Tropical Storm Harvey had not stopped raining on Texas before the first estimates emerged as to how many billions of dollars in damages would result from the storm. Initial estimates from insurance companies like Hannover Re put the number at $3 billion. In a note to clients, JP Morgan estimated that the insurance industry could lose $10 to 20 billion from Harvey, making it one of the top 10 costliest hurricanes to hit the U.S. Enki Holdings, a consultancy that calculates the risks and costs of various natural disasters, said Monday afternoon that its estimates for Harvey damages had reached $30 billion. It’s likely, though, that none of these estimates will end up being accurate. “It’s a pretty tough business—you don't really know what’s on the ground,” Tobias Geiger, a researcher from the Potsdam Institute for Climate Impact Research, told me, about forecasting the impact of disasters. “A good ballpark would be if you’re off by a factor of two.”

There are two different types of damages tallied from natural disasters: direct damages, which are caused by harm to physical structures like buildings and the belongings inside of them, and indirect damages, which are caused by people losing their incomes and jobs. Both direct and indirect damages are best tallied months or years after a storm has taken place because people have made insurance claims, and if they’re not insured, they know how much money they spent rebuilding. But dozens of companies try to predict damages earlier than that, both because it is useful for insurers to know what their potential costs may be, and because government officials may need to offer economic aid for residents.

Every company that predicts damages from disasters has a different method for doing so. Enki, for example, has a computer simulation that uses the laws of physics to estimate what the forces of nature such as wind, waves, and flood waters would do to the properties it is predicted to hit. The company comes up with the values of the properties in a storm’s path using data it has tabulated on what sits on parcels of land across the country. Risk Management Solutions (RMS), a catastrophic-risk-modeling company based in California, has constructed a hurricane model that simulates tens of thousands of potential hurricanes to advise the insurance industry on the likely impacts of Harvey. RMS can use wind speeds to calculate what percentage of the home may be damaged by the storm, said Tom Sabbatelli, a senior product manager with the company. Once it gets information from insurers on the how much particular properties are worth, RMS can estimate the cost of the damages based on those percentages.

These models are pretty accurate when accounting for the damage caused by wind speed and storm surges, which are typically the two factors that cause the most damage from a hurricane. But Harvey was not a typical hurricane. Rather than just hitting the coast and moving on, it crept inland and lingered, causing huge amounts of flooding in Houston and surrounding areas. Watson estimates that 30,000 square miles have seen over 10 inches of rain. Experimental flood models he’s run show that impacts from Harvey are four to five times higher than those of a typical hurricane. “If Harvey had been a normal Category 3 or 4 hurricane, we’d be talking about a $4 billion storm—from a national perspective, it would not have been a huge event,” he said. “The problem is that Harvey moved inland and turned into a big, wet, tropical storm.”

Economists are now trying to calculate how to factor in the flooding damage that Harvey is causing. But damage from flooding is much more difficult to model than damage from wind. While the general wind structure of a hurricane is well understood by scientists, flooding is more circumstantial. Small factors can easily aggravate flooding—a trash can could block a drainage ditch, for example, and flood a block of homes, while the drainage ditch in the next block worked as planned, Watson said. Wind damage also usually harms the outside of a structure, while flood damage is heavily dependent on what’s inside the structure too, which can vary from house to house, Sabatelli said. And while instruments on the ground can accurately measure how fast the wind is blowing during a storm, there are not enough instruments to determine a flood's extent with great granularity, block to block or home to home. “There will always be uncertainty in flood estimates,” he said.

Still, there are two past storms that can be compared to Harvey when it comes to calculating the amount of damages—Hurricane Ike and Tropical Storm Allison. Ike in 2008 had lower wind speeds than Harvey, but had a high storm surge and a great deal of coastal flooding. It cost $30 billion, which is about $38.4 billion in today’s dollars, according to the National Oceanic and Atmospheric Administration (NOAA). Tropical Storm Allison, which dumped 30 to 40 inches of rainfall in Texas and Louisiana in 2001 and produced severe flooding in Houston, cost about $8.5 billion, which is about $11.9 billion in today’s dollars, according to NOAA. The amount of rain Allison dumped on Houston was catastrophic, but Harvey released as much rain in two and a half days as Allison did in five, Watson said. “It’s probably too early to tell, but my guess is there is going to be parts of Houston that will be uninhabitable for at least months,” said Tatyana Deryugina, a professor of finance at the University of Illinois who studies the economic impact of disasters.

The fact that Houston is the nation’s fourth-largest city will also push the damages up, simply because there were more people and properties in the storm’s path. It is a bigger city today than it was when Allison hit. The city of Houston had 2.3 million residents last year, an 18 percent jump from the population in the year 2000, according to Census data. As more people move to coastal regions, scientists predict that the costs of storms could continue to rise because more people and properties will lie in the path of big weather events.

In addition, economic forecasters have also discovered that each new storm can generate costs that had not been problematic in the past. Hurricane Sandy, for instance, caused tens of millions of dollars in additional damage because of the valuable equipment like servers and financial documents stored in basements of some of the buildings that were affected. The costs of things like road signs and guardrails often add up more quickly than most forecasters anticipate—Watson estimates the costs of replacing street signs and repairing road damage after Katrina reached $800 million.

Still, one bright spot might be that indirect damages from Harvey might not be particularly bad. In a 2014 study, Deryugina used tax-return data to track the long-term economic impact of Hurricane Katrina on its victims. She found that though Katrina uprooted many people from their homes, the storm did not harm their long-term economic prospects. Though incomes of the average Katrina victim were lower than that of counterparts in other cities the year after the storm, by 2008, the average hurricane victims had incomes that were higher than comparable people in other cities. “What we found is that people bounce back pretty quickly,” she told me. On the other hand, New Orleans was a city with declining economic opportunities when Katrina hit, which meant that people forced to relocate often found better opportunities when they moved, she said. Houston was booming economically before Harvey, and if people lose jobs in Houston because of the storm, they might have trouble finding better opportunities elsewhere.