Local forecasters knew Hurricane Harvey would be a bad storm. Their models suggested it would drop 10 to 15 inches of water across the city the first day after landfall, says Jeff Lindner, a meteorologist with the Harris County Flood Control District. “But if we go back,” he says, those models “were a bit low on the rainfall totals that Saturday night–Sunday morning period.” Northeast of Harvey’s eye on August 25, the hurricane encountered something scientists did not realize might contribute to the torrents that broke centuries-old rainfall and flood records: the City of Houston.

Houston’s skyscrapers and concrete expanses likely exacerbated floods by altering the storm itself and making it rain harder, according to new research published in Nature Wednesday. “It’s very intriguing,” says Kerry Emanuel, an atmospheric scientist at Massachusetts Institute of Technology, who did not work on the study. “We know cyclones are sensitive to characteristics of the surface—mountains, streams, marshland. This new twist is that cities have become big enough to tangibly alter the storm.”

To see how urban areas could do that, the researchers built a computer model of Harvey and the Houston area. “Then we removed the urban areas from Houston and replaced them with cropland,” says Gabriele Villarini, an environmental engineer at The University of Iowa and an author on the study. This allowed the scientists to see how the storm would likely have behaved if Houston did not exist. According to the model, urbanization in Houston may have caused Harvey to release far more rain than it would have if plains had existed in Harris County instead of neighborhoods. “The presence of urban areas enhanced all the things you need to get heavy precipitation,” Villarini says.

Like most hurricanes, Harvey whipped across stretches of trees, hills and grasslands when it made landfall. The ruggedness of natural landscapes typically slows these storms down and—like water rushing over pebbles—forces the air above to bubble higher into the sky. There, in the upper reaches of the atmosphere, the storm cools and thus drops more rain. Tall buildings and sprawling suburbs like Houston’s artificially jack that ruggedness to new levels, Emanuel says, making the effect that land has on cyclones even more extreme.

According to Villarini and his colleagues’ model, storms physically change in a couple ways under such circumstances, Emanuel says. First, the artificial ruggedness of an urban area slows air down. Whenever air slows in a hurricane, he says, it gets shunted toward the center of the storm and up into the sky. “That increases rainfall everywhere [in a metropolitan area],” he adds.

A storm moves particularly slowly over downtown areas where buildings are tallest, but the winds bearing down from outside the city are still moving quickly. “So, [the storm] is piling up on the city,” Emanuel says. “And it has no place to go but up. Then you have the reverse situation downwind of the city, where air is falling back down.” That causes even more rain on the side of a city where the hurricane first plows into—in Harvey’s case, the southeast side—and less rain on the far side of the city.

Overall these effects lead to more rainfall from tropical storms, Emanuel says. “It’s very noticeable. It may be on the order of 50 percent more rain. This is a powerful enough effect that you can see it through all the noise in [weather] systems,” he says, adding that the exact amount is highly variable because storm systems are so unpredictable and chaotic.

Urbanization has the added consequence of increased paved surfaces like concrete or asphalt sidewalks and roads, Villarini notes. “Because no rainfall goes into the ground because of this, more water runs off on the surface,” he says. That makes flooding worse. When he included this into his analysis, he found Harris County’s risk of a Harvey-level flood had increased up to 21-fold due to Houston’s presence.

If such findings are true, they would have significant implications for urban planners, Lindner says. “If we could see storm system after storm system produce the same rainfall patterns, where one side of the city sees more rainfall, that would be great,” he says. “We could target those areas for larger projects or more stringent building codes.”

But until he sees the model applied to other cities or other storms with consistent results, Lindner says it does not help him forecast what future hurricanes will be like or help planners with flood interventions. “I can’t use this study, because this is one storm—one event,” he says. And in Lindner’s experience hurricanes do not have much of a pattern when it comes to flooding. “Just by working in the [20]15 and ’16 storms, I can tell you the heaviest rain did not fall in the eastern part of Houston,” he says.

One thing does seem clear, however. Rainfall from hurricanes is increasing and will continue to do so, says Christina Patricola, a climate scientist at Lawrence Berkeley National Laboratory, who was not involved in the work. “The development of cities and climate change are both human influences that are exacerbating the rainfall from cyclones,” she says. In her own work, also published Wednesday in Nature, she shows climate change may lead to a 15 to 35 percent increase in rainfall and likely made hurricanes Irma, Maria and Katrina deliver 5 to 10 percent more rainfall than they would have in the absence of climate change.

Coastal cities will need to think very hard about how to prepare for both the increase in rainfall from climate change and continued urbanization, says Sam Brody, an urban planning professor at Texas A&M University at Galveston who did not take part in the research. “After Harvey, the City of Houston—it still astonishes me to say this—passed the strongest building regulation for flooding,” he says. New Houston buildings must now be two feet above the 500-year flood line. “But [we need] to consider future conditions systematically, and not just in reaction to storms.”