Editor’s Note: On July 31, the Senate approved a bill temporarily reauthorizing the National Flood Insurance Program until November 30, 2018.

The National Flood Insurance Program, which covers some 5.2 million property holders in the U.S., was slated to get a badly needed overhaul today. The Senate’s task—which includes hammering out reforms that address the changing math of flood risk—has already been pushed back three times since November. Yet lawmakers still have not compromised on how to fix a broken system, so a reauthorization of the NFIP will almost certainly be punted again, to July 31.

The NFIP, which is run by the Federal Emergency Management Agency, is struggling because it is trapped in a downward spiral of ballooning claims without the resources to cover them. The program has been unable to sustain itself since Hurricane Katrina in 2005, but last fall, after a hurricane season that was unprecedented in both severity and frequency, administrators announced the program had maxed out the $30.4 billion it had been authorized to borrow from the U.S. Treasury. Although Pres. Donald Trump signed a disaster relief bill that forgave $16 billion of its debt, the NFIP continues to be steamrolled by extreme weather events, including the recent series of back-to-back nor’easters that have clobbered the Northeast.

In its current form, the NFIP is “ill-suited to deal with the challenges we face today and the flood risks we face five, 10, 50 years from now,” says Robert Moore, a senior policy analyst at the National Resources Defense Council. Pretty much everyone agrees this is true, regardless of party politics. In fact, SmarterSafer, a collation made up of insurance companies, progressive environmental organizations and right-leaning think tanks, is one group advocating for sweeping change (pdf).

Among the major reasons why the NFIP cannot keep up with the growing number of claims is that it assesses risk based on outdated science. In the wake of Hurricane Sandy it became widely known FEMA was demarcating flood zones using data from the 1980s, which failed to predict the inundations brought by the 2012 storm. Meanwhile development continues to bloom in increasingly vulnerable floodplains, creating more properties that the NFIP pays to repair—sometimes over and over again.

Since the chaos of Sandy, a growing number of researchers have analyzed flood risk on a more comprehensive and granular level. Their results have exposed an even more troubling gap between FEMA guidelines and recent estimations of risk.

One study, published in Environmental Research Letters last month (pdf), found the number of Americans exposed to serious flooding is about three times higher than previous estimates. That is partly because inland flooding is not as well considered as coastal flooding. According to the paper, maps delineating riverine and rainfall-driven flooding “are only partially complete nationwide, and no comprehensive estimate of U.S. population exposure currently exists. Where they are available, FEMA flood maps are of varying age and levels of quality.” They also have “notably poor coverage” of streams and creeks.

According to FEMA’s calculations, only 13 million people live within a 100-year-flood zone (that is, a place that has a one-in-100 chance in a given year of being hit with the worst flood in a century). But by doing a comprehensive evaluation of floods driven by rivers and rainfall and creating higher-resolution maps, the authors reported nearly 41 million people in the U.S. live within a 100-year-flood zone. In another example of glaring discrepancies, a 2016 study by Robert Criss of Washington University in Saint Louis found current estimates of 100-year-flood elevations could be off by as much as 6.6 feetin parts of the Upper Mississippi River. Numbers like these imply that many more Americans should be buying flood insurance, which would bring in more premium dollars to the program.

Sea-level rise, coupled with other factors such as shifts in wave action and storm surges, is changing those numbers, too. By 2050, according to a recent article in Nature, “some places can expect to see what is currently considered a 100-year-flood event recur as often as every one or five years on average.” FEMA mostly uses models based on historical storm data, which, of course, do not account for these predicted effects of climate change. “One of the biggest problems is that no flood map produced by FEMA contemplates a future that looks any different from the past,” Moore says. FEMA, for its part, is self-aware about its woeful maps; the NFIP’s chief, Roy Wright, recently told the Miami Heraldthat all Floridians should have flood insurance and “quit focusing just on the lines.”

For a while some legislators suggested they were wary of changing flood-insurance standards because there was not enough data to draw on. Now, “the science is already much, much further ahead than what FEMA is working with,” says Thomas Wahl, a coastal engineer and oceanographer at the University of Central Florida. Wahl published a paper in Nature Communications in 2015 on the effects of compound flooding, which is when increased river discharge, precipitation and storm surges happen at the same time. This helped drown communities near Houston during Hurricane Harvey.

From Wahl’s perspective, “the biggest gap is the fact that FEMA’s maps do not connect inland flooding and coastal flooding,” he says. “FEMA creates a flat map from the river side of things and a flat map from the storm surge side of things, and they just overlay them, which assumes that these two things are completely independent. But most tropical storms bring a lot of rain and storm surges. We understand why these events happen simultaneously, but what we haven’t done is include that information into risk assessments,” Wahl notes. Some of that backlog is bureaucratic: FEMA only uses officially approved models, he says, and the processes of approval can slow down the inclusion of newer, better data. Without a better understanding of the true risks, it is trickier for lawmakers to agree on solutions for how to make the NFIP financially stable.

As a growing body of research illuminates the current and future scope of flooding, adaptation strategies are beginning to look more attractive than the traditional (and bankrupting) cycle of rebuilding. In addition to making building codes stronger, installing hard flood barriers and funding projects such as raising homes on pilings, the once-disparaged concept of “buyouts” is slowly being embraced. Instead of repairing certain homes that flood over and over again, the government “buys out” your property at its pre-storm value using NFIP money, then permanently returns the land to nature. This more radical option gets people out of harm’s way and can be cheaper than repeatedly paying to repair the same property. The NRDC, for one, is advocating for a better pathway to buyouts as part of NFIP reform.

Uprooting from home is an emotionally and financially complex decision, but people who are repeatedly flooded seem to be growing more open-minded to relocation strategies. Take Harris County, Texas, for example. Prior to Hurricane Harvey, “the county had purchased 3,000 flood-prone homes over 20 years” in buyout programs, Moore says. “After Harvey hit they had 3,800 new application requests for home buyouts in just eight weeks.” The governor of Texas requested $1.5 billion from the federal government for buyouts alone.

But FEMA’s buyout process can take years to complete. In the aftermath of a disaster, homeowners often face an onerous choice: Hold off on repairs and live in limbo while awaiting a verdict on whether they qualify for a buyout or start investing in reconstruction so they have a place to live in the interim. The slowness and opacity of the buyout process disincentivizes many eligible homeowners from participating.

Streamlining the buyout option could make those decisions easier while improving the long-term financial viability of the NFIP. According to data the NRDC obtained via the Freedom of Information Act, FEMA has identified about 30,000 “severe repetitive loss properties” that get routinely flooded and repaired. Of those properties, more than 75 percent are worth less than $250,000, with the average price being $110,000. “These severe repetitive-loss properties have already incurred an average of $134,000 in damages” that the NFIP has paid out, Moore says. So in many cases, “we’ve already bought their properties.”

Incorporating recent research into the NFIP could help establish a framework for identifying which properties and communities are most vulnerable to repetitive flooding and create more efficient standards to determine buyout eligibility. Armed with more knowledge about how often their houses might flood over the life of a mortgage, homeowners could make informed decisions. Communities, too, might preemptively plan how to relocate families, mitigating psychological and financial stress while also maintaining their tax bases. That way, when catastrophic damage occurs the response is not purely reactive.

While the particulars of insurance reform are still being determined, one thing that is for certain is that flood risk is going up. “I believe we have the science now that we can inform policy makers much better than we could several years ago,” Wahl says. “It’s very site-specific and there is still uncertainty, but we are seeing disasters strike all the time, so I think we are at the point where we can make some decisions.”

If, in July, reform gets kicked down the road yet again, it might only take one expensive hurricane to force Congress to decide whether they must increase the NFIP’s borrowing limit beyond $30.4 billion or forgive even more debt, Moore says. Alternatively, Congress could try to force the Senate’s hand to pass a bill on July 31 by allowing the NFIP to lapse. In that case, the NFIP’s ability to borrow money would plummet to a cap of $1 billion and homeowners would not be able to renew their policies—a precarious scenario at the start of hurricane season.