Rainwater stored on rooftops could help cities beat extreme heat

Rain that already fell on those roofs is the exception, and a new model of one Tokyo neighborhood estimates that catching it can cut the energy air conditioning uses.

How much it saves depends on the timing more than the hardware. A bigger tank matters less than the temperature that starts the spray.

Zhonghua Zheng, an associate professor in the Department of Earth and Environmental Sciences at the University of Manchester (UoM), built the system with PhD researcher Junjie Yu and four co-authors.

They wrote the tank and the sprinklers into the Community Land Model Urban, a simulation of how a city’s roofs, walls, and streets trade heat and water with the air. The NSF National Center for Atmospheric Research built that model; the Manchester team added the plumbing.

In an interview with Earth.com, Zheng was asked which finding he most wanted to highlight.

“One important finding is that the temperature at which sprinkling begins matters more than either tank size or sprinkling intensity,” he said.

________________________________________________________________________

The roof’s own heat starts the spray

Watering a roof works. A 2025 study found that spraying can lower a roof’s surface temperature by 10°F to 35°F (5.5°C to 19.5°C). The constraint has always been the water.

Rain landing on a roof normally evaporates, puddles or runs off down the drain. In this design, the runoff goes into a tank instead and waits there until the roof gets hot enough to need it.

The sprinklers switch on when the roof surface passes a set temperature, and the team let that threshold range from 95°F to 149°F (35°C to 65°C). Roofs get much hotter than the street air, so the number refers to the roof itself.

Yu and Zheng sampled 400 combinations of tank size, spray rate, and trigger temperature, then ran each through the model. A higher trigger generally meant more air conditioning energy. With the trigger set low, the spray rate began to matter too.

Roof runoff is too dirty for drinking or washing, so the tank feeds only the sprinklers, and only from June through September. In the best runs, the water came on around 9 a.m. and stopped around 7 p.m.

Bigger tanks stopped paying off

The neighborhood in the model is a 0.23-square-mile patch of Yoyogi, Tokyo, roughly 110 football fields, with 1,996 roofs and buildings averaging 30 feet (9 m) tall.

Tank size here is expressed as the depth of water that would cover the roof when the tank is full. The smallest tank holds about 0.17 inch (4.37 mm) and the biggest holds 0.79 inch (19.99 mm).

The team’s search had two goals, less air conditioning energy and the smallest tank that could deliver it. It ran on a machine-learning copy of the climate model, because the full version is too slow.

Among the best combinations, the neighborhood with the smallest tank used 1,600.76 megawatt-hours of air conditioning a year. Moving to the 0.17-inch tank saved about 49 megawatt-hours. Saving the next 49 took storage 4.6 times larger.

So the first half of the saving comes at little cost, and the rest costs almost five times the tank. Water behaves the same way.

“Once the roof was sufficiently wet, adding more water produced diminishing cooling and energy-saving benefits,” Zheng told Earth.com.

A roof cools as the water on it evaporates. Once the surface is soaked, extra spray just lies there, and in the model a long, gentle soak beat a short, heavy one.

Extreme heat hours fell across the block

The street cools for two reasons. Wet roofs evaporate water, and buildings running less air conditioning push less waste heat into the urban heat island around them.

Over ten simulated summers, the sprinklers cut the time roofs spent at unusually high temperatures by roughly 590 to 1,220 hours. Those were temperatures the roofs reached only 5% of the time without the system.

Street-level air cooled less, but it still cooled: hours above 95°F (35°C) fell by 86 to 298 across the decade, or about 9 to 30 hours a summer.

Comfort improved least. Water sprayed on a roof adds moisture to the air, and humidity is part of what makes heat dangerous.

The humidex combines temperature and humidity into a single reading of heat stress. The block spent only 3.5 to 19.5 fewer hours above 46, the level the authors treat as great discomfort. With the largest tanks, those hours started to climb again.

________________________________________________________________________

Tanks also caught water that would have poured off the roofs. Extreme runoff eased steadily as storage grew, by five hours at the small end and 51 at the large one.

Heatwaves arrived less often

A heatwave here is three days in a row at or above 95°F (35°C). The team counted them twice, once breaking the run whenever a cooler day interrupted it, and once allowing one cooler day inside.

In the simulations, the block had fewer heatwave days under both counts. The biggest tanks cut the most, and the events arrived less often and peaked slightly cooler.

Average length didn’t follow. Under the stricter count it edged up, because the sprinklers were ending short heatwaves and leaving the long ones to make up more of what remained. Under the looser count it fell.

Warmer years produced bigger savings

The ten summers weren’t equally hot, and the hotter ones saved more energy. That pattern was weak in runs with small tanks and moderately strong with the biggest ones. Yearly rainfall tracked the savings less closely than temperature did.

Zheng said the optimized system lowered air conditioning demand, extreme roof and air temperatures, heatwave days, and extreme runoff. The bigger savings in hotter years, he said, suggest the benefits could become more valuable as the climate warms.

The World Health Organization estimates that about 125 million more people were exposed to extreme heatwaves in 2016 than in 2000.

Asked by Earth.com whether it would work in a drier city than Tokyo, Zheng said it might, with changes to suit local conditions.

“Limited rainfall could make it less effective or require another water source,” he said.

A pilot has to come first

Nothing here was measured on a real roof. The city model behind it has been checked against a tower of instruments in Yoyogi, and it tracks the measured radiation closely and the heat and moisture flows less well. But the tank and the sprinklers don’t exist outside the code.

One quirk of the model rides with the findings. Air just above the roofs mixes freely with air in the streets below, so where that link is weaker, the street-level cooling would come out smaller.

Mold grows on a roof that stays wet, sprinkling wears at roofing material, rules on collecting rainwater differ from city to city, and tanks and their upkeep cost money. Reflective cool roofs – another way to cool the same buildings – use no water at all.

Zheng said costs, regulations and moisture risks all need assessing. Then comes the part no simulation can do.

“A city would need pilot testing, suitable roofs and drainage, sufficient local rainfall, and systems for monitoring and maintenance,” Zheng said.

Only a real tank on a real roof, through one Tokyo August, can turn these estimates into measurements.