---
title: 'An Engineer''s Perspective on the Texas Floods'
source: 'https://youtube.com/watch?v=3FfMzWa6LKg'
video_id: '3FfMzWa6LKg'
date: 2026-07-28
duration_sec: 1394
channel: 'Practical Engineering'
---

# An Engineer's Perspective on the Texas Floods

> Source: [An Engineer's Perspective on the Texas Floods](https://youtube.com/watch?v=3FfMzWa6LKg)

## Summary



## Transcript

This is an animation of the weather radar  in central Texas starting at noon on July 3,   2025. You can see there was torrential rain across  the state throughout the afternoon from remnants   of Tropical Storm Barry. But focus on this area  northwest of San Antonio. Around midnight on July  
4, a severe storm gets stuck in this area  and just stays in place for several hours.   When you put it in context with the rest of  the system, it looks kind of insignificant,   but that little storm dropped enough rain to raise  the Guadalupe River higher than ever in recorded  
history, at least in the upper part of the basin.  The water quickly rushed through summer camps,   RV parks, and rural communities in  the middle of the night. And the   result was one of the deadliest inland  flooding events in the past 50 years.
I live not too far from some of the worst-hit  areas, and although my family wasn’t directly   affected by the weather, it’s been a  tough situation for me to wrestle with,   personally. I spent the better part of  my career as an engineer thinking about  
flooding and designing projects to cope  with it. I’ve worked on and played in   the Guadalupe River. And I have kids who are  getting close to summer camp age. As a dad,   it’s almost impossible to comprehend a tragedy  like this. As an engineer, I’ve dedicated a large  
part of my professional career to understanding  events exactly like it. So, as I’ve ruminated   about this flood over the past few months,  I’ve collected some thoughts that might be   worth putting into the world. Let’s take a look  at this event through an engineering lens, talk  
a little bit about how technical and regulatory  decisions play out in the aftermath of tragedy,   and see if any lessons become apparent. I’m  Grady, and this is Practical Engineering.
One of the fundamental problems we face in  engineering, and really life in general,   is that we can’t predict the future. That  sounds like a ridiculous thing to say,   but out of that uncertainty comes the framework  for how we think about so many things. Because,  
we have to make all kinds of decisions - many  of them with extremely high stakes - in the   face of the unknown. In civil engineering,  a lot of the loads we account for come from   the most classically volatile and  unpredictable aspect of the earth:  
the weather. Wind, ice, snow, waves,  and rain - you cannot look ahead 50   or 100 years and know what forces a structure  will be subjected to. You just have to guess. And that’s a pretty hard thing to do,  especially because you tend to have two  
opposing forces pushing your guess around. On  the one hand, caution dictates overestimating   forces to leave a wide margin of safety, but on  the other hand, costs and budget constraints tend  
to push the estimate the other way. I can  make this dam taller or this bridge higher,   but it’s going to cost me a lot more money, and  maybe it’s not necessary. So how do you draw the   line? The same way we try to predict the future in  so many other parts of life: we look to the past.
Surely past performance is an indicator of future  results, right? I know that’s a stock line,   but what else do we have? Over the years, we have  gone to considerable lengths to apply historical  
data to predictions of future floods. Of course,  this gets pretty complicated. One of the resources   widely used in the United States for decades  is Technical Paper 40, published in 1961. It  
represents a monumental effort to compile rainfall  data across the contiguous United States, find   probability distributions that fit the data, and  map the results. It’s divided up by duration and  
recurrence interval, so you get this big group of  separate maps. But what is a recurrence interval? I’ve talked about the so-called 100-year  flood in a few of my videos, but it’s a   concept so widely misunderstood  that it’s worth explaining again,  
especially because it’s so relevant to the  Guadalupe River flood in July. We can’t   really use historical data to determine  when a flood might happen in the future,   but we can make an estimation about how probable  one might be. The bigger the flood, the lower the  
probability that it might occur. So there’s a  relationship between probability and magnitude.   In hydrology, we often express the probability  as a quote-unquote “return period,” which means,  
on average, how many years you would expect  to pass before you see that magnitude equalled   or exceeded again. But that “on average” is  doing some heavy lifting in the definition.
This terminology is debated endlessly in the  hydrologic community because saying something like   the 100-year flood has an underlying implication  that storms are cyclical; that somehow if a   particular magnitude of storm was to occur,  we might have a period of security before it  
happened again, or the flipside: that if a flood  hadn’t occurred in some time, we might be more   “due” for it. And that’s just not how it works.  Every year, the atmosphere rolls the metaphorical  dice to see what the biggest one is going to be.  
The odds of rolling a two or snake eyes in craps  are 1 in 36, but if you go 35 rolls without a   snake eyes, the odds of rolling it on the next  one haven’t changed. The dice don’t remember  
what happened before. No one calls snake eyes the  36-roll throw because we understand it’s possible   to do it twice in a row, and it’s possible to go  a lot more than 36 rolls without getting one. So  
why do we call it the 100-year flood? Probably  because the only good alternative is the storm   with a 1% annual exceedance probability. Just  doesn’t roll off the tongue. But it is the   technically correct definition: the 100-year  rainfall is the depth of precipitation (over a  
given duration) that has a one percent probability  of being equalled or exceeded in a given year.   It’s a tough concept to wrap your head around,  but it’s fundamental to engineering hydrology.
If you take a look at these maps, you can  see that the 100-year rainfall over a 24-hour   duration in Kerr County, Texas is around 9.5  inches (or about 240 millimeters). But again,   this is from 1961. And it’s based entirely  on historical data. So there are decades of  
rainfall not included in this analysis, not  to mention limitations in the statistical   methodology and data processing methods of  the time. TP 40 wasn’t the only resource for   precipitation frequency data in the US,  but it was probably the most widely used  
until Atlas 14 came along, or is coming  along (it’s still a work in progress).   NOAA has been working to update this information  with the entire historical record and more   rigorous statistical methods. For most of the  US, this is easy to navigate online. Just mark  
a spot on the map and you get this table of  values and confidence intervals for a range   of durations and return periods. And you can  see that the 100-year, 24-hour precipitation   in Kerr County is 11.5 inches (or nearly 300  millimeters). That’s a pretty big jump from  
the 1961 estimate - an increase of about  20 percent. What was the 100-year rainfall   in 1961 is now just the 50-year storm AND look  at those confidence intervals! 8 to 16 inches.
I know this is kind of long-winded, but the  whole point I’m trying to make here is the   tremendous uncertainty we have when it comes  to hydrology. In some ways, this rainfall data   is extremely rigorous, and I couldn’t even  begin to explain some of the statistical  
methods used to develop it. It serves a really  important purpose in the world of engineering,   planning, and emergency management. But in  another sense, it’s almost meaningless. And I   can show you a few of the reasons through  the lens of the Guadalupe River Flood.
Here’s an hourly map of the rainfall  that hit central Texas on July 4, 2025.   That yellow area is the watershed for the upper  Guadalupe River. When I loop through it again,  
you can see that cell right there caused the  majority of the flooding you probably read about   on the news. It was there and gone in four hours.  More rain came in later that morning and the next   few days, but this was a classic flash flood:  A relatively short burst of heavy rainfall on  
a small, steep, rocky basin, where most of it  runs off into a river within minutes or hours.   Here’s the thing: hourly rainfall records  weren’t very common until the 1940s. I  
counted about 100 rain gauges used by Atlas  14 within a 50 mile radius of Hunt, Texas,   where most of the fatalities occurred.  None had hourly records before 1940,   and of the group that did collect hourly data,  only four had a record longer than 70 years. That  
might seem like enough data to understand flooding  in the area, but let me show you why it’s not. Here’s that loop of rainfall again. What do you  see on this map? Because I’ll tell you what I see:  
enormous spatial variability. If you were  to pick four random pixels on this map,   how good a picture do you think it would give you  of what really happened? That’s essentially what   we’re doing with rainfall frequency analysis.  Compared to modern data collection methods,  
like the radar rainfall I showed, our  historical records are extremely sparse,   especially for data that varies so significantly  across space. Imagine trying to recreate the Mona  
Lisa from scratch with just a dozen random pixels. estimate flood probabilities have never  even seen an event of the magnitude we’re   trying to use them to predict. There’s  a whole lot of extrapolation going on.
To hammer this point home: This is the  24-hour rainfall totals for the flood,   and you can see that even within this single  watershed, some areas saw extreme precipitation,   while others just got an inch or 25  millimeters of rain. And actually,  
I mapped the percentage of the 100-year rainfall  that this storm amounted to, and you can see,   at least in the Upper Guadalupe Basin, only a  small area got close to the 100-year rainfall.  
For most of the watershed, this was  more like a 2- or a 5-year storm. And here’s what makes this even tougher:  When we’re talking about flooding, we don’t   actually care too much about rainfall.  We care about the outcome of rainfall,  
specifically the rise in a river or stream.  Here’s the graph of a stream gage upstream   of Hunt during the flood. You can see that,  starting around 2:00 on the morning of the 4th,   the river rose by 20 feet or 6 meters in  three-and-a-half hours. A little further  
downstream, similar story. Starting at 2  AM, the river went up 35 feet or nearly 11   meters in 3 hours before the gage broke. That is a  staggeringly fast increase. In a hydrologic sense,  
it’s practically a wall of water. And the  results were devastating. In Kerr County,   there just wasn’t enough time to coordinate an  evacuation. More than 100 people were killed,   many of them children. So a rain gauge here, or  here, or here would have completely missed the  
fact that the watershed it was within  was experiencing the flood of record. That’s the value of measuring the thing you  actually care about. Just like precipitation,   you can take historical stream gage data,  fit it to a probability distribution,  
and get a sense of the likelihood of major  floods in the future. But these gages are even   more sparse in coverage than rain gauges,  their records often don’t go back as far,   they can go offline, ironically as a result  of flooding, completely missing the peak.  
Engineers or hydrologists actually often visit the  affected area and map the high water line after a   flood to validate and confirm the data from stream  gages (or to fill in the gaps if one breaks). So,  
although they serve an extremely important  role, most of the time when engineers are   trying to predict flooding or its effect  on infrastructure and the built world,   instead of using stream gages, they’re using  hydrologic models to convert rainfall into  
runoff and flooding, a process that introduces  a whole new set of uncertainties into the mix. And there’s one more thing. Everything we’ve been  talking about so far is predicated on a crucial   underlying assumption: temporal stationarity,  basically, the idea that the distribution of  
extreme events doesn’t change over time - or put  another way - that future precipitation can be   represented by past observations. But, even though  those past observations are relatively sparse,  
in a lot of cases, we can already see that it’s  probably not a great assumption. I understand this   is a point of pretty strong contention in the  public discourse. But within the professional   community of hydrologists, engineers, and climate  scientists, it’s not really a question of “is the  
climate changing” but more a question of how much,  how quickly, and where the effects of that are   most pronounced. For example, in the Texas Volume  of Atlas 14, the team tested for long-term trends  
in the data. They found some scattered weather  stations that did show an increase in extreme   rainfall over time; most of them didn’t. Other  studies have found more pronounced increases by   looking at only the past few decades. So there are  no broad statements that capture the complexity  
of the situation as we understand it, and  importantly, this is a tough thing to figure out. Say you have 100 years of historical data. How  many 100-year floods happened within that time?  
Could be a few. Could be none. So, especially for  very extreme events on the 1-in-a-century scale,   there’s a lot of uncertainty when it comes to  teasing out any trends. That said, there is a  
strong consensus among the various climate models  and recorded data that a warming atmosphere has   already resulted in an overall increase in  the intensity and frequency of rainfall,   a trend that will likely continue. And you can  see why that poses a problem. Particularly for  
infrastructure with a design life of 50 to 100  years, we need to design not just for the storms   of today but those decades in the future,  and our current methods of doing that is,   on average, systematically underestimating  them if we assume a stationary climate.
Just to be clear, I’m not trying to blame a  flood on climate change. Although attribution   studies can estimate the contribution  of extra energy in the climate system,   there’s no way to ascribe any particular weather  event to global warming deterministically. For  
many places, it might not even be a major source  of uncertainty compared to all the other factors   I’ve mentioned when it comes to predicting  the magnitude of future floods. My point   is that it’s just one more confounding  aspect of estimating flood risks. And  
it gets to the heart of the entire issue.  Because why does any of this even matter? There‘s been a lot of discourse about what  should have happened before the storm and   what should be done in its wake. But before you  can take any action to mitigate flood impacts,  
you have to know what the actual risks are. On  the upper Guadalupe, we’ve seen it with our eyes,   but how many similar watersheds just got lucky  that night, or really, any night? I think you’ll   agree with me that this is complicated stuff. And  humans are notoriously bad at using probabilities  
and risks to make decisions. Almost nothing in  our biology is optimized for long-term, rational   decision-making about rare and extreme events.  Almost every day of everyone’s lives, there’s not  
a flood. That makes it really tough to consider  it as a priority and devote resources toward   preparations. And I think part of the problem  is that we rarely talk about the uncertainties.
Even within the field of engineering, where we  should know better, we have a strong tendency   to treat everything deterministically. It  sure makes things a lot simpler. Take the   bold number in the table, plug it into  your equations and computer models,  
and just forget that those uncertainty bands even  exist. In some ways, it makes sense. Ultimately,   you do have to choose a number: how high to  build a bridge or how large a culvert to install,   or how wide to make a spillway. But, in a lot of  cases, those decisions get translated into a sort  
of confidence that doesn’t actually exist. The  concept of the floodplain is a perfect example. In the US, a lot of the framework for how we  think about and prepare for floods comes out  
of the National Flood Insurance Program.  And to participate in this program,   communities are required to regulate what  happens in the floodplain, or more specifically,   what and how things get built there. And so, a  fundamental part of regulating the floodplain  
is deciding where it actually is and isn’t. We’re  not going to dive into that process, but billions   of dollars have been invested in making these  maps and keeping them up to date in the US. If you take a look at one, it’s a lot to parse  depending on the location. There are quite a few  
different hazard areas with different meanings.  The simplest for riverine locations is the base   flood, essentially the 100-year flood. Some  maps show the 500-year flood as well. Many   maps show the floodway, which is kind of the  main part of the channel needed to pass floods,  
so it’s usually regulated more strictly.  But there’s something I notice when I look   at floodplain maps. All of these zones are  bordered with nice crisp lines. You’re inside   the floodplain here, and you’re outside of  it here. And property owners often go to  
great lengths to refine these maps; to shift the  line just slightly and reduce their regulatory   responsibilities. But consider everything  we’ve talked about with estimating flood   risk and ask yourself, what’s the difference  in the risk profile between here and here?  
Is it enough to have a sharp line between  them? And if not - if the true situation is   more nebulous - is the map doing a good job  of communicating flood risk to the public? Because, just to be clear, that is one of the  stated purposes of floodplain maps. Of course  
you need to delineate zones clearly to be able  to regulate where permits are required and where   buildings can be built and so on. But, to me at  least, it sends a complicated message to have   this binary definition of inside the floodplain or  outside of it as a way to explain to individuals,  
homeowners, renters, and the general public  about the risks that they’re actually exposed to. You look at these maps and there is  absolutely no indication about uncertainty,   despite the fact that almost every step of  the process that goes into creating them has  
huge margins of error. And then, when we get  more historical data, or land uses change,   or our understanding of the floodplain evolves,  and we try to change the map, that immediately   sows distrust. You hear it all the time (at  least if you run in similar circles as I do):  
“We’ve had two hundred-year floods in the past  5 years. These engineers don’t know what they’re   talking about…” Part of that, of course, is just  a misunderstanding about what the hundred-year   flood actually means, but part of it is that  we don’t do a good job communicating risk and  
uncertainty well. The meteorologists get the same  thing. People get salty when forecasts are wrong   without any acknowledgement at all that the job  is essentially predicting the future. You know,   it’s wizard stuff. Weather is really  complicated, and I think we have a  
lot of room to grow in how we discuss and  disseminate the things we don’t know for sure. Because flooding is capricious. If you look  back at the maps from July 4, you can see a   lot of places where rainfall was more intense than  in Kerr County and the Guadalupe River. Many areas  
of central Texas received more than the 100-year,  24-hour precipitation from Atlas 14. And there   were severe storms and flooding across the region  in the days that followed as well. But nearly all   the fatalities happened in this one place. I  don’t have a good answer for why. Maybe some  
combination of timing, warning systems, the rural  location, differences in floodplain regulations,   and plain bad luck. I think scientists, engineers,  and emergency planners can probably learn a lot  
by simply comparing the flooding between  Kerr County and some of the other areas   in central Texas hit by this storm system, and  why the outcomes were so drastically different. My heart goes out to the victims and their  families who were affected by this flood.  
of risks can go so underappreciated that  we wouldn’t bat an eye at having such a   large population of people sleeping in  the floodplain of a flashy watershed.
I think there are a lot of lessons to learn here,  but the one that keeps coming back to me is about   communication. People can’t act to reduce  their risk unless they can internalize what   it actually is. Professionals think about these  issues every day; they have technical training,  
knowledge, and experience to make informed  decisions about infrastructure, land use,   and zoning. But most people don’t have the  same cognizance of the hazards. You can’t   blame them. It’s a crazy world we live  in, and even individuals who live, work,  
and play in areas at risk of flooding might not  come face-to-face with the danger in their entire   lives. Like I said, weather is complicated,  and we don’t all have the headspace to try and  
understand spatial variability, annual exceedance  probabilities, climate stationarity, and so on. So I think the professional community  has a responsibility to improve how we  
communicate flood risks to the public,  not only for accessibility but honesty.   We need to have language that anyone can  grasp, but we also need to be better about   acknowledging uncertainty. It sounds  counterintuitive, but I think facing  
the limitations of our understanding head-on  actually instills more trust than pretending   like we have all the answers. And when people  understand those uncertainties, they get a   deeper appreciation for how flood hazards vary  across the landscape, giving them more insight,  
not less, to prepare for what’s ahead. Thanks  for watching, and let me know what you think.