Exclusive: FEMA maps may be missing over 200K high-flood-risk Harris County properties, study shows 5%
By Yilun Cheng0% Katherine Oung0%
7/30/2026, 4:00:00 AM
BS Summary: This article contains 14 faulty reasoning types, including Framing Effect, Availability Heuristic, and Halo Effect, with Negativity Bias as the most egregious example at 12% saturation with 263 hits. Analysis detected 908 faulty-reasoning hits from 2,187 analyzed words, generating a BS Score of 19.7% and a BS Rank of 5% (21,211 of 22,308 articles). This article is better (less manipulative) than 95.10% of the article peer group.
In Harris County, the number of properties facing 100-year-level flood risk may be more than twice as large as FEMA’s current maps suggest, according to a Texas A&M University study.
For the past decade, flood researchers at the university’s Institute for a Disaster Resilient Texas have worked to develop a tool that offers disaster risk information to property owners across dozens of counties impacted by Hurricane Harvey in 2017.
Using detailed federal insurance claims data not available to the public, they built the model based on where flooding has historically caused damage.
Despite increasingly advanced climate science, many homebuyers are still left without a clear picture of the risk they may be taking on, said Sam Brody, the institute’s director and the lead technical expert on Gov.
Greg Abbott’s post-Harvey commission.
FEMA’s maps for Harris County are decades out of date, and their long-awaited replacements will likely take years to finalize.
Meanwhile, state disclosure laws still leave significant gaps, and some major real estate platforms have resisted publishing independent climate risk estimates.
“Every time I gave a talk in Texas, people would come up to me afterward and say, ‘I wish I had known these things, and I would not have bought my house, but now I’m stuck.’
Or I’d hear, ‘I’ve lived in my house for 30 years.
I never flooded, and now I’m flooding.
Can you tell me why?’’
Brody said.
“I realized we had enough data and knowledge to tell people their risk.”
Overall, FEMA places about 184,800 county properties in its regulatory floodplain — areas with at least a 1% annual chance of flooding.
The IDRT model, on the other hand, estimates that about 389,500 properties have at least a 1% annual chance of sustaining flood damage.
That figure is closer to the 345,400 properties that would fall inside FEMA’s regulatory floodplain under a new set of maps released by the federal agency in February.
The proposed maps would mark the first comprehensive update to Harris County’s floodplain boundaries in two decades, but local officials said they could still take years to finalize.
The Houston Chronicle has integrated the IDRT tool’s Harris County data into a searchable map that allows residents to enter an address and see the property’s estimated flood risk.
Readers can also visit the researchers’ Buyers Aware platform to explore flood hazards in other counties.
Like any model, the Texas A&M researchers’ approach has strengths and limitations, according to multiple independent climate risk experts.
A key advantage, they said, is that it draws on actual flood damage claims rather than relying on engineering assumptions about how water moves, as FEMA and many commercial flood-risk apps do.
“The concept of a floodplain makes sense in places that have a river, then a floodplain, then hills and mountains beyond that.
It’s always going to flood in roughly the same place,” said James Dolls-Gollins, a civil and environmental engineering professor at Rice University.
“But Houston is very flat, with a lot of concrete, and gets torrential rain, so it can flood anywhere.”
With the IDRT model, “you get to bypass a lot of the assumptions because you’re actually looking at real events,” Dolls-Gollins said.
How the model works
The idea for Buyers Aware started, in part, because Brody was tired of being asked by relatives and friends if their houses would flood.
“I can automate this,” he remembered thinking.
So one weekend in 2016, a former student flew into Houston, and the two got to work.
Together they figured out how to turn flood data Brody had gathered from previous projects into something the public could use.
SEARCHABLE MAP: Is your Houston home outside of the floodplain?
New research says think again.
By the end of the weekend, they had built a rudimentary version of what would become Buyers Aware.
It covered only Harris County and looked little like today’s tool.
Still, people found it.
Following one storm that year, the site drew roughly 30,000 views in a week, Brody said.
Then Harvey hit, and the project received millions in state funding.
That allowed the team to hire staff, expand to more counties and obtain more claims, elevation data and soil data to build the machine-learning algorithm that is now the basis of its flood risk assessments.
The algorithm learns how seven property-level factors are associated with historical flood insurance claims, and then uses those patterns to estimate flood hazards for other properties.
Below are the seven factors it analyzes.
Seven risk factors that can determine your flood risk
Elevation
Lower-lying areas are more likely to collect water or sit in the path of runoff.
Distance to coast
Properties closer to the coast face greater exposure to storm surge and high tides and generally have higher flood risk.
Distance to stream
Properties closer to a stream are more vulnerable to flooding because they are more likely to be affected when water overflows its banks.
Height above nearest drainage
Those with less vertical separation from nearby drainage points – such as a stream, lake or coastline – are more vulnerable to rising water.
Land roughness
Vegetation and other ground cover can create a rougher surface and slow moving water.
This typically reduces flood risk.
However, in low-lying areas, for example, rougher surfaces may cause water to pool and increase flood risk.
Soil drainage
How quickly water can move through the soil when saturated.
Water that drains more slowly is more likely to remain at or near the surface, increasing the likelihood of flooding.
Imperviousness
The share of surrounding land covered by urban impervious surfaces such as pavement, buildings and concrete that prevent water from soaking into the ground.
Impervious cover creates more runoff and can overwhelm drainage systems, increasing a property’s risk.
Unlike FEMA’s floodplain boundaries – which divide properties into those inside or outside a certain designated flood zone – the IDRT model shows risk on a continuum.
When the two models display different levels of risk, Brody said residents should still heed FEMA’s designation because it determines whether a homeowner must carry flood insurance and other requirements.
Understanding Flood Risk Levels
Floodway
Rivers, bayous and adjacent land where water flows most forcefully during storms.
Development is usually heavily restricted in these areas.
100-Year Floodplain
Areas with a 1% chance of flooding in any given year.
In Harris County, these properties could flood with about 17 inches of rain in 24 hours.
Significant risk over a typical 30-year mortgage period.
500-Year Floodplain
Areas with a 0.2% annual chance of flooding.
Houston experienced three "500-year floods" from 2015 to 2017, including Hurricane Harvey.
Note: With climate change intensifying storms, these historical risk calculations may underestimate current flood dangers.
Learn more here.
A higher degree of risk on Buyers Aware can be a reason to look more closely at a property and consider added protections, from buying flood insurance to improving drainage, adding flood vents, sealing entry points or, in more extreme cases, elevating a home.
A lower risk level, meanwhile, could offer some reassurance about a property’s level of safety, he said.
A FEMA spokesperson acknowledged the agency’s maps do not capture every place that could flood.
Third-party tools use different data and methods and can provide complementary information.
“It’s very important to recognize that flooding happens everywhere and can occur outside high-risk areas,” the spokesperson said.
“While maps help communities manage risk, they do not guarantee that flooding will not happen outside of these boundaries.”
A growing field
For homebuyers, piecing together a property’s flood history can often be daunting and sometimes impossible, experts said.
Texas law requires sellers to disclose information, such as if a home is in a floodplain, has flooded before or has been the subject of a flood insurance claim.
But the law has limits.
An owner who has held a property only briefly, for instance, does not have to disclose flooding that occurred before they owned it or pass along details provided by a previous seller.
There is also no central database preserving those disclosures as homes change hands.
At the same time, FEMA maps, one of the main tools available to the public, were primarily designed to help determine insurance requirements and guide development regulations, not to communicate individual properties’ flood risk to the public, several researchers said.
Brody’s team is one of a growing number of organizations trying to fill that gap.
MIND THE GAP: Harris County finally has new drafts of FEMA flood maps.
Here's what they're still missing.
Many academic groups, nonprofits and private companies have developed their own tools.
And each approach comes with trade-offs, according to Brett Sanders, a civil and environmental engineering professor at the University of California, Irvine.
In a 2024 paper he co-authored, Sanders and his colleagues examined the flood hazard data used by First Street, one of the most well-known climate risk companies in the U.S.
The researchers found “concerning levels of uncertainty” in First Street’s data at the municipal and property levels, warning that these national-scale models may miss important local conditions such as small differences in the landscape, drainage systems and flood control infrastructure.
That is part of the appeal of the Texas A&M researchers’ approach, Sanders said.
The algorithm sidesteps the kind of uncertainties engineering models must account for — from predicting how much rain will fall and where the water will go to how much damage it may cause and minute local factors that can be tricky to capture.
“It’s a really promising approach,” Sanders said.
“Machine learning methods really work well when you have good data and when you can have a lot of data.
So it may be that Houston’s a good place for this because it has a history of flooding.”
Lisa Knoll, First Street’s chief marketing officer, defended the company’s work, saying it uses high-resolution elevation data and incorporates levees, flood walls and other flood control measures.
But Knoll acknowledged that national models cannot fully capture some highly localized conditions, particularly in dense urban areas where street-level drainage systems can influence flooding.
While the IDRT model is not constrained by those uncertainties, the same reliance on real-world claims that sets it apart might also create potential blind spots in the data, some researchers said.
Because the model learns from properties that flooded and filed insurance claims, its predictions may be less certain in neighborhoods where far fewer homeowners have flood insurance, said James Doss-Gollin, the Rice University professor.
“Lots of people don’t have flood insurance, so that’s a big limitation,” Doss-Gollin said.
Brody said Texas floods frequently enough that his team has enough data to compensate for the limitation, and they have tested whether the properties represented in its data systematically differ on factors used by the model, such as elevation and impervious surfaces.
“In Texas, it is so saturated.
We have more data than we know what to do with,” Brody said.
“But we can always have better data and more data.”
The transparency gap
Experts say the Texas A&M researchers have been unusually transparent about how their model works, and its limitations.
That is far from the norm in today’s climate risk research, according to Adam Pollack, an Iowa University professor focused on flood risk management.
In a 2026 paper led by Pollack, researchers examined over 250 highly cited climate risk studies published in 2021 and 2022 and found that only 4% fully shared both their data and code.
He said that was one reason he preferred the IDRT’s approach to many commercial products – including First Street’s – whose underlying methods can be harder for outside researchers to scrutinize.
“We’re so lucky that it’s being transparent about its limitations,” Pollack said, pointing to the peer-reviewed paper laying out the Texas A&M researchers’ methodology and performance tests, as well as disclaimers throughout the Buyers Aware website.
But it has been challenging to get IDRT’s app directly in front of homebuyers.
Brody’s institute spent about two years working with the Houston Association of Realtors to integrate its flood risk estimates into HAR.com – the association’s home search portal and one of the most widely used real estate websites in Texas.
Last year, however, shortly before the project’s planned launch, the association abandoned the efforts amid complaints from real estate agents who questioned high-risk ratings on some properties and worried they could hurt sales.
Bill Baldwin, the association’s spokesperson, told the Chronicle at the time that leaders simply could not convince enough members that it was the right time to display the risk ratings directly on listings.
A flood protection advocate himself, Baldwin nevertheless said he believes pressure to make such information more accessible will only grow.
“The consumer adamantly wants this information,” he told the Chronicle.
“They’re going to find the information somewhere.”
For Pollack, tools like Buyers Aware show why improving the science is only part of the equation.
The other challenge, he said, is making that information accessible and useful enough to guide real-world decisions.
“Flood risk is really uncertain,” Pollack said.
“But we can make good decisions in the face of uncertainty.”
Analysis
Hover over highlighted words in the article to view the associated bias or fallacy analysis.