic kilometres a year.
2. Rodell, M. et al. (2018). Emerging trends in global freshwater availability. Nature 557. GRACE satellite observations 2002 to 2016.
3. UNESCO and IGRAC, United Nations World Water Development Report 2022: Groundwater, making the invisible visible. About 1,000 cubic kilometres a year abstracted; India the largest user.
4. Jasechko, S. et al. (2024). Rapid groundwater decline and some cases of recovery in aquifers globally. Nature 625. 170,000 wells in 40 countries.
5. Famiglietti, J.S. (2014). The global groundwater crisis. Nature Climate Change 4.
6. Photographs: opener: Hermiston area aerial (34522051522) by Sam Beebe (CC BY) via Wikimedia Commons; inline: Woman operating an electrical pump starter for irrigation in Raichur by Vraj Acharya (CC BY-SA) via Wikimedia Commons; inline: Crops Kansas AST 20010624 by NASA (Public domain) via Wikimedia Commons.

In 1974 a chemist at the Rotterdam water works, Johannes Rook, found chloroform in the city's tap water and, after ruling out every other source, worked out that the works was making it: the chlorine added to disinfect the water was reacting with the faint tea colour of the river Rhine, the dissolved remains of leaves and soil that every surface water carries, and turning a little of it into chloroform. An American team at the federal environment agency found the same thing in the same year in the water of several American cities. Every chlorinated supply in the world had it, and had had since the first chlorination of a public supply in 1908, at levels of tens of micrograms a litre that nobody had looked for because nobody had the instrument to see it.

This article is about what those compounds are, how a water works makes them without meaning to, what the health studies have found in fifty years of looking, how the works have reduced them, and why the answer was never to stop adding chlorine.

What they are

Chlorine in water is a strong oxidant, and it reacts with whatever it meets. What it meets, in a river or a reservoir, is natural organic matter: humic and fulvic acids, the brown dissolved residue of dead vegetation that gives a moorland stream its colour and a reservoir its faint tint, at a few milligrams a litre. Chlorine breaks those large molecules and attaches to the fragments, and among the products are a family of small compounds with a carbon atom carrying three halogens, which is what trihalomethane means: chloroform is the commonest, and where the water carries bromide, from the sea or from rock, the bromine versions form too. Alongside them form the haloacetic acids, and a longer list of compounds at lower concentrations that the chemists are still cataloguing.

The amount depends on how much organic matter there was, how much chlorine was added, how warm the water is and how long the two have been together; a supply from a peaty upland reservoir in a warm summer, chlorinated hard and held for days in a long pipe network, makes the most. A supply from a deep well with almost no organic matter makes almost none.

What the studies found

Chloroform causes cancer in rats given large doses, which was known in the 1970s, and the question since has been what tens of micrograms a litre over a lifetime do to people. The answer has come from epidemiology, above all from studies of bladder cancer, which is the cancer most consistently linked. A pooled analysis of several studies in 2004 found that people who had drunk chlorinated water with high trihalomethane levels for decades had a risk of bladder cancer somewhat higher than those who had not, by a factor around one and a half at the highest exposures, with the risk also tracking showering and swimming, because the compounds are breathed in and absorbed through the skin as well as drunk. A study across the European Union in 2020 estimated that a few percent of the bladder cancers in the Union might be attributable to the byproducts, concentrated in the countries with the highest levels, and that bringing every supply down to the levels of the best would prevent most of them.

The effect is small and real, of the order that the regulators write limits for. The studies of reproductive outcomes, of other cancers and of the wider list of compounds have been less consistent, and the research goes on. The limit of eighty or a hundred micrograms a litre is set, as such limits are, to hold the lifetime risk at a level the regulator judges acceptable against the alternative.

How the works reduce them

The lesson the works drew from 1974 was to attack the organic matter rather than the chlorine. The coagulation and filtration that the plain water articles on this site describe were tuned to remove more of the colour: a slightly lower pH, a higher dose of the coagulant, and the dissolved humic material comes out with the floc, so that there is less for the chlorine to react with. Activated carbon, granular beds that adsorb organic matter, was added at the works that needed it. Ozone or ultraviolet light was used for the first, heavy disinfection and chlorine kept for the last, light dose that has to hold through the pipes, which is the residual the disinfection article describes. And in the United States, where pipe networks are long and warm, many works switched the residual to chloramine, chlorine bound to ammonia, which is a weaker disinfectant and forms far fewer trihalomethanes, at the cost of a different set of byproducts and, in Washington in the early 2000s, the release of lead from old pipes that the lead article on this site describes.

The result is that the levels in most regulated supplies have fallen by half or more since the 1980s, and the supplies that remain high are the ones with peaty water, small works and long networks, which are the ones the European study pointed to.

The trade

The discovery of 1974 produced, in some places, the wrong response, and the water profession has told the story since as a warning. Chlorine stopped typhoid and cholera in the cities that adopted it, in the space of a decade, and the death rates from waterborne disease in the industrial world fell to nearly nothing by the 1930s because of it; the plain water articles on cholera and disinfection give the figures. A byproduct that raises the risk of one cancer by a fraction over a lifetime, at the levels of the 1970s, is a small harm set against the largest public health gain of the twentieth century, and the regulators of the 1970s and 1980s said so and set limits that kept the chlorine and cut the byproducts. Where the balance has been lost, in supplies that cut disinfection to reduce the byproducts or to save money, the disease has come back, which the Legionella and cholera articles on this site both record.

The rule the profession settled on, and that the World Health Organization writes at the head of its guidance, is that microbial safety comes first and byproducts second, and that a works must never reduce disinfection to reduce byproducts. The way to have both is to take the organic matter out. The way to have neither is to stop chlorinating, and that way is the one that kills.

The shower and the pool

The byproducts are not only drunk. Chloroform and its cousins are volatile, and a hot shower in a house with high levels releases them into the steam, where they are breathed in, and a study of exposure routes found that a ten minute shower could deliver as much as a day's drinking. The swimming pool is the extreme case: chlorine at pool doses reacting with the sweat, urine, skin and sunscreen of the swimmers makes the compounds at levels many times a tap's, along with the chloramines that give an indoor pool its smell and the swimmers their red eyes, and the studies of competitive swimmers and of pool attendants find the exposures, and some of the effects, that the drinking water studies find at lower levels. None of this argues for less chlorine in the pool, which is where the microbial risk is highest and most immediate; it argues for ventilation, for showering before swimming, and for the same principle as at the works, which is to give the chlorine less to react with.

Other disinfectants, other byproducts

Every disinfectant makes something. Ozone, which many European works use, reacts with bromide in the water to make bromate, which is regulated at ten micrograms a litre and is harder to remove than the trihalomethanes it avoids. Chloramine makes fewer trihalomethanes and more of a class of nitrogen containing compounds that the chemists rank as more potent by weight, at lower concentrations. Chlorine dioxide leaves chlorite. Ultraviolet light makes almost nothing and leaves no residual, which is why it is used at the works and cannot be used alone in a network. The choice a works makes is between byproduct families, on the chemistry of its own water, and the works that has removed the organic matter first has the least of any of them to worry about.

What it teaches

The chlorine byproducts are the plain water story about a fix that had a cost, found sixty six years after the fix, and managed since without giving up the fix. The chlorine reacts with the leaves in the river to make a compound that raises a small risk over a lifetime; the works answer by taking the leaves out first and adding the chlorine last; and the limit is set where the risk is small and the disinfection is whole. The tap water in a regulated supply carries a few tens of micrograms a litre of chloroform and no cholera, and that is the right side of the trade.

A hundred micrograms a litre, found in 1974, and a century of chlorine that was never the thing to give up.

Sources

1. Rook, J.J. (1974). Formation of haloforms during chlorination of natural waters. Water Treatment and Examination 23; Bellar, T.A., Lichtenberg, J.J. and Kroner, R.C. (1974). The occurrence of organohalides in chlorinated drinking waters. Journal AWWA 66.
2. US EPA, Stage 1 and Stage 2 Disinfectants and Disinfection Byproducts Rules: total trihalomethanes 80 micrograms per litre; five haloacetic acids 60.
3. European Union, Drinking Water Directive 2020/2184: trihalomethanes total 100 micrograms per litre; haloacetic acids 60.
4. Villanueva, C.M. et al. (2004). Disinfection byproducts and bladder cancer: a pooled analysis. Epidemiology 15; Villanueva, C.M. et al. (2020). Bladder cancer attributable to trihalomethanes in the European Union. Environmental Health Perspectives 128.
5. WHO (2022). Guidelines for Drinking Water Quality, 4th edition with addenda: chapter on disinfection byproducts and the priority of microbial safety.
6. Photographs: opener: Aerial view of Silicon Valley Clean Water treatment plant, September 2023 by Pi.1415926535 (CC BY-SA) via Wikimedia Commons; inline: Wentworth Falls water chlorination plant by 2hu4u (CC BY-SA) via Wikimedia Commons; inline: Indoor Swimming Pool in Melbourne VIC Australia by KeepActive Australia (CC BY-SA) via Wikimedia Commons.

The two taps in a kitchen are connected to different things. The cold one is connected, in most houses, to the main in the street, and the water that comes out of it is the water the utility treated and tested, with the residual chlorine still in it and a few hours old. The hot one is connected to a tank or a heater inside the house, and the water in it has been drawn from the same main and then held, at fifty or sixty degrees, in a steel or copper vessel, for hours or days, in contact with the vessel's lining, the sediment at its bottom, the anode that protects it from corrosion and every metre of pipe between the tank and the tap. The utilities and the plumbing codes of every country say the same thing: drink and cook from the cold tap. The myth is that it does not matter, and the reason it matters is the pipe rather than the heat.

This article is about what the hot water in a house has been through, why it carries more of the plumbing than the cold, what the tank grows if it is too cool, where the advice comes from and where it is out of date.

What hot water dissolves

Metal dissolves into water faster when the water is hot. That is the whole of the chemistry, and it applies to every metal a domestic system is made of: the copper of the pipes, the zinc and lead in brass fittings, the lead of old service pipes and old solder, the nickel of chrome plating. The lead article on this site describes how lead gets from a pipe into a glass, and the rate at which it does so roughly doubles for every ten to fifteen degrees of temperature, so that water at sixty degrees standing in a lead soldered copper system for a night picks up several times what cold water in the same pipe would. Copper behaves the same way, and a hot tap that runs blue green or leaves a green stain in the basin is a system dissolving itself into the water.

The tank adds its own. A steel tank is lined with glass or protected by a magnesium or aluminium anode that is designed to corrode instead of the steel, and the anode dissolves into the water as it does its job; the sediment at the bottom, which is the scale of the hard water article and whatever the mains delivered over the years, is stirred into the hot water when the tank is drawn; and an old tank with a failing lining sheds rust. None of it is dangerous at the concentrations an ordinary system produces. All of it is in the hot water and none of it is in the cold.

Where the advice comes from

The rule to drink from the cold tap is a century old, and its origin explains its force. In the houses of the early twentieth century, in Britain above all, the cold kitchen tap was fed directly from the main and everything else in the house, including the hot system, was fed from a cistern in the roof: an open tank, often uncovered, in which the water sat for days, collected dust, insects and, in the stories every plumber tells, dead birds, and from which the hot tank was filled. The kitchen cold tap was the one tap in the house connected to clean water, and the rule was to drink from it and nothing else. That system is still in millions of British houses, and where it is, the hot water has been through a loft tank before it was heated.

The American version of the rule comes from lead. The service pipes of the older cities, which the lead article describes, were lead, the solder was lead until the 1980s, and the federal agency's advice since the 1990s has been to run the cold tap before drinking and never to use hot water for drinking, cooking or, above all, making up a baby's bottle, because a baby's small body takes lead hardest and a bottle made with hot tap water is the easiest way to give it some.

The tank and Legionella

The second reason is alive. Legionella, which the article on it describes, grows in still water between about twenty five and forty five degrees, and a hot water tank is a vessel designed to hold still water at a temperature that, if the thermostat is turned down to save energy, sits in that range. The plumbing codes that specify sixty degrees for storage and fifty at the tap do so for that reason: sixty kills the bacterium in minutes. A household that has turned its tank down to forty five to save on the bill, and a system with long dead legs of pipe where water sits warm, are the conditions that the outbreak investigations find. The bacterium is a hazard in the shower, where it is breathed in as mist, rather than in the glass, and the advice to keep the tank hot is about the shower. The advice to drink cold is about the pipe. The two together are the reason the codes were written.

Where the advice is out of date

A new house with plastic pipe, lead free brass, a combination boiler that heats water as it passes and no tank at all produces hot water that is a few seconds old and has touched almost nothing. In that house the hot tap is, in practice, the cold tap with a heater on it, and filling the kettle from it is harmless, if wasteful of the energy the boiler spent. The problem is that the tap gives no sign of which house it is in, the plumbing of a house changes with every owner, and the rule that works everywhere is the one that assumes the worst. Utilities, which cannot inspect every house, write the rule for the house with the loft cistern and the lead solder, and it costs the new house a minute at the kettle.

There is also the boil. Water boils faster from the hot tap, which is the myth's whole appeal, and the saving is the energy the tank already spent heating it, which is usually more than the kettle's, because the tank kept the water hot while nobody used it. The fast boil is a slow one that started an hour ago.

What the taps are for

The kitchen has two taps because the water in them is for different things. The cold is for drinking, cooking and anything that goes into a body, from the pipe that leads most directly to the treated supply; the hot is for washing, from a system that is designed to be hot and clean enough for that and is not tested for anything else. A kettle filled from the cold takes a minute longer and carries whatever the utility tested for. A kettle filled from the hot carries the tank.

The rest of the house

The same reasoning sorts the other taps. The bathroom cold tap, in a British house with a loft cistern, is fed from the cistern and carries what the cistern holds, which is why the old rule named the kitchen tap and no other; in a house plumbed since the 1990s, with every cold tap on the main, the bathroom cold is as good as the kitchen's. The garden tap is on the main and is fine to drink from, if the hose is not, because a garden hose lying in the sun grows what a warm pipe grows and sheds what its plastic is made of. A water cooler or a fridge dispenser is a small tank with a filter, and it is the filter jug article's story in a colder box. The rule, everywhere, is the same one: the shortest path from the treated supply to the glass, through the fewest things and the least standing time, and the hot system is the longest path in the house.

What it teaches

The hot tap myth is the smallest on this site and it turns on the same point as the largest: the water in a glass is the pipe it came through. The utility's responsibility ends at the meter and the house's plumbing begins, and the hot system is the part of the house's plumbing that has had the longest and warmest contact with the water. The rule to drink cold was written for lead pipes and loft cisterns and it still fits millions of houses, and the house that it does not fit cannot be told from the tap.

Sixty degrees in the tank, several times the metal from the same pipe, and a rule a century old that still holds.

Sources

1. US EPA, Basic Information about Lead in Drinking Water: use only cold water for drinking, cooking and making baby formula; hot water dissolves lead more quickly.
2. UK Drinking Water Inspectorate and Water UK, advice on drinking from the cold kitchen tap; stored hot water and cistern fed systems.
3. Schock, M.R. and Lytle, D.A. (2011). Internal corrosion and deposition control. In Water Quality and Treatment, 6th edition. Temperature and metal release.
4. Health and Safety Executive (UK), HSG274 Part 2: Legionella in hot and cold water systems; storage at 60 degrees, distribution above 50.
5. Building codes: UK Water Supply (Water Fittings) Regulations 1999; US Uniform Plumbing Code on potable water and water heater temperatures.
6. Photographs: opener: Kitchen water tap 20190408 by Santeri Viinamäki (CC BY-SA) via Wikimedia Commons; inline: Rheem home gas water heater tank - top of tank by Infrogmation (CC BY-SA) via Wikimedia Commons; inline: Water running out of a faucet by Pixelmaniac (CC0) via Wikimedia Commons.

A litre of sunflower oil, on the arithmetic this site uses, carries about six and a half thousand litres of water, which is more than a kilogram of chicken and about the same as a kilogram of pork. It is a large number, and it is one of the least worrying on this site, for the reason the wool article gives: almost all of it is rain, and it fell on land that grows sunflowers because sunflowers are what grows there. The crop is the world's fourth vegetable oil, after palm, soy and rapeseed, and half of the traded supply came, until 2022, from two countries at war with each other.

This article is about where a sunflower's water goes, why an oil carries twice the water of its seed, how it compares with the other oils in the cupboard, what a crushing plant uses, and what a war showed about a crop whose water was never the constraint.

The crop

The sunflower is an American plant, domesticated on the plains of what is now the United States and taken to Europe in the sixteenth century as a curiosity, and it became a crop in Russia, where the Orthodox fast allowed its oil when it forbade most others, and where the black earth of the steppe and the hot dry summer suited it. Ukraine and southern Russia are still its heartland, growing about half of the world's seed between them, with Argentina, Romania, Bulgaria, Turkey, France and Hungary behind. It is a summer crop, sown in spring and harvested in September, and it is grown on rain: a tap root that reaches two metres down lets it draw on water stored in the soil from winter, and it will set seed in a summer that would fail a maize crop.

A hectare gives two to three tonnes of seed, and the seed is about forty five percent oil. The crushing plant recovers most of that and the cake that remains, high in protein, goes to animal feed, so that the water of the field is divided between the oil and the meal in proportion to their value, and the oil takes the larger share.

Where the water is

The sunflower's footprint is more than nine tenths green. The crop's need over a season is six hundred to a thousand millimetres, which the steppe's rain and stored soil moisture supply in an ordinary year, and the blue share, irrigation, is a few percent, mostly in the drier fields of Turkey, Spain and the American plains. The grey share, for the fertiliser and the pesticide that run off, is modest, because the sunflower is a lightly fertilised crop that follows a cereal in the rotation and uses what the cereal left. The number is large because the yield of oil per hectare is low compared with palm, which gives four times as much oil from the same land, and not because the plant is thirsty.

That is the argument the wool article makes and it applies here with one difference. The steppe is farmland, and the rain that grows sunflowers could grow wheat or maize instead, so the green water has an alternative use in a way that the rain on a Welsh hillside does not. The comparison that matters is with the other oils, and on that comparison the sunflower sits in the middle, using rain that falls on the driest of the oil growing regions and asking for almost nothing from a river.

The crushing plant

The seed is dried, dehulled, flaked, cooked and pressed, and the cake left by the press is washed with a solvent, usually hexane, to take out the last of the oil, and the solvent is recovered and used again. The plant's water goes into the steam that cooks the flakes and strips the solvent, into cooling, and into the refining of the crude oil, which is washed with water and a little alkali to take out the free fatty acids and the gums, and then bleached and deodorised. It is a few cubic metres of water per tonne of oil, a small fraction of the field's, and the effluent carries the soapstock and the gums, which are sold, and a fatty, high oxygen demand stream that a treatment plant of the kind the food article describes handles without trouble. A crushing plant on the Black Sea coast at Odesa or Mykolaiv, where most of Ukraine's oil was made and shipped, is a modest water user in a country with rivers.

The war

In February 2022 Ukraine and Russia between them supplied more than half of the sunflower oil traded in the world, most of it through the Ukrainian Black Sea ports, and within a month the ports were closed, the crushing plants near the front were shut, and the oil that Europe, India, Turkey and the Middle East fried with was on the wrong side of a blockade. Supermarkets in Britain and Spain limited customers to a bottle or two; India, the largest importer, switched to palm and soy; the price of every vegetable oil rose with it, because the oils substitute for each other in the fryer and the price of one is the price of all.

The water was never the question. The crop was in the ground, the rain had fallen, and the oil was in tanks at a port that ships could not leave. Exports resumed through a negotiated corridor in the summer, then by rail and river through Romania and Poland, and by 2023 Ukraine was shipping most of what it had grown, from a smaller area, with the mines and the front line having taken a share of the fields. The lesson for this site is one it makes elsewhere about virtual water: the water in traded food moves through ports, and a port is a narrower thing than a rain belt.

Sunflower against olive

The oil most often compared with sunflower in the European kitchen is olive, and the comparison runs the other way from the one the cook expects. Olive oil carries about twice the water per litre, and a larger share of it is blue, because the olive grows in the dry Mediterranean summer and the intensive groves of Spain, which make half the world's supply, are irrigated from rivers and aquifers that the Barcelona and Murray Darling articles show are already spoken for. A litre of Spanish olive oil in a dry year is a litre with a river in it. A litre of Ukrainian sunflower oil is a litre of rain on the steppe. The olive is the better oil by most measures a cook cares about; the sunflower is the better one by the measure this site keeps.

The field in a dry year

The sunflower's reputation for drought tolerance holds until it does not. The root finds the water the winter left in the soil, and in a summer with no rain at all the crop draws that store down and sets a smaller head with lighter seed, so that the yield falls by a third or a half and the water per kilogram of oil rises by the same. The droughts of 2012 and 2022 on the steppe did exactly that, and the growers in the drier parts of Spain, Turkey and Argentina who irrigate do so for the difference between a crop and a poor one, with a few hundred millimetres in July, which is the blue share in the table above. Sunflower is also the crop the steppe's farmers plant when the wheat has failed or the market has shifted, and the area swings by millions of hectares from year to year with the price, which is why the footprint, an average, describes no particular field. A wet year on the steppe gives oil at four thousand litres a kilogram; a dry one gives it at nine. The average in the tables, which is what a footprint is, lands between the two and describes neither.

What it teaches

Sunflower oil is the product on this site whose large number hides a small burden and whose supply broke for a reason that had nothing to do with water. The crop follows the sun across the driest farmland in Europe on rain and a deep root, its factory is a light water user, and its water is rarely anyone else's. The war showed the world how much of its frying depended on one coast, and the rain kept falling on the fields behind it.

6,800 litres a kilogram, nearly all of it rain on the steppe, and a port that mattered more than any of it.

Sources

1. Mekonnen, M.M. and Hoekstra, A.Y. (2011). The green, blue and grey water footprint of crops and derived crop products. Sunflower seed about 3,366 litres per kilogram; sunflower oil about 6,792; rapeseed oil 4,301; olive oil 14,431; palm oil 4,971.
2. FAOSTAT and USDA Foreign Agricultural Service, sunflower seed and oil production and trade: Ukraine and Russia over half of world exports before 2022.
3. FAO, Crop water requirements: sunflower 600 to 1,000 millimetres over the season; deep rooting and drought tolerance.
4. European Commission, Best Available Techniques for the food, drink and milk industries: oilseed crushing and solvent extraction, water use and effluent.
5. Ukrainian Ministry of Agrarian Policy and UkrAgroConsult, sunflower area and yields 2021 to 2023.
6. Photographs: opener: Sunflower field in Kosharka by Yakudza (CC BY-SA) via Wikimedia Commons; inline: Harvested sunflower field by CosyCobra (CC BY-SA) via Wikimedia Commons; inline: Sunflower oil bottles in Dnipro 2 by Alex Blokha (CC BY-SA) via Wikimedia Commons.

Sydney is the largest city in Australia and it drinks, for the most part, from a single dam. Warragamba, sixty kilometres west of the harbour, holds about two thousand billion litres behind a concrete wall across the Nepean gorge and supplies about four fifths of the city's water. When it is full the city has three or four years of ordinary use stored in the valley behind it. When it falls, the city has a problem with no second answer, and twice in twenty years it has fallen far enough that the answer had to be built.

This article is about what Sydney's water is, what the two droughts did, what the desalination plant at Kurnell was built for and why it spent most of its life switched off, and why that, rather than being a scandal, is the plan working.

One dam

Sydney's water comes from a ring of dams in the sandstone country south and west of the city, built between the 1880s and the 1960s as the city outgrew each in turn, and Warragamba, finished in 1960, is the last and by far the largest. The catchment is about nine thousand square kilometres of mostly forested and farmed country, closed to public access near the water as Melbourne's is, and the water is soft, low in dissolved solids and treated at Prospect, the largest water filtration plant in the southern hemisphere, by the coagulation and filtration the plain water articles describe.

The city uses about one and a half billion litres a day, of which households take the larger part at around two hundred litres a person. In an average year the dams catch more than that and the surplus spills. In a dry one they do not, and because the storage is so large and so concentrated, the city's fortunes rise and fall with one valley's rain.

The first drought

The Millennium Drought that emptied Melbourne's reservoirs did the same to Sydney's, more slowly. From 1998 the dams fell, and by February 2007 Warragamba was at about a third, with the city on restrictions that banned sprinklers and hoses on hard surfaces and, by the end, allowed a hand held hose on a garden only on two days a week. The government looked at the rate of fall, at a catchment that had had eight dry years, and at the forecast, and decided the city could not depend on rain returning in time.

The answer was a desalination plant at Kurnell, on the peninsula south of the harbour, drawing seawater from the Tasman and sending it through reverse osmosis at 250 million litres a day, with a wind farm built inland to offset its electricity. It cost about two billion dollars and it was finished in 2010, by which time the drought had broken, the dams were at ninety percent, and the plant's water was not needed. It ran for two years to prove itself and was put on standby in 2012.

Seven years idle

For seven years the plant sat at Kurnell doing nothing, and the argument about it never stopped. It was leased to private owners who were paid a standby charge whether it ran or not, which the newspapers reported every year as the cost of a white elephant. A tornado tore the roof off part of it in 2015 and the repairs took three years. The dams, meanwhile, stayed full, and the city's use, which the drought restrictions had cut from about 350 litres a person to about 200, stayed low, because the habits and the fittings had changed.

The rule that governed the plant was written into its contract: it would be switched on when total storage fell below sixty percent and off when it rose above seventy. The rule was the whole point. A desalination plant, as the desalination article on this site argues, is expensive water and it makes sense only when the alternative is no water; a plant that runs all the time in a city with a two thousand billion litre dam is wasting electricity, and a plant that runs only when the dam is low is doing what it is for.

The second drought

In 2017 the rain stopped again, and this time the dam fell faster than anyone had modelled. The catchment was dry and warm, the inflows were the lowest on record, and Warragamba went from full in 2016 to below sixty percent in January 2019, when the rule triggered and the plant was restarted. It reached full output within weeks and ran through the year. The city went on to restrictions in June 2019, the first in a decade, and to tighter ones in December, as the dam approached forty percent and the government began to plan for what it had never had to plan for before: the dam running out.

The plan, drawn up in the second half of 2019, was to double the desalination plant, to build a second one, to build the pipes to bring purified recycled water into the dams, and to cut use further. It was published as the drought reached its worst and then, in February 2020, the rain came: a week of it, filling Warragamba from about forty percent to over eighty in ten days and flooding the valley below. The restrictions were lifted, the plant ran on until the dams passed the threshold, and the plan was kept.

What the second drought taught

The lesson Sydney drew, written into its water strategy in 2022, was that its one dam is the problem and the plant is the answer. A catchment that can go from full to forty percent in three years, in a climate where the dry spells are getting longer and the inflows lower, cannot be the city's only source, and the only source that does not depend on rain is the sea. The plant at Kurnell is being doubled to five hundred million litres a day, a third of the city's use, and a second plant is being designed for the north. Purified recycled water, of the kind Singapore drinks and Orange County puts into its aquifer, is being planned for the dams, against the public reluctance that has stalled it in Australia before. And the wall of Warragamba, which the government proposed to raise for flood control, was left alone in 2023 after the cost and the drowning of the valley above were judged too high.

The plant's critics have gone quiet, because the seven idle years now read differently. A plant that was switched on within days when the dam fell, and that carried a sixth of the city through its worst year, cost the standby charge for exactly the reason the standby charge exists.

The water nobody wants to drink

The cheaper answer than a second desalination plant is the one Sydney already makes and throws away. The city's sewage works treat about a billion litres a day and discharge most of it to the ocean through three deep outfalls off the northern beaches, and the technology to purify a share of it to drinking standard and put it into Warragamba is the one the Orange County and Singapore articles on this site describe: membranes, reverse osmosis, ultraviolet light and a long rest in the dam. It costs less than desalination, because the water is already fresh, and it uses less than half the electricity. It has been proposed for Sydney in every drought since 2006 and rejected in every one, by governments that read the polling. Toowoomba, in Queensland, voted against it in 2006 by a margin of two to one, and the word that won the vote was the one the campaigners used on the posters. The 2022 strategy names purified recycled water as a source to be built by the 2030s and does not say when, which is how a government keeps an option without spending on it. Perth, the Australian city with the least rain, has been putting purified recycled water into its aquifer since 2017 and drinking it since, and nobody there has noticed.

What it teaches

Sydney is the city on this site that answers the argument the desalination myth article makes. Desalination is expensive, energy hungry and the wrong first answer for a city that can cut its use and catch its rain; and for a city of five million with one dam in a drying climate, it is the only answer that does not depend on the sky. The two are reconciled by the switch: a plant that sits idle in the wet years and runs in the dry ones, paid for as insurance and used as insurance. Melbourne built the same and left it idle for a decade. Cape Town nearly ran out for want of one. Sydney is doubling its own.

250 million litres a day, idle for seven years, and running when it counted.

Sources

1. WaterNSW, Greater Sydney dam levels and storage: Warragamba capacity 2,027 GL; system total about 2,600 GL. Lowest levels about 32 percent in February 2007 and about 42 percent in February 2020.
2. Sydney Water, Water Conservation Report and annual reports: residential use around 200 litres per person per day; total demand around 1,500 to 1,700 ML per day.
3. Sydney Desalination Plant, plant facts: 250 ML per day, commissioned 2010, on standby 2012 to 2019, restarted January 2019 when total storage fell below 60 percent.
4. NSW Government, Greater Sydney Water Strategy (2022): desalination expansion, purified recycled water, and demand reduction.
5. Bureau of Meteorology, Special Climate Statement 70: the 2017 to 2019 drought in eastern Australia.
6. Photographs: opener: Sydney Harbour Bridge & Opera House (20231001) (53322902670) by Takeshi Aida (CC BY-SA) via Wikimedia Commons; inline: Warragamba Dam (January 2014) by Goran Has (CC BY) via Wikimedia Commons; inline: Aerial view of Warragamba Dam, Silverdale and Wallacia by Bidgee (CC BY-SA) via Wikimedia Commons.

Nairobi has been on water rationing since 2017. Every part of the city, from the tower blocks of Upper Hill to the tin roofed lanes of Kibera, has its supply turned off by rota for one or more days a week, and the schedule, which was announced as a drought measure when the city's main dam fell to a fifth, was never lifted, because the drought was only the occasion. The city needs about 800,000 cubic metres of water a day and the utility can supply about 500,000, and the rota is how the difference is shared.

This article is about why a city at the foot of some of the wettest hills in East Africa is short of water, where the shortfall goes, what the tunnel under the Aberdares was built to fix, and why the people who pay most per litre are the ones with no pipe.

Where the water comes from

Nairobi's water is not from Nairobi. The city sits on a dry plateau at 1,700 metres, with a river through it that was never large enough to drink from and is now an open sewer, and its water comes from the Aberdare range, sixty kilometres to the north, where the rain that the plateau does not get falls on forested hills. The Ndakaini dam on the Thika river, finished in 1994, holds about seventy million cubic metres and supplies about five sixths of the city; an older dam at Sasumua and a few boreholes and springs supply the rest. The water is treated at Ng'ethu, on the way down, and carried to the city by pipe.

The dam was built for a city of two million. Nairobi County has four and a half million people and the metropolitan area has more, and the growth has been fastest in exactly the districts the pipe network reaches least.

Where it goes

Of the 500,000 cubic metres that enter the city's pipes each day, close to half is never sold. Some of it leaks from a network that is older than the dam, some is taken through illegal connections, some is delivered to customers whose meters do not work, and the regulator's reports have put the figure at between forty five and fifty percent for a decade. The leaks article on this site describes what non revenue water is and why it is hard to cut; Nairobi's is at the high end, and cutting it to thirty percent would give the city a fifth more water without a new dam. The utility knows this. The money to do it, which is money to dig up the pipes of a city that cannot stop for the work, has not been there.

The rest reaches taps, and the taps are unevenly placed. The suburbs of the west and north, where the city's wealth is, have piped water in every house and storage tanks on every roof, so that a day off the rota is a day the tank covers. The informal settlements of the south and east, where a third of the city lives on a few percent of its land, have a handful of standpipes and kiosks fed from the mains, and when the rota turns the mains off, the kiosks close.

The price of a jerry can

The person in Kibera buys water in twenty litre jerry cans from a kiosk or a vendor with a handcart, at a few shillings a can in an ordinary week and several times that when the rota has closed the kiosks and the vendors are the only supply. Worked out per litre, that price is five to ten times what a household in Karen or Muthaiga pays for metered water from its own tap, and it is paid by the people with the least. The pattern is the one Lagos and Karachi show on this site, and it is the same arithmetic: piped water is cheap because the pipe was paid for long ago, and the people without the pipe pay for the truck, the cart and the middleman instead.

The utility has tried to fix this from the middle. Prepaid dispensers, where a card is charged and water is drawn at the mains price, have been installed in some settlements and have cut the price where they work; the vendors, who lose their trade, have broken some of them. Kiosks run by community groups on a bulk tariff work where the mains reach and fail where they do not. The fix that works everywhere is the pipe, and the pipe is what the settlements were built without.

The tunnel

The supply side answer, planned since the 1990s and built through the 2010s, is the Northern Collector Tunnel: nearly twelve kilometres bored under the Aberdare foothills to take water from three rivers, the Maragua, the Gikigie and the Irati, and deliver it into the Thika river above Ndakaini, so that the dam refills in a dry season from catchments it was never connected to. The first phase, finished in stages from 2022, adds about 140,000 cubic metres a day, roughly a quarter more than the city had, with a new treatment plant to match. It was late, over budget and opposed by the farmers downstream of the three rivers, who argued, with reason, that water taken from their rivers to Nairobi's dam is water they no longer have.

It is also, on the arithmetic above, about a decade of growth. The city adds around a hundred thousand people a year, the demand rises with them, and the second phase of the tunnel, which would take more rivers, is not yet funded. Nairobi has built the largest water project in its history and closed perhaps half the gap the rota was created to manage.

What the rain does

The Aberdares are wet and the plateau is dry, and both are becoming less predictable. The long rains of March to May and the short rains of October to December fill the dam in a normal year, and the years since 2000 have included several in which one or both failed: 2017, when Ndakaini fell to about twenty percent and the rota began, and the drought of 2021 to 2023, the worst in the Horn of Africa in forty years, which the dam survived only because the tunnel's first water had arrived. In a wet year the dam spills and the city, with its pipes and its rota, cannot use the surplus. Storage is the missing piece, and the next dam, on the Thika's neighbour, has been at the planning stage since before the tunnel was dug.

The river through the city

Nairobi is named for its river, and the river is the other half of the water story. The Nairobi river and its tributaries run through the city from the Ngong hills to the Athi, and for most of their length they carry the sewage of the districts that have no sewer, the runoff of the industrial area and the waste of the markets on their banks, so that by the time the water leaves the city it is the kind the wastewater articles on this site describe as raw. About half the city is connected to a sewer, and the two treatment works at Dandora and Kariobangi were built for a smaller city and receive more than they can treat; the other half uses pit latrines, which fill and are emptied, when they are, into the same river. A city that brings in half a million cubic metres a day sends most of it out again, and the river is where it goes. The government has cleared the banks and planted them more than once, and the plan for the sewers, like the plan for the pipes into the settlements, is the part of the city's water that has waited longest.

What it teaches

Nairobi is the city on this site where the gap between demand and supply has been made permanent and administered, and where the cost of it has been shifted onto the people least able to pay. The rota is honest in a way that a leaking network pretending to be adequate is not, and it is also a decision to manage scarcity rather than end it. The tunnel closes half the gap for a decade. The leaks, if fixed, would close most of the rest. The pipe into the settlements, which is the only thing that would end the jerry can price, is the part of the plan that has waited longest.

500,000 cubic metres a day against 800,000, half of it lost, and a jerry can that costs the most where the pipe never came.

Sources

1. Nairobi City Water and Sewerage Company, water supply and rationing programme (2017 onward): production about 525,000 cubic metres a day against demand of about 810,000; Ndakaini (Thika) dam supplies about 84 percent.
2. Athi Water Works Development Agency, Northern Collector Tunnel Phase 1: 11.8 kilometres, 140,000 cubic metres a day, from the Maragua, Gikigie and Irati rivers.
3. Kenya Water Services Regulatory Board (WASREB), Impact Reports: non revenue water in Nairobi about 45 to 50 percent.
4. Water and Sanitation Program (World Bank), pricing of water from kiosks and vendors in Nairobi's informal settlements.
5. Kenya National Bureau of Statistics, 2019 census: Nairobi County 4.4 million; metropolitan area larger.
6. Photographs: opener: Nairobi skyline P1000021 by Lmwangi (CC BY-SA) via Wikimedia Commons; inline: Water in Kibera Slum by Mutungadavid (CC BY-SA) via Wikimedia Commons; inline: Ndakaini Dam 02 by Ahero dala (CC BY-SA) via Wikimedia Commons.

Istanbul is a city of sixteen million people built on the strait between two seas, and in the second week of January 2021 its water utility said, in public, that the reservoirs held about forty five days of supply. They were at 19.7 percent, the lowest level in fifteen years, after a year in which the winter rain had barely come. The imams were asked to pray for rain at Friday prayers, and the mayor asked the city to shower for less time, and a city that had never in living memory run dry looked at its taps. Then, in February and March, it rained hard for weeks, the reservoirs rose past half, and the story ended, for that year.

This article is about why a city on the sea, between two more, is short of fresh water, and why the answer has been the same for a hundred and fifty years: another reservoir, further away.

A city that outgrew its water

Istanbul had about a million people in 1950. It has sixteen million now, more than any city in Europe, spread along both shores of the Bosphorus and forty kilometres into the hills on either side, and almost all of that growth happened in the sixty years after the reservoirs that were meant to supply it were built. The Ottoman city drank from aqueducts and cisterns, some Roman, that brought water from the forests to the north; the republic built Terkos, a lake on the Black Sea coast, into a reservoir in the 1880s, and Ömerli, on the Asian side, in the 1970s, and a ring of smaller dams on both sides through the 1980s and 1990s. Each was sized for the city as it then was, and the city doubled before each was finished.

The catchments are small, low hills of forest and scrub within an hour of the city, and the rain that falls on them falls in winter: Istanbul's summer is long and dry, and from June to October the reservoirs only fall. A wet winter fills them. A dry one, of which there have been several since 2000, does not, and the city goes into summer with a deficit that the next winter has to make up. Two dry winters in a row, which the region's climate now produces more often than it did, leave the city with a deficit it cannot make up by rain at all, and that is what happened in 2020.

The catchments under the city

The reservoirs' other problem is that the city has grown over them. Ömerli's catchment on the Asian side was forest and farmland when the dam was built and is now suburb, with the roads, the sewers and the informal housing that follow sixteen million people, and the water that runs into the reservoir carries what a suburb sheds. The forests north of the city, which hold the Terkos and Alibeyköy catchments, have been cut by the third airport, the third bridge and the motorway that serves it, and the canal that the government proposes to dig from the Black Sea to the Marmara, parallel to the Bosphorus, would pass through the catchment of the Sazlıdere reservoir and take it out of use.

Every one of these is an argument the city has had, and lost, in the last twenty years. The reservoirs closest to the city, which are the cheapest water it has, are being spent on the city's growth, and the water is being brought from further away to replace them.

The pipe from the Melen

The answer, planned in the 1990s and built in stages since, is the Melen. The Büyük Melen is a river in the hills east of the city, two hundred kilometres away in the province of Düzce, and the plan was to take its water through a pipe and tunnel under the Bosphorus to both sides of Istanbul, at first from a weir on the river and then, when a dam was finished, from a reservoir behind it that would hold more than all the city's existing reservoirs together. The first stage opened in 2010 and the Melen became the city's largest single source, carrying, in a dry year, more than a third of what it drinks.

The dam was the problem. Built as a large rockfill structure in the 2010s, it cracked in its clay core before it was finished and the water behind it had to be drawn down while the core was rebuilt, and the completion date slid, year by year, through the second half of the decade and past the crisis of 2021. Without the dam the Melen system depends on the river's flow, which is high in winter and low in summer, and gives the city what the river has rather than what it needs. With the dam, the system is the insurance a city of sixteen million needs against a dry winter. The dam is now being finished, late, and it is the reason the utility says the next dry winter will be different.

The winter of 2021

The crisis of 2021 was a dry autumn after a dry year. The winter rain of 2019 and 2020 had been below average, the summer of 2020 had been hot, and by November 2020 the reservoirs were at a third and falling, with the rain that should have started in October still absent. Through December and into January the level fell by a fraction of a percent a day, and the utility, which publishes the figure every morning, watched it go through thirty, twenty five and twenty. The mayor said in early January that the city had about forty five days of water at the rate of use, and the calculation was straightforward: the reservoirs held about 170 million cubic metres and the city was drawing about three million a day, with the Melen river's winter flow keeping the figure from being worse.

What the city did was mostly ask. Use fell a little. There was no rationing by district, because the utility judged that the rain would come before it was needed, and it did: a wet February and a wetter March lifted the reservoirs past fifty percent by April, and by summer the crisis was a memory. The city had come closer than it had ever come, and the reason it did not run dry was the weather. The utility has since said that the forty five day figure was conservative, because the Melen river could have been drawn harder and the wells around the city brought in, and that the real margin was perhaps twice that. Ninety days of water for sixteen million people is the margin of a city that has no margin.

What comes next

Istanbul's plan for the next dry winter is the same as its plan for the last one: finish the Melen dam, connect more of the rivers to the east, and cut the leakage, which the utility puts at around a fifth of what it supplies and which is worth, on its own, a reservoir. Desalination has been discussed and set aside, because the city sits between two seas but has no shortage of rivers within two hundred kilometres, and a pipe is cheaper than a membrane. Demand is the harder half. A city that adds a quarter of a million people a year, most of them in apartments with modern fittings, uses more each year regardless of what each person does, and the utility's forecast is that the Melen, at full development, buys the city about two decades.

The catchments are the part of the plan no one controls. A canal through one, an airport in another and a suburb over a third are decisions made for other reasons, and each takes a piece of the cheap water the city had. The reservoirs will be refilled by the winters that are wet enough. The question, in a warming eastern Mediterranean where the dry winters are becoming more frequent, is how many of those there will be.

What it teaches

Istanbul is the city on this site that shows what growth does to a water supply. Every reservoir built for it was the right size for a city that had already been left behind, and the response, each time, has been a longer pipe to a further river. The winter of 2021 was the arithmetic catching up: sixteen million people, a dozen shallow reservoirs, and a dry season that lasted three months longer than the storage did. The rain came. The city is finishing a dam that should have been finished before the rain was needed.

19.7 percent, forty five days, and a dam two hundred kilometres away that was not yet ready.

Sources

1. İSKİ (Istanbul Water and Sewerage Administration), daily reservoir occupancy data: 19.7 percent on 14 January 2021; capacity about 868 million cubic metres.
2. İSKİ, annual reports: daily supply about 3 million cubic metres; the Melen system's share; non revenue water.
3. Turkish State Hydraulic Works (DSİ), Greater Melen project documents: 190 kilometre transmission line, capacity 1.18 billion cubic metres a year at full development.
4. Turkish State Meteorological Service, precipitation records for Istanbul 2020 and 2021.
5. World Bank and OECD reviews of Istanbul's water supply and urban growth.
6. Photographs: opener: Istanbul as seen from the Galata Tower (8396801424) by Jorge Láscar (CC BY) via Wikimedia Commons; inline: Ömerli Barajı ve Gün Batımı by Emregaznevi (CC BY-SA) via Wikimedia Commons; inline: Terkusu Gölü (Terkos) - btv1b10100798w by Bibliothèque nationale de France (Public domain) via Wikimedia Commons.

A semiconductor fab is a factory the size of several football pitches in which the air is filtered a thousand times cleaner than an operating theatre, and the thing it uses most of, after electricity, is water. A 300 millimetre wafer of silicon, on its way to becoming a few hundred chips, passes through several hundred steps of depositing, etching, polishing and cleaning, and between most of them it is rinsed in water of a purity that the ultrapure water article on this site describes: water with almost nothing left in it, made at the fab from the town supply at a cost of a litre and a half in for every litre out. A large fab uses thirty to fifty thousand cubic metres a day. The largest chipmaker in the world, in Taiwan, uses about 150,000 a day across its plants, which is the water of a city of half a million people.

This article is about what a fab does with that water, what happened when the rain that supplies it stopped, and why an industry that needs water this badly is building its next plants in deserts.

What a wafer does with water

The chips in a phone are built up in layers on a disc of polished silicon, each layer patterned by light, etched, filled and polished flat, and each step leaves something on the surface that the next step cannot tolerate: a residue of etchant, a film of polishing slurry, particles a few nanometres across. Between steps the wafer is cleaned, in baths and sprays of ultrapure water and chemicals, and then rinsed in ultrapure water alone until the water coming off it is as pure as the water going on. A modern chip has sixty or more layers of patterning and each has several cleans; the count of rinses per wafer runs into the hundreds. The rinse has to be ultrapure because the chips' features are now a few nanometres across, smaller than most of what a town's water carries, and a single particle or a single ion of the wrong metal left on the surface can short a transistor. The purity is a manufacturing tolerance, and the water plant is part of the production line.

Ultrapure water is made at the fab in a plant that is a water treatment works in miniature and in reverse: reverse osmosis, degassing, ion exchange, ultraviolet light and fine filtration, taking the town's drinking water and stripping it of everything the town leaves in. The plant rejects about a third of what it takes in as concentrate, so a fab's water intake is larger than its ultrapure demand, and the rest goes to cooling towers, to the scrubbers that clean the exhaust, and to the plant itself.

Where the fabs are

The world's leading edge chips are made in a handful of places, and the largest concentration is in Taiwan, in the science parks at Hsinchu, Taichung and Tainan, on an island whose rain falls in the typhoon season and is stored in reservoirs that silt up faster than they can be dredged. The fabs have first claim on the reservoirs in law and practice, because the chips are the island's largest export and its strategic guarantee, and in an ordinary year the reservoirs hold enough for the fabs, the cities and the rice paddies of the western plain. In 2021 there was no ordinary year.

The rest are in South Korea, in Japan, in the United States, in Singapore, in Israel and in Germany, and the pattern in each is a cluster of fabs in a place that was chosen for its engineers and its politics rather than its rain. The new fabs, announced since 2020 with subsidies from every government that wants chips made at home, are in Arizona, Texas, Ohio, Kumamoto in Japan, Dresden and, in the largest new cluster of all, in the dry west of the United States, where the water is the question every plan has to answer first.

Taiwan, 2021

In 2020 no typhoon made landfall on Taiwan for the first time in fifty six years, and the reservoirs that the typhoons fill went into the dry season low. By the spring of 2021 the Baoshan reservoirs that supply Hsinchu were below ten percent and the island was in its worst drought in half a century. The government did what its water law allows: it suspended irrigation on about seventy four thousand hectares of rice paddy in the west, compensated the farmers, and kept the water for the cities and the fabs. When the mains could not keep up, the chipmakers hired fleets of tankers, dozens of trucks a day per fab, drawing from wells and construction sites, and at the height of the drought the largest company was trucking in a tenth of what it used. The fabs did not stop. The rice, for a season, was not planted.

Rain returned in June and the reservoirs refilled, and the drought became the reason for the recycling and reclaimed water plants that every fab on the island has since built or expanded. It is also the year in which the world noticed that the chips in everything it buys depend on the rain in one island's hills.

What recycling means in a fab

The rinse water that comes off a wafer is, by any ordinary measure, still very clean, and the industry has learned to sort it. Lightly used rinse water is collected and sent back through the ultrapure plant, which takes it more easily than town water; the concentrate from reverse osmosis goes to cooling towers and scrubbers, which do not need purity; and the truly dirty streams, the acids, the slurries and the solvents, are treated separately and, at the best plants, largely recovered. The leading fabs in Taiwan now reuse between eighty five and ninety percent of the water they take in, by their own accounting, and the science parks have built plants that take the city's treated sewage and polish it to the fab's intake standard, the NEWater route that Singapore's fabs have drunk from for twenty years.

The limit is the same as for any recycled water. Every pass concentrates what is left, and the last stream, the brine from the last membrane, has to go somewhere. At a coastal fab it goes to the sea. At an inland one it is evaporated to a solid and landfilled, at the cost the zero liquid discharge article on this site describes, which the chip price can bear and few other products could.

The fabs in the desert

The new American fabs are in Arizona, Texas and Ohio, and the largest are in the Phoenix suburbs, which the Phoenix article on this site describes as a city living on a river already spoken for and an aquifer being drawn down. A fab there is a water user the size of a suburb in a place where the suburbs are told to stop growing, and the companies building them have made the same promise: near total recycling, a reclaimed water plant of their own, and a net return of water to the aquifer through the purchase and restoration of water rights elsewhere in the basin. Whether the promise holds will be known in the first dry decade. The reason the fabs are there anyway is that the water was never the first question; the engineers, the subsidies and the politics were, and the water was engineered to fit. The same is true of the fabs planned for the Gulf and for India, where the water will be desalinated or drawn from rivers already shared with farms, and where the choice Taiwan made in 2021 has been written into the permits before the first wafer is cut. A fab's water is the one industrial demand that a government will meet before every other, and the industry knows it.

What it teaches

The semiconductor fab is the industry on this site that uses water most intensely and has done the most to use it again, because it has both the need and the money. Its water is a city's, made pure at a cost most industries could not pay and recycled at a rate most could not reach, and it is still, on an island in a dry year, water that a rice farmer did not get. Taiwan's choice in 2021 was made in public and it will be made again, there and in Arizona, and the answer will be the chips, because that is what the water was allocated for.

150,000 cubic metres a day, ninety percent of it used twice, and a rice crop that was not planted so that it could be.

Sources

1. TSMC, Sustainability Report 2022 and 2023: total water withdrawal, process water recycling rate above 85 percent, water use per wafer layer, and the 2021 drought response.
2. Taiwan Water Resources Agency, reservoir levels 2021: Baoshan and Baoshan Second below 10 percent; irrigation suspended on about 74,000 hectares.
3. Frost, K. and Hua, I. (2019). Quantifying spatiotemporal impacts of the interaction of water scarcity and water use by the global semiconductor manufacturing industry. Water Resources and Industry 22.
4. Semiconductor Industry Association and SEMI, water use benchmarks per wafer and ultrapure water specifications (SEMI F63).
5. Intel and Micron water stewardship reports: reclaimed water at Arizona and Idaho sites.
6. Photographs: opener: 5C2A5953R - 49913961083 – Silicon Wafer 20200519 by Rob Bulmahn (CC BY) via Wikimedia Commons; inline: TSMC Fab 18 May 2025 by 4300streetcar (CC BY) via Wikimedia Commons; inline: Hsinchu paoshan 2nd reservoir by Liaon98 (CC BY-SA) via Wikimedia Commons.

An oil well is, for most of its life, a water well that also produces oil. Oil and gas sit in porous rock alongside the water that was there first, seawater trapped when the rock was laid down and made saltier since, and when the well is opened the water comes up with the oil. In a new field the oil share is high; in an old one, after decades of pumping and of water injected to push the oil toward the wells, the water share is nine tenths and more. Worldwide the industry lifts about three barrels of water for every barrel of oil, roughly 250 million barrels of water a day, and it has to do something with all of it.

This article is about the water at every stage of oil and gas: the water that comes up with it, the water pushed down to get it, the water a refinery uses to turn it into fuel, and what happened when one state disposed of its water too fast.

Produced water

The water that comes up with oil is called produced water, and it is among the worst water on this site. It is typically several times saltier than the sea, sometimes ten times, with dissolved oil, benzene and other hydrocarbons, heavy metals, the chemicals added down the well, and, in some formations, naturally occurring radium leached from the rock. It comes up hot and under pressure. It cannot be discharged to a river, cannot be used on a field, and cannot be drunk, and the industry's answer for a century has been to separate the oil from it and put it back underground.

Most of it goes back into the oil bearing rock it came from, injected through wells at the edge of the field to push the remaining oil toward the producing wells, which is called waterflooding and is the reason old fields produce so much water: what is injected comes back round. The rest goes into disposal wells drilled into deep formations that hold nothing anyone wants, where it stays. A small share, in a few places, is treated and reused. The United States alone produces about three and a half billion litres of it a day.

The water that goes down first

The wells drilled in the last fifteen years in the United States, and increasingly elsewhere, are made with water. Hydraulic fracturing drills a well sideways through a layer of shale for two or three kilometres and then pumps water, sand and a small share of chemicals into it at pressures high enough to crack the rock, the sand holding the cracks open so that oil and gas can flow. The water per well was about ten million litres in 2011 and, as the wells grew longer and the fracturing more intense, rose several fold, so that a well in the Permian basin of West Texas now takes fifteen to forty million litres and the largest more. A single well pad with a dozen wells uses the annual water of a small town, in a week, in a desert.

The water comes from where it can. In Texas it comes from aquifers, from farmers who sell their allocation, and from rivers when there are any; in Pennsylvania from streams and the town supply. A share of it, ten to forty percent, comes back up in the first weeks as flowback, mixed with the formation's own produced water, and the rest stays in the rock. The industry has learned to recycle the flowback for the next well, and in the Permian the reuse of produced water for fracturing has grown to a large share, because it is there and the aquifer is not.

Oklahoma

The water has to go somewhere, and in Oklahoma in the 2010s it went into disposal wells drilled into the Arbuckle formation, deep limestone that lies on the crystalline basement rock. The volume rose with the drilling boom, to more than a billion barrels a year, and the pressure in the formation rose with it, and the pressure reached faults in the basement that had been stable for millions of years. Oklahoma had recorded a couple of earthquakes of magnitude three or more in an average year before 2009. In 2015 it recorded about nine hundred, more than California, including several above magnitude five that damaged buildings in Cushing, where the country's oil is stored, and in Pawnee and Prague.

The cause was not disputed for long; the maps of injection volume and of earthquakes were the same map. The state ordered injection cut in the areas of highest activity, by about forty percent, and the earthquakes fell over the following years, though they have not stopped. The same has been seen, at smaller scale, in Texas, Kansas, Alberta and the Netherlands, where the gas field at Groningen was shut in 2024 because of the earthquakes its extraction caused. Water put into deep rock does what water in rock does: it lubricates.

The refinery

The refinery is the industry's other water user, and it uses water the way a chemical plant does: for cooling, above all, and for the steam that drives the distillation and the crackers, and for washing the salt out of the crude before it is heated. A refinery uses between one and two and a half barrels of water for every barrel of crude, most of it cooling, and its effluent carries oil, sulphide, ammonia and phenols that a treatment plant of the kind the effluent article describes takes out before discharge. Refineries sit on coasts and rivers for the same reason ammonia plants do, and the older ones, in the American Gulf and in the ports of Europe and Asia, have been the source of some of the oil in the harbours around them.

Offshore

The platforms in the North Sea, the Gulf of Mexico and off Brazil and West Africa produce water too, in the same ratio, and they have nowhere underground to put it that is cheap to reach. Offshore produced water is separated from the oil on the platform, in vessels and hydrocyclones that bring the oil content down to a few tens of milligrams per litre, and discharged to the sea, under a limit that in the North Sea is thirty milligrams of oil per litre and falling. The volume is enormous: the North Sea fields discharge hundreds of millions of tonnes of produced water a year, warm and salty and carrying a trace of oil and the additives, into a sea whose fish are caught and sold. The studies that have looked for harm have found it close to the platforms and not far away, and the regulators have pushed the limit down and the reinjection up, so that the newer fields put their water back into the reservoir even offshore. The oil in the sea from produced water, year on year, is larger than the oil from the tanker spills that make the news, and it arrives at a milligram a litre rather than a slick.

The water in a tank of fuel

Added up, from the well to the pump, a litre of petrol carries a few litres of fresh water: a fraction of a litre in the refinery, a litre or two if the crude came from a fracked well, and a little in the pipeline and the depot. It is one of the smaller numbers on this site, a hundredth of a litre of milk's, and it is the reason oil rarely appears in the tables of thirsty products. The produced water, which is many times larger, is left out of those tables because it was never fresh, and the water that a spill or an injection well spoils is left out because it was never used. Both belong in the account, which is the point of this article.

What it teaches

Oil and gas is the industry on this site that handles more water than almost any other and is rarely counted as thirsty, because most of its water was never fresh and most of it goes back underground. The three barrels with every barrel of oil are ancient seawater, returned to the rock; the fracking water is fresh, and it is the part that competes with a town or a farm; the injected water is the part that moves faults. The industry's water problem is one of disposal rather than supply, and it has been solved, for a century, by putting the water somewhere deep and hoping the rock is quiet. In Oklahoma it was not.

Three barrels of water for every barrel of oil, forty million litres down a well, and nine hundred earthquakes in a year.

Sources

1. Veil, J.A. (2020). US produced water volumes and management practices in 2017. Ground Water Protection Council. About 3.8 billion litres a day in the United States; about 3 barrels of water per barrel of oil.
2. Kondash, A.J., Lauer, N.E. and Vengosh, A. (2018). The intensification of the water footprint of hydraulic fracturing. Science Advances 4. Water per well up to 770 percent higher between 2011 and 2016.
3. Ellsworth, W.L. (2013). Injection induced earthquakes. Science 341; Oklahoma Geological Survey, earthquake counts 2009 to 2019.
4. Fakhru'l Razi, A. et al. (2009). Review of technologies for oil and gas produced water treatment. Journal of Hazardous Materials 170. Global produced water about 250 million barrels a day.
5. US EPA, Effluent guidelines for petroleum refining; refinery water use about 1 to 2.5 barrels per barrel of crude.
6. Photographs: opener: LostHillsPumpjacksSunset by Arne Hückelheim (CC BY-SA) via Wikimedia Commons; inline: Secretary Doug Burgum fracking site visit Pittsburg, Pennsylvania on April 3, 2025 - 4 by US Department of the Interior (Public domain) via Wikimedia Commons; inline: Preemraff Lysekil oil refinery on a foggy night by W.carter (CC0) via Wikimedia Commons.

About half of the nitrogen in the protein in a human body came through an ammonia plant. The nitrogen in the air is inert, three quarters of every breath and useless to a plant, and until 1913 the only ways of turning it into the form a crop can use were bacteria, lightning and the mining of guano and nitrate beds. Then Fritz Haber and Carl Bosch found how to join it to hydrogen at high pressure over an iron catalyst, and the fertiliser that resulted, spread on the world's fields through the twentieth century, is the reason the world's population could quadruple. The world now makes about 180 million tonnes of ammonia a year, and it uses water twice to do it: once as an ingredient, and then, in far larger quantities, as a coolant.

This article is about the water in an ammonia plant, why the plants are where they are, what changes if the hydrogen comes from water instead of gas, and why the water that ammonia affects most is the water it ends up in.

The recipe

Ammonia is one nitrogen atom with three hydrogens. The nitrogen is taken from the air, which is free. The hydrogen has to be made, and at almost every plant in the world it is made by steam reforming: natural gas, which is mostly methane, is mixed with steam at high temperature over a catalyst, and the methane and the water are both broken up, the carbon leaving as carbon dioxide and the hydrogen from both molecules going forward to the synthesis loop. The arithmetic of the reaction is that about half of the hydrogen in the ammonia came from the methane and about half from the water. In round numbers, a tonne of ammonia contains the hydrogen of about a tonne of water.

That water has to be pure, because it becomes steam in a boiler and passes over a catalyst that scale would ruin, so the plant makes it by the demineralisation the power station article describes, and it makes rather more than it uses because the steam also drives the compressors that push the gas to two hundred atmospheres. The boiler feed water is the one stream in the plant that is treated to a standard, and the plant's water chemist spends most of the day on it, because a boiler tube that scales or corrodes at those pressures fails without warning. Everything else the plant does with water is cooling, and cooling water needs only to be wet and cheap. A plant making a million tonnes a year, which is a large one, makes a few thousand tonnes of demineralised water a day.

The heat

The larger water use is heat. Making ammonia is a sequence of hot and cold steps: the reformer runs at nine hundred degrees, the shift and carbon dioxide removal steps at a few hundred, the synthesis at four hundred and fifty, and between each the gas has to be cooled, and the ammonia itself is condensed out of the loop by chilling. All of that heat, after what can be recovered as steam, is carried away by water, and the quantity is large enough that every ammonia plant in the world is on a river, an estuary or a coast, or has a cooling tower field beside it that evaporates five to fifteen tonnes of water for every tonne of product.

Once through cooling takes river water in, warms it, and returns it, and the water is not consumed; the effect is on the river's temperature, which the power station article discusses. Towers consume the water and return nothing. The plants in dry places, in the Gulf, in the American plains and in western India, use towers, and the water they evaporate is water drawn from the same aquifers the fertiliser they make will later reach.

Where the plants are

Ammonia is made where gas is cheap. The largest producer is China, which makes a third of the world's supply, much of it from coal rather than gas, by a process that uses more water and far more carbon; then Russia, the United States, whose plants were built on the shale gas of the last fifteen years, India, and the Gulf, where gas that would otherwise be flared is turned into ammonia and urea for export. The plants are large, a million tonnes a year and more, and they sit at the junction of a gas pipeline, a source of cooling water and a port or a railway.

The effluent is small and specific. The condensate from the process carries ammonia and a little methanol and is stripped and reused; the cooling tower blowdown carries the salts the tower concentrated; and an accidental release of ammonia itself, which is toxic to fish at a fraction of a milligram per litre, is the risk the plants are regulated for. A well run ammonia plant discharges little, and what it discharges is warm. The plants that are not well run are a different matter. Ammonia complexes in the older industrial districts of China, India and the former Soviet Union have been the source of some of the largest fish kills on record, when a leak or a flood sent a few tonnes of ammonia into a river, and the storage of ammonium nitrate, the fertiliser made from ammonia, is the reason for the explosions at West in Texas in 2013 and at the port of Beirut in 2020. The water in the plant is a small story. The chemistry the plant makes is a large one.

Green ammonia

The industry's carbon problem, which is about two percent of the world's emissions, has produced a plan to make the hydrogen from water alone. Electrolysis splits water into hydrogen and oxygen with electricity, and the hydrogen article on this site gives the arithmetic: nine litres of water for every kilogram of hydrogen, so about a tonne and a half of water in every tonne of ammonia, as an ingredient, with no methane and no carbon dioxide. The projects announced for the 2020s, in Saudi Arabia, Oman, Australia, Chile and Namibia, are all in deserts, because that is where the solar power is, and they will make their water by desalination, which adds the brine problem and a little energy to a process that is mostly electricity anyway.

Green ammonia would use more water as an ingredient than the gas route and about the same in cooling, and it would be made in places with none, from the sea. The water is a small part of the cost and the desalination plant is a small part of the project, and neither is the obstacle; the electricity is. The interesting question is what the plants will be for. Half of the projects are designed to make ammonia as a way of shipping hydrogen, to be cracked back into hydrogen at the other end for fuel, and the other half to make fertiliser, and the fertiliser will do what fertiliser has always done.

The water it ends up in

The water that ammonia affects most is the water it ends up in. Most of the world's ammonia becomes urea and ammonium nitrate, and most of that is spread on fields, and about half of what is spread is taken up by the crop. The rest goes into the air as ammonia and nitrous oxide, and into the soil as nitrate, which the rain carries down into aquifers and sideways into rivers, at the rates and with the consequences the nitrate and nitrogen articles on this site describe: wells above the drinking limit across every intensive farming region, rivers feeding the algae and the dead zones at their mouths, and a limit set by the blue baby cases of the 1940s that the fertiliser of the 1970s made ordinary. A tonne of ammonia has a tonne of water in it and a few tonnes more in the cooling; the nitrate it becomes can push the concentration in a cubic kilometre of groundwater past the limit.

What it teaches

Ammonia is the industry on this site whose water use is almost a technicality and whose water effect is among the largest of any. The plant splits water to make it, cools with a river to condense it, and discharges little; the fertiliser it becomes is in the wells of every farming district on Earth. The green version would take its water from the sea and put no carbon in the air, and it would make the same fertiliser. The nitrogen cycle, once opened, does not close at the plant gate.

180 million tonnes a year, a tonne of water in each, and nitrate in the wells where it lands.

Sources

1. International Fertilizer Association, ammonia production statistics: about 180 to 185 million tonnes a year; China, Russia, the United States, India and the Middle East the largest producers.
2. IEA (2021). Ammonia Technology Roadmap: energy use, emissions, and the water and electricity needs of electrolytic ammonia.
3. Erisman, J.W. et al. (2008). How a century of ammonia synthesis changed the world. Nature Geoscience 1. About half of the nitrogen in human protein from Haber Bosch.
4. US Geological Survey, Nitrogen (fixed) Ammonia, Mineral Commodity Summaries.
5. European Commission, Best Available Techniques reference document for the manufacture of large volume inorganic chemicals: ammonia plant water use and effluent.
6. Photographs: opener: Urea process plant UFFL 01 by Mar11 (CC BY-SA) via Wikimedia Commons; inline: FERTILIZER PLANT NEAR EL CENTRO - NARA - 548851 by Charles O'Rear (Public domain) via Wikimedia Commons; inline: Urea process plant UFFL 02 by Mar11 (CC BY-SA) via Wikimedia Commons.

A cannabis plant grown outdoors in the hills of northern California drinks, at the height of summer, about twenty two litres of water a day. That figure comes from a study the state's wildlife agency did in 2015, when the plant was still illegal to grow and the grows were hidden in the forests of Humboldt and Mendocino counties, and it made the crop, per plant, one of the thirstiest in the state: about twice a grapevine, and, over a season, thousands of litres for a plant that yields about half a kilogram to a kilogram of dried flower. The total, for a crop grown on a few thousand hectares, was small. The timing and the place were the problem, and they are the reason the crop is on this site.

This article is about what a cannabis plant uses, why the number per kilogram is high and the number per crop is not, what the hidden grows did to the streams they drew from, why indoor growing trades water for electricity, and what legalisation changed.

The plant

Cannabis is a large, fast growing annual, planted in spring and harvested in October, that in a hot dry summer puts on a metre of height a month and a canopy that transpires like a small tree. Outdoors in California it is grown in beds or large pots, watered by drip or by hand, and its need rises through the summer to the twenty two litres a day the study measured, roughly six millimetres a day over the area a plant covers, which is what a tomato or a maize plant uses in the same heat. The number is a plant's, and a plant is large. Per hectare, at a few thousand plants, cannabis uses about what a vineyard or an orchard uses, and less than an alfalfa field.

Per kilogram of product it is high, because the product is a small part of the plant, a few hundred grams of dried flower from a plant that weighed several kilograms wet, and the arithmetic is the one the wine article does for grapes: thousands of litres per kilogram of a crop sold by the gram.

The streams

The grows of the illegal years were in the hills of the Emerald Triangle, the three counties of northern California where the redwood and Douglas fir forest is cut by small steep streams that run high in the winter rains and fall, by August, to a trickle. Those streams are where the coho salmon and the steelhead spawn, and they were already in trouble from logging and roads. The grows were on cleared patches in the forest, out of sight, and their water came from the nearest stream through a pump and a black plastic pipe, in August, when the plants needed most and the stream had least.

The 2015 study mapped the grows from aerial photographs in four watersheds, counted the plants, multiplied by the daily need, and compared the result with the streams' measured summer flow. In three of the four the grows' demand exceeded the entire flow of the stream in the low months, and the study's authors had found, in the field, stretches of stream that had been pumped dry with dead fish in the pools. The grows also left fertiliser, rodent poison and diesel from the generators, and the streams carried those too. A later study of the same region found that groundwater pumping for cannabis and for the houses that came with it was lowering the summer flow of streams that had never had a pump in them.

Indoors

The other way to grow the plant is under lights, in a warehouse or a basement, where the water use is lower because the climate is controlled and the runoff is caught, and the electricity use is enormous. A study in 2012 put indoor cannabis at about one percent of the electricity used in the United States, for the lamps, the air conditioning that removes their heat, and the dehumidifiers that remove the water the plants transpire, which is then, in the best set ups, collected and reused. A kilogram of indoor flower carries a small water footprint and a carbon footprint of a few tonnes, which is the trade that the data centre article on this site describes in another form: evaporate the water outdoors or burn the fuel to keep it in.

What legalisation changed

California legalised cannabis for adult use in 2016 and the water rules came with the licences. A grower now needs a permit that specifies the water source, and a stream diversion is allowed only in the winter months, when the flow is high, with the water stored in tanks or ponds on the property for the summer; pumping from a stream in the dry season is forbidden. The rules pushed the crop out of the forest and onto farmland in the valleys, where the water comes from a well or a district with a right to it, and where the grows are metered and inspected. The hidden grows did not vanish, because the illegal market did not, and the enforcement raids of each summer still find pumps in the streams. But the licensed crop, which is most of what is sold, now draws on stored winter water or on farmland rights, and the salmon streams have one fewer pump in them.

The same pattern has followed legalisation elsewhere. Colorado, Oregon and Washington wrote water rules into their licences; Canada's crop is mostly indoors and its footprint is electric. The countries that grow cannabis illegally on a large scale, Morocco's Rif above all, grow it on rain in the hills and, increasingly, on wells in the plains, and the wells are falling.

The Rif

The largest cannabis growing region in the world is the Rif, the mountains of northern Morocco, where the crop has been grown for centuries on rain in the hills and where it is, since 2021, legal for medical and industrial use and still mostly illegal for the rest. The traditional crop, a small local variety sown in spring and cut in late summer, was grown on terraces on the winter rain alone, and it fitted the mountain: a low yield, a long season, no pump. The hybrid varieties that arrived from Europe in the 2000s yield several times as much and need water through the summer, and the growers who planted them dug wells and ran pumps, so that the streams of the Rif, which fed the coastal towns and the orchards below, now run low in August the way the streams of Humboldt did. The government's legalisation law favours the local variety for its lower water need, and the market, which pays for the hybrid, does not. It is the pattern this site sees in every crop that moves from rain to a well: the yield goes up, the season gets longer, and the water that was falling on the hillside for free is replaced by water that somebody downstream was counting on.

The number per gram

A gram of dried cannabis flower, on the outdoor figures above, carries three to five litres of water, about what a gram of coffee carries and a fraction of what a gram of beef does. Sold by the gram, it is a crop whose water hardly registers against its price, which is why the growers who could pay for water never economised on it and why the ones who stole it from a stream had no reason to stop. The tomato article on this site makes the point in reverse: a product cheap by the kilogram, grown on the same water, gets a greenhouse and a recycling system because the water is a visible share of the cost. Cannabis got a pump and a pipe because it was not. The legal price has since fallen by more than half in California, and the licensed growers who now count every cost have moved, as the tomato growers did, toward drip lines, moisture sensors and stored winter water, because the water is now a line on a spreadsheet rather than a stream behind the trees.

What it teaches

Cannabis is the crop on this site that shows the difference between a footprint and a harm. Its water per plant is high and its water per crop is small, and neither number is the reason the fish died. The reason was a pump in a shrinking stream in August, in a place chosen for being hidden, and the fix was a permit that moved the pumping to winter and the crop to farmland. The footprint did not change much. The harm did.

Twenty two litres a day for a plant, and a stream pumped dry in the month the salmon needed it.

Sources

1. Bauer, S. et al. (2015). Impacts of surface water diversions for marijuana cultivation on aquatic habitat in four northwestern California watersheds. PLOS ONE 10. About 22.7 litres per plant per day; demand exceeding streamflow in three of four watersheds.
2. Zipper, S.C. et al. (2019). Cannabis and residential groundwater pumping impacts on streamflow and ecosystems in northern California. Environmental Research Communications 1.
3. Mills, E. (2012). The carbon footprint of indoor cannabis production. Energy Policy 46. About 1 percent of US electricity.
4. California State Water Resources Control Board, Cannabis Cultivation Policy (2017, updated 2019): forbearance periods, storage requirements, stream diversion rules.
5. Wilson, H. et al. (2019). First known survey of cannabis production practices in California. California Agriculture 73.
6. Photographs: opener: Cannabis Ganja Farm (18456071382) by Cannabis Pictures (CC BY) via Wikimedia Commons; inline: Commercial Cannabis Greenhouse Facility by Cannabis Tours (CC BY-SA) via Wikimedia Commons; inline: Yebel Tidighine, Ketama, Rif central, Marruecos by Jesús Ruiz Villena (CC BY) via Wikimedia Commons.

In 2023 the storm overflows of England's sewers discharged into rivers, lakes and the sea for 3.6 million hours. That is the sum, across about fourteen thousand overflow points, of the time each one was letting out a mixture of rainwater and sewage, and it works out at more than four hundred years of continuous spilling in one calendar year. Every hour of it was legal, or nearly, because the overflows were built to do exactly that, and had been doing it since the sewers were laid. What was new was that someone had counted.

This article is about what a combined sewer is and why it was built, what an overflow lets out and where, how monitoring turned a hidden practice into a public one, what London has built to stop it, and why the rest of the fix will take decades.

One pipe

The sewers of London, Paris, New York, Manchester and every city that built its drains in the nineteenth century were built to carry two things: the rain that fell on the streets and roofs, and the sewage from the houses, which until then had gone to cesspits and the gutter. The engineers had one pipe to lay and they put both in it, and the pipe was sized for ordinary rain and ordinary sewage, taken together, down to the river or, after the treatment works came, to the works. The problem is a storm. A heavy rain puts more water into the pipe in an hour than the works can take in a day, and a pipe that cannot pass it fills and backs up into cellars and streets. The engineers' answer was an overflow: a weir in the sewer, set at a height, over which the excess spills into a channel that leads to the nearest river or the sea.

The overflow is a designed relief valve rather than a fault, and every combined sewer has them, at intervals, wherever there was a watercourse to spill into. The newer suburbs and the newer cities, built after the 1950s, have separate pipes for rain and sewage, and no overflows, and their problem is different. The old cities have the one pipe and the weirs, and in a country where it rains often, they spill often.

What comes out

What comes over the weir is the pipe's contents at that moment: rain, a lot of it, and sewage, diluted by the rain but not treated. It carries the things the wastewater article on this site lists, bacteria and viruses from every toilet upstream, the organic load that the BOD article describes, ammonia, the phosphorus that feeds algae, the wipes that the flushable wipes myth is about, and whatever the streets contributed, which is oil, metals and litter. In a heavy storm the dilution is large and the spill short; in a light rain on a pipe already near full, or from an overflow whose weir is set low, the mixture is close to sewage and the spill can last for days.

Where it goes matters. An overflow into a large tidal estuary is diluted quickly; one into a chalk stream in a dry summer, or onto a bathing beach on a weekend, is not, and the public argument in England has been about those: rivers with the highest bacterial counts in Europe, swimmers ill after a race, a beach closed on the hottest day of the year.

The count

For a century the overflows spilled unseen. A weir in a sewer has no dial, and the water companies, which own the sewers in England, reported what they chose to. From 2016 the regulator required monitors on each overflow that record when it is spilling and for how long, and the coverage rose from a few percent to nearly all by 2023. The numbers, published annually, are what made the story: hundreds of thousands of spills a year, millions of hours, some overflows spilling on dry days, which the design does not allow and which suggested pipes so full or works so undersized that any flow tipped over the weir. A regulator's investigation in 2024 found that the companies and the regulators had, for years, permitted what the law did not, and the fines and the plans followed.

The count did not change the sewers. It changed what was known, and knowing turned a plumbing detail into a national argument about whether the private companies that own the pipes had spent on them what they had charged for.

The tunnel

The fix for a combined sewer is storage: a place to hold the storm flow until the works can treat it. It can be a tank under a park, a larger pipe, or, in a city the size of London, a tunnel. The Thames Tideway Tunnel runs for twenty five kilometres beneath the river from west to east at a depth of up to sixty metres, is seven metres across, and intercepts the thirty or so overflows that put about forty million tonnes of sewage into the tidal Thames each year. When it rains, the overflows fill the tunnel instead of the river, and when the storm passes the tunnel is pumped out to the works at Beckton and treated. It cost about four and a half billion pounds, took eight years to dig, and connected in stages through 2024 and 2025; in its first storms it captured, by the project's count, nearly all of what would have gone into the river. Paris built a similar system for the Seine before the 2024 Olympics, so that the swimming events could be held in it, and the river was swimmable on most days that summer and not on the others.

Why the rest is slow

London had one river and one tunnel. England has fourteen thousand overflows on hundreds of rivers, and the plan to bring them to a standard runs to 2050 at a cost the government estimates in tens of billions of pounds, to be paid through water bills. The engineering is not difficult; it is a tank at each overflow, or a separate rain sewer, or the sustainable drainage that keeps rain out of the pipe in the first place by letting it soak into the ground, which is the cheapest fix and the slowest, because it means rebuilding the streets. The United States has been at the same task since the 1990s under consent decrees, city by city, at a similar pace.

The overflows will spill, in a country where it rains, until the storage is built. The argument is about how fast, and about who pays, and about whether the companies that owned the pipes for thirty years should have been building the storage during them.

Keeping the rain out of the pipe

The other fix works at the top of the system. Every square metre of roof, road and car park that drains into a combined sewer adds its rain to the pipe, and a city that has paved itself over since the sewers were laid sends them several times the storm water they were built for. Sustainable drainage, which is the name for letting the rain soak into the ground where it falls, takes the load off: permeable paving, rain gardens along the kerb, green roofs, ponds in the parks that fill in a storm and drain over days. Philadelphia, under a consent decree for its overflows, chose this route over a tunnel and has spent two decades turning streets and school yards into places where rain sinks, at a cost per litre kept out of the sewer well below the tunnel's. Copenhagen did the same after a cloudburst in 2011 flooded the city. It is slow, because it is a rebuilding of the surface of the city block by block, and it is the only fix that also cools the streets and refills the aquifer, which is why the plans that run to 2050 lean on it.

What it teaches

Storm overflows are the part of the plain water story that shows what a treatment works cannot do: it can only treat what reaches it, and a sewer built to let the rest out will let it out. The overflows were designed, they were legal, and they were unseen, and the monitors turned a century of quiet spilling into a number that could not be argued with. London built a tunnel. The rest of the old cities have the same pipes and the same weirs, and the same rain.

3.6 million hours in a year, fourteen thousand weirs, and a river that was clean only once the counting started.

Sources

1. Environment Agency, Event Duration Monitoring data 2023: 464,056 spill events, 3,606,170 hours, from about 14,300 storm overflows, monitor coverage 100 percent.
2. Thames Tideway Tunnel, project facts: 25 kilometres, 7.2 metre diameter, about £4.5 billion, connected 2024 to 2025, designed to capture about 95 percent of the sewage that entered the tidal Thames.
3. US EPA (2004). Report to Congress on Impacts and Control of CSOs and SSOs: about 860 communities with combined sewers; about 850 billion gallons of overflow a year.
4. Defra, Storm Overflows Discharge Reduction Plan (2022, updated 2023): targets to 2050 and estimated cost.
5. Office for Environmental Protection, investigation into regulation of combined sewer overflows (2024).
6. Photographs: opener: Sewer outfall to the river Aire - geograph.org.uk - 7550732 by Stephen Craven (CC BY-SA) via Wikimedia Commons; inline: Bell Wharf CSO outfall in London by Z22 (CC BY-SA) via Wikimedia Commons; inline: Thames Tideway Tunnel by Matt Brown (CC BY) via Wikimedia Commons.

There is a jug in the fridge of millions of kitchens with a cartridge in its neck, and the water is poured through it before it is drunk on the belief that the cartridge makes the water safer. It does not, in any way the word usually means. What the cartridge does is take out the chlorine, which is the thing that was keeping the water safe, and give it a bed of warm wet carbon in which to grow whatever the chlorine was holding back. The jug makes the water taste better. That is a real thing and worth having. Safety is a different thing, and the jug, for most of the reasons water is unsafe, does nothing or makes it slightly worse. The distinction matters because the belief has a cost: people who trust the jug skip the test that would tell them whether their water has a problem, and people who need a real filter buy the one that was designed for a different job.

This article is about what is in a cartridge, what it removes and what it leaves, why the filtered water can carry more bacteria than the tap, what the jug is good for, and when a filter in the home is the right answer.

What is in the cartridge

A jug cartridge holds two things: granules of activated carbon, and beads of ion exchange resin, in roughly equal measure, held in a plastic shell with a mesh at each end. Water trickles through by gravity in a minute or two. The carbon adsorbs chlorine and the organic compounds that give water a taste or smell, and it does so well, which is why jug water tastes flat and clean. The resin swaps calcium and magnesium, the hardness, for hydrogen or sodium, which softens the water a little and stops the scale in the kettle, and it picks up some lead and copper in passing. That is the whole of it. A cartridge is a taste device, and the standard it is tested against, in the United States the one numbered 42, is a standard for taste and appearance, with a separate standard, 53, for things that affect health, to which most jugs are certified for one or two items, usually lead and the cysts of two parasites, and nothing more.

What the tap already did

The water from a tap in a country with a regulated supply has been through the whole of the plain water pillar on this site: coagulation, filtration, disinfection, and the testing that the disinfection article describes, against a list of parameters that runs to fifty or more, at a compliance rate above 99.9 percent in Britain and the United States. The chlorine in it, at half a milligram a litre or less, is there so that whatever gets into the pipe between the works and the tap is killed before it is drunk. The water is safe when it leaves the tap, by the standards a jug is not tested against, and the jug's job is to remove the thing that the works added to keep it that way.

The exceptions are real and specific. A house with lead pipes, which the lead article on this site describes, has a problem the jug can reduce while the cartridge is fresh and the flow is slow. A well with nitrate has a problem the jug cannot touch. A town whose water has failed, in a boil notice or an outbreak, has a problem the jug cannot touch either, and the notice will say so.

The bacteria in the jug

The finding that surprised the researchers who first looked at it, in Germany in the 1990s, was that the water coming out of a jug in an ordinary kitchen carried more bacteria than the water going in, often many times more and in some jugs thousands of times. The reason is the carbon. Activated carbon is a vast surface, hundreds of square metres in a gram, kept wet at room temperature, collecting the organic matter that bacteria feed on, and with the chlorine that would have killed them removed at the first pass. Within days a biofilm forms, and after the four weeks that the manufacturers give the cartridge, and the eight or twelve that most people give it, the cartridge is releasing bacteria into every jug of water poured through it.

The bacteria are mostly harmless, the ones that live in any wet place, and a healthy adult drinks them without effect; the studies did not find outbreaks of illness, only counts. The point is narrower and it is the myth's undoing: a device bought to make water safer adds to it the thing the water was treated to remove. The manufacturers' instructions say to keep the jug in the fridge and change the cartridge on time, and both help; a jug on a sunny worktop with a cartridge from the spring is a small warm aquarium.

What it is good for

The jug is good for taste, and taste matters, because the largest reason people give for buying bottled water is that their tap water tastes of chlorine, and the bottled water article on this site shows what that costs. A jug that stops a household buying bottles is a jug that has paid for itself in water and in plastic several times over, and it should be in the fridge and its cartridge changed on time. It is also good for the kettle, in a hard water area, and for a household that likes its water soft. It is a kitchen appliance, and judged as one it is a decent thing.

When a filter is the right answer

There are households that need a filter, and for them the jug is the wrong one. A house with lead pipes needs a filter certified for lead, fitted at the tap, and a plan to replace the pipe. A well with nitrate, arsenic or bacteria needs the specific treatment the plain water articles describe: reverse osmosis or ion exchange for nitrate and arsenic, ultraviolet or chlorine for bacteria, tested and maintained. A supply with PFAS needs granular carbon or reverse osmosis at a scale a jug cannot hold, and a supply with hardness alone needs a softener or nothing. In every one of those cases the first step is a test, which tells the household what is in the water, and the second is a filter matched to it. The jug is bought without the test, for a problem it was never designed to solve.

The plastic and the money

A cartridge changed every four weeks is thirteen a year, each a plastic shell with a hundred grams of carbon and resin inside, and the jug itself is replaced every few years when it clouds. The makers run return schemes for the shells and most cartridges go in the bin. Against the bottled water the jug replaces, which is the case for it, the plastic is small: a year of cartridges weighs less than a fortnight of bottles for a household that drank two litres a day from them. Against the tap alone, the jug is a year of plastic and a hundred pounds or so for water that was already fit to drink and tasted of the chlorine that kept it so. A glass jug of tap water left in the fridge for an hour loses most of its chlorine to the air and costs nothing, which is the answer for the household that dislikes the taste and does not need the softening. A jug kept for the taste, filled from the tap, changed on the day the indicator says and stored cold, is a reasonable thing to own. The mistake is only in what it is believed to do.

What it teaches

The filter jug is the appliance on this site that shows what a myth costs when it is cheap. Millions of people pour their water through a cartridge in the belief that it is protecting them, when what it is doing is removing the protection they had and growing a garden in its place. The tap water was safe. The jug makes it taste better, and that is the whole and honest case for it. Safety is a test, a specific filter for a specific finding, or a supply that meets the standard, and for most of the people who own a jug, the supply already does.

Four weeks of cartridge life, the chlorine gone on the first pass, and a jug that adds what the tap took out.

Sources

1. NSF/ANSI Standard 42 (aesthetic effects: chlorine, taste, odour) and NSF/ANSI Standard 53 (health effects: lead, cysts, specific organics); which claims jug filters are certified for.
2. Daschner, F.D. et al. (1996). Microbiological contamination of drinking water in commercial household water filter systems. European Journal of Clinical Microbiology and Infectious Diseases 15. Bacterial counts in filtered water up to 10,000 times higher than tap.
3. Wu, J. et al. (2021). Are Point of Use water filters effective? A review of removal of lead, arsenic, nitrate and microbes. Environmental Science: Water Research and Technology 7.
4. Drinking Water Inspectorate (England and Wales) and US EPA, water quality compliance rates above 99.9 percent for regulated parameters.
5. Purdue University and University of Arizona studies on carbon filter biofilm growth and the effect of cartridge age.
6. Photographs: opener: Brita water filter in use by BrokenSphere (CC BY-SA) via Wikimedia Commons; inline: Person fills glass with water at kitchen sink closeup by Shixart1985 (CC BY) via Wikimedia Commons; inline: Pasteur filters (CP 2615), National Museum of Health and Medicine (4712460336) by National Museum of Health and Medicine (CC BY) via Wikimedia Commons.