In October, KFF Health News published an analysis identifying 170 hospitals with the highest flood risk in the United States. Not only was I intrigued by the data presentation, which allows journalists to find vulnerable hospitals in their coverage areas, but also by the analysis: The KFF Health News team did not rely on the U.S. Federal Emergency Management Agency (FEMA) maps of flood risk that are journalists’ most accessible resource, despite their known limitations.
In this conversation, KFF Health News data editor Holly Hacker and I discuss how she and her team tackled this data-driven project.
How did this project come to be?
The impetus was Hurricane Helene last year, when there was flooding in North Carolina, Virginia and Tennessee. There was a little rural hospital in Erwin, Tenn., Unicoi County Hospital, where the floodwaters got so high so fast that they ended up having to evacuate patients and staff by helicopter from the rooftop. We’ve heard of hurricanes and devastation in Florida, in the New York area, along the coast, but this was something inland.
This hospital was built along a river, mainly in what FEMA calls a 500-year floodplain. What that means is, in any given year, there’s a 0.2% chance of a major flood occurring. Another reporter, Lauren Sasser, and I produced a story a year ago focused on Unicoi County and Erwin, and I pulled the FEMA flood maps for Tennessee and looked at Tennessee hospitals.
For this project, we wanted to find out: If it happened there, where else might it be possible?
I worked on this project remotely with Brett Kelman, who’s a reporter in Nashville, Tenn., and Daniel Chang, who is a reporter out of Miami, Fla.. My role was mainly data spatial analysis. Brett and Dan did a lot more talking to experts and other folks, going to some of these hospitals, getting pictures, as well as writing stuff.
Why didn’t you use the FEMA flood risk maps for this analysis?
The issue with FEMA maps is that in some cases they are outdated and do not account for climate change or a broader range of flooding causes. Our initial story a year ago showed FEMA maps and some of the hospitals, so you shouldn’t completely disregard them. And if you go online, there’s no one FEMA map; it’s really a collection of maps that get stitched together. And in some extreme cases, if you go online and you try to pull up the map, it’ll say, “No digital map available.” So then, you can hunt behind the scenes, and find a paper map from the 1980s or the 1990s. And so sometimes it’s also a matter of just what is the most up-to-date information.
The type of flooding that FEMA largely does not encompass is called pluvial flooding. It’s essentially when you get a lot of intense rainfall, and the ground can’t absorb it quickly enough. So, we know about coastal flooding, we know about rivers that tend to swell above their banks. But in the case of pluvial flooding, the question is: When you have a lot of hard ground or pavement and the rain falls, where does it go?
A company based in the UK called Fathom has been doing some sophisticated climate modeling, and their model incorporated pluvial flooding a lot better. They agreed to share their flood map data with us.
Fathom also predicted flood depths. They use satellite imagery and sophisticated computer modeling to predict flood depths. We set a minimum of at least a foot of flooding, but in some cases, you could see flood waters as deep as 15 feet. Even if the flood depth is 1 or 2 feet,, if you’ve got a lot of your equipment or generators or patient rooms on the first floor, that can be an issue.
Once you got that data from Fathom, what did you do with it?
We mapped all the hospitals in the United States as well as the flood data from Fathom, and then looked at where the overlap was. We tried to be conservative, so we actually looked at our initial results, a list of more than 300 hospitals, to see where at least a foot of flooding could potentially hit the building or the access roads. And then also we looked at how in some cases, Fathom found hospitals that FEMA would not have identified. We did not end up using the future models from Fathom, but they also have projections for 2040, for 2050, looking into the future.
I should also stress that what we didn’t do was speak to what, if anything, a hospital has done to protect itself. For example, there was a hospital that we mentioned in Coney Island, and they had been flooded years ago, and then they got a grant to prepare that building with windproof windows and a flood barrier.
Here’s where it gets nerdy. Some databases’ map point for each hospital was actually the building, and others actually map it to the middle of the street, which isn’t the same thing. To make sure our locations were accurate, we drew a buffer around the hospital to delineate its general area. When we looked at those locations with the flood layer, we generated an initial list of any hospital with at least 20% of the hospital’s area having at least a foot of flooding. We ended up with about 320 hospitals using that threshold. And then screensharing my GIS program over Zoom, the three of us went hospital by hospital to look at the predicted flooding.
Sometimes, all the flooding was off to the side, or there’s a big road that’s flooded, but there’s another road that leads out. So we tried to be conservative, because we didn’t want to seem like we were fearmongering. If anything, experts said this is probably an undercount.
FEMA categorizes 100-year floods and 500-year floods. Why are those terms confusing? How can journalists help the public comprehend them?
This idea of a 100-year flood can mislead people to think, “Oh, well, something happened 10 years ago, so it won’t happen in 90 years.” That’s not how it works. It means that in any given year there is a 1 in 100 chance of having this level of devastation. And it’s the same thing with a 500-year flood: It’s not that this is going to happen one time in 500 years, but it’s just a way of expressing the probability. These probabilities are based on past weather. They don’t take into account that these things are becoming more likely over time.
How are the federal cuts that many government agencies, including FEMA, experienced this year relevant to this reporting topic?
There’s a program that FEMA had announced that was supposed to be more probabilistic, communicating an area’s chance of flooding. It’s called the Future of Flood Risk Data Initiative, and they were trying to get that to incorporate more data. That initiative was announced under the first Trump Administration, but there really hasn’t been much of an update. And so, that was also the relevance of this story: This is happening as FEMA and NOAA, including the National Weather Service, have had cuts. Does this leave emergency planners and others even more vulnerable or without enough resources to plan, prepare and respond to something like this?
How does flood risk intersect with other challenges that rural hospitals are facing?
There’s a lot of tough choices that rural hospitals are having to make now, and then on top of that, you’re talking about something that could happen.
How did you find Fathom and decide to use their data?
People had talked to us before, mentioning First Street, and we had seen First Street cited a lot in some other media, such as The Washington Post and The New York Times. We reached out to them, and they weren’t interested in working with us, for whatever reason. And so we kept asking around.
Some flood and climate experts told us about Fathom, which was started by UK scientists, some of whom had worked at First Street. We were emailing back and forth, seeing if this could work, and then I saw that they were on the the list of speakers for the 2024 annual American Geophysical Union meeting in DC, which many climate scientists attend. So, I went to one of the panels where Fathom was speaking, and then got to meet some of the people. What was helpful about that was seeing that they were presenting at an academic conference. They’ve done peer-reviewed research. That also gave us more confidence. We were fortunate that Fathom was willing to share their data, because normally they charge for it.
What can journalists who don’t want to solely depend on FEMA maps do to explore flood risks in their coverage area?
Talk with whoever handles flood management or preparation in the area. It turns out for Charleston, West Virginia, five of the six hospitals were looking at a major flood risk, which was pretty unusual. So we talked to some folks; a whole department at West Virginia University is studying climate risk.
You can find out what the impact of a flood in a particular area would be, and you can also find out where else these researchers and officials are looking. Some researchers are actually developing their own models for their regions of concern. There could be something more specific to your area. A lot of the researchers who study this stuff were really interested in talking.
There’s a group called the Association of State Floodplain Managers. Journalists could talk to them about what data they use and what their thoughts are on this issue. It took us a while of just asking around. It’s just good old-fashioned sourcing.
Part two of this interview will be published later this week.
