[Note that this article is a transcript of the video embedded above.]

Do you remember the summer of 2022 when a record drought had gripped not only a large part of the United States, but most of Europe too? Reservoirs were empty, wildfires spread, crop yields dropped, and rivers ran dry. It seemed like practically the whole world was facing heatwaves and water shortages. But there was one video that warned against hoping for rain, at least not for big storm right at first. Rob Thompson, a meteorologist and professor at the University of Reading, shared a little backyard experiment: cups of water being inverted on top of grass with varying moisture levels in the soil. The results seemed to show that the dry soil absorbed the water much more slowly than the wet grass or normal summer conditions. This video was shared across the internet as a viral reminder that, contrary to what you might think, droughts can increase the impact of flooding. But is that actually true? Does dry soil absorb moisture more slowly than wet soil, and could a storm after a drought cause more runoff and worse damage than if the ground was already wet? No matter what your intuitions say, the answer’s a little more complicated than you might think. And of course, I built some garage demonstrations to show why. I’m Grady, and this is Practical Engineering. Today we’re exploring the relationship between droughts and floods.

Of all of the natural disasters we face, floods are among the worst. There have been more than 30 floods in the US since 1980 that caused over a billion dollars in damages each! And that’s not including hurricanes. In fact, floods are so impactful that I’ve already made a whole series of videos about how dangerous they are and many of the ways that engineers work to reduce the risk of flooding or at least reduce the damage they cause. Many of those flood infrastructure projects are based on a “design storm,” essentially a made up flood used to set the capacity or height of a structure. For example, the storm gutters on your street might be designed to carry the 25-year storm. Many spillways for dams are designed for a flood that is unlikely to ever occur called the Probable Maximum Flood. Of course, we just can’t run full-scale tests on flood infrastructure. Despite architects and contractors saying we always rain on the parade, civil engineers can’t call down a flood of a particular magnitude and duration from the heavens. And even if they could, it would be an ethical gray area. So, engineers who design water infrastructure instead use models to help estimate various magnitudes of flooding and predict how the built environment will respond.

There are all kinds of hydrologic models that can simulate just about every aspect of the water cycle you could imagine, but modeling basic storm hydrology is actually pretty simple. It’s usually broken into three steps. Precipitation is exactly what you would expect: how much rain actually falls from the sky and hits the surface of the earth? Transformation describes what actually happens to those raindrops as they run along the ground and the timing of how they combine and concentrate. But in between precipitation and transformation, there’s a third step. Because not all those rain drops run off and reach a river or stream. Some of them get stuck in puddles and ponds (called abstractions), some evaporate, and some soak into the ground.

I say all this to point out that the engineers and scientists who study flooding have put a great deal of thought and research into the how, where, how much, and why rainfall soaks into the ground. It’s the third leg of the “estimating how bad floods can be” stool (a stool, by the way, I spent a good part of my education and professional experience sitting on). And of course, there’s a litany of factors that affect how much precipitation is lost to infiltration into the earth versus how much runs off into rivers and creeks: temperature, vegetation, season, land use, soil type, and more. But one of the factors is more important than any other: soil moisture. And it shouldn’t be that surprising. How much water is being held between those tiny grains of silt, sand, or clay plays a pretty big role in how much more water can flow in.

Maybe you’re starting to see what I’m getting at here (and I promise the demos are coming but I think it’s important to know the theory first). One of the most beloved mathematical expressions of hydrologists everywhere is Horton’s equation. Looks a little intimidating, but it’s much simpler as a graph. This logarithmic curve shows the infiltration rate we can expect during a rain event of a given magnitude over time. At first, when the soil is driest, the rate of infiltration is highest. As rainfall continues to soak the soil, less water is absorbed, and the infiltration rate slowly approaches a steady state.

The inputs to Horton’s equation are fine for a laboratory, but they’re not really easy to estimate in a real world scenario, so most hydrologic models don’t use it. One of the simpler infiltration models actually used in engineering is the Curve Number method, originally developed by the Soil Conservation Service in the 1950s. Here, instead of esoteric laboratory variables, infiltration rates are tied to actual soil types and land uses we can estimate in the field, and this is meant to be dead simple. You too can be a civil engineer by simply picking the right number from a table and feeding it into a model. In fact, let’s try it out. My backyard is an open space, I would say in good condition, with mostly clay which is hydrologic soil group D. So my curve number should be 80.

I won't make you go through the calculations, because we can make the computer do them. This is the Hydrologic Modeling System, a free piece of software available from the US Army Corps of Engineers. I’ll plug in my backyard curve number, plug in a storm with a constant rainfall over a day, and push go. The bars show the total amount of precipitation for each time step. The red portion shows the losses and the blue portion shows the runoff. At first, all the precipitation goes toward losses as the rainfall gets caught in abstractions. But once the puddles fill up, some runoff starts to occur. You can see that, for a constant rate of precipitation, runoff increases over time, and infiltration goes down, just like we saw with Horton’s equation. I know we’re in the weeds just a bit, but I think it’s important to know that we have technically rigorous ways to describe our intuitions of how floods work. The Curve Number method (along with many others) are used across the world by engineers to characterize floods and even to calibrate hydrological models to actual floods. Of course models are never perfect, but at least they’re based on real science. Water fits into the spaces, the interstices, between soil particles. The more water there already, the harder it is for more water to flow in. But you don’t need a graph. You can see it for yourself.

I hammered a clear tube into my Curve Number 80 backyard, and we can watch the water flow into that clay soil with grass of quote-unquote “good condition.” This is actually a crude version of an actual scientific test apparatus called an infiltrometer, but this isn’t strictly scientific. The real test involves hammering the tube deeper to prevent lateral spread and maintaining a constant level to remove water pressure as a variable. But, hey, this is just a youtube demo, and I wanted to push my kid on the swing instead of babysitting the water level in a clear tube for 45 minutes.

I did take the time to graph the water level for the duration of the experiment so you can see the results more clearly. The level drops quickly at first and slows down to roughly a constant rate, just as the theory predicted. Some of that slowdown is because of the decreased water pressure over time, the variable I didn’t control, but it’s mostly because the soil became saturated, making it harder for water to infiltrate.

Just for fun, I ran another experiment in the garage with a tube full of sand. FYI, that’s roughly equivalent to “Natural Desert Landscaping” with an associated curve number of 63. Are you feeling like an expert at this yet? It’s a little harder to tell in the sand because the water flows so quickly, but it does in fact flow more quickly at the beginning before the sand is saturated. Once it saturates, the infiltration is more or less constant, just like we would expect. The reason for the sand demo is this: we’ve left out a key consideration so far which is the initial conditions. How much water is in the soil at the start of the event? If it’s a lot, you would assume there would be less infiltration. If the soil is dry, you would assume infiltration would be greater. Is it true? Let’s try it out!

Again, the sand is maybe a little bit too porous for this demonstration, and my method for adding the water isn’t so precise either. But, just paying attention to how quickly the tube fills up with water with the valve fully opened, the dry sand takes longer. That’s because more of the water is infiltrating into the soil. The wet sand is like starting halfway down the Horton curve. But that wasn’t a super satisfying result, so I put some potting soil into the tube next (Curve Number 86). I ran it once dry, then ran it again after the soil was saturated, and lined the shots up side by side. This time you can clearly see the difference. Water infiltrates into the unsaturated soil much more quickly, but once it does, it infiltrates about the same rate as the already wet example.

We have a word for this: antecedent conditions. Most of the factors we talked about that affect infiltration rates don’t stay the same over time. They change. Many hydrologic models use average conditions as a starting point, but the real world isn’t very average. Vegetation is seasonal; temperatures fluctuate; watersheds experience fires and droughts (hint hint). How wet a watershed was before the storm is an important factor in determining how much runoff will occur. According to all the theory and practical examples I’ve shown, a wet watershed will absorb less precipitation, so flooding will be worse. And the opposite is true for a dry one. More water will soak in, making flooding less impactful. But, that seems contrary to the video I showed in the introduction, and do you really think I would make a video called “Do Droughts Make Floods Worse” if the answer was just, “no”?

It turns out that certain kinds of soil, when they become very dry, also become hydrophobic. They actually repel water. This is not a super-well-understood phenomenon, but it seems that under very dry conditions, waxes, plant root excretions, and the action of bacteria and fungi create a layer at the surface that reduces a soil’s affinity to water. If you’ve ever forgotten to water a houseplant for a while, you may have experienced this yourself. It’s hard to get the water to soak in at first, and many gardeners will actually fully submerge a potted plant to properly water it.

Because it’s a finicky phenomenon, I had a little trouble creating water repellent soil in the garage, but luckily, hydrophobicity is interesting enough to be a fun kids toy. I bought some hydrophobic sand and put a layer of it on top of my regular sand to simulate this effect of soil water repellency. You can see clearly that the repellent layer slows down the infiltration of the water. It still gets through, but it happens a lot more slowly compared to if it weren’t there. So, why doesn’t this effect show up in the theory (or least the theory of flood modeling)?

There are a few reasons: number 1 is that most hydrophobic soil effects disappear pretty quickly after the soil gets wet. It just doesn’t last that long, as you know if you’ve dealt with it in your potted plants. Number 2 is that it’s a phenomenon that hasn’t been well-characterized in terms of what soils experience repellency and under what conditions. There’s no nice table for an engineer to look up values. But number 3 is the biggest one: there are other antecedent factors that just end up being more important. Very high soil moisture before a flood is much more likely to lead to severe flooding than very low soil moisture in most cases. The extreme example of this is rain-on-snow flooding, which contributed to the 2022 flooding in Yellowstone National Park. But there is one big exception to this rule: fires.

When organic stuff burns, some of that volatile material creates hydrophobic properties in the underlying soil, reducing its ability to absorb rainfall. That effect plus the loss of vegetation on the surface means that the potential for flooding after a fire increases dramatically. Storms after wildfires are known to create massive floods, mudslides, and erosion, so there is a lot of research into understanding this phenomenon.

So what’s the answer? Are floods worse after a drought? Dry conditions do kill plants and grasses that slow down runoff, they create hydrophobic soils that briefly keep water from soaking into the ground. And, they often make fire conditions worse which in turn, can lead to more impactful floods. But droughts also leave the soil drier than average, increasing its ability to soak up rainfall. In many cases, a flood after a good soaking rainfall is going to generate far more runoff than a flood after a drought.

Rob told me he was completely surprised by the response to his video, especially since he only spent a few hours making it. His goal was to show that, under certain conditions, flash floods can be worse when the underlying soil is very dry. But I suspect if his demo lasted a little bit longer (and his setup was a little more rigorous), the results may have looked a little different. And on the other hand, most models used by engineers to estimate floods assume that infiltration always goes up as soil moisture goes down, completely neglecting the fact that some soils lose their affinity for water at very low moisture levels. One statistician famously said that, “All models are wrong, but some are useful.” And even something as simple as the flow of water into the soil has so many complexities to keep track of. Like most answers to simple questions in engineering and in life: the answer is that’s it’s complicated.