The St. Anthony Falls Laboratory was built into the banks of the Mississippi River near downtown Minneapolis in 1938. Since then researchers there have used it to study rivers, constructing large, elaborate physical models that make use of the Mississippi itself.

One of the SAFL engineers, Robert Gabrielson, built a working model of a river in a 10-foot by 13-foot box with an artificial turf floodplain and a miniature river channel running down the center. He made the levees out of PVC trim — flexible, plastic strips — and piped actual Mississippi River water through the channel, along with a steady supply of sediment.

He adjusted the water’s flow rate to simulate normal river conditions as well as floods. The lab is equipped with sophisticated scanners that record high-resolution elevation and water surface data. We added tiny wooden houses to represent communities on the floodplain. Though the model isn't to scale, nor a scientific study, it effectively simulates the basic physics of the conditions facing many of the people who live near the nation's 100,000 miles of levees.

The four river scenarios we modeled were: no levees; two levees of the same height on both sides; a higher levee on one side than the other; and a single "setback" levee, built further back, which preserves some of the natural floodplain.

For each scenario, we gradually increased the flow rate of the water running through the channel — from 50 gallons per minute to almost 200 gallons per minute in some scenarios.

Use the buttons below to see what happens as the flow increases.

Just like in the real world, when we added levees to the model, water ran faster and higher when trapped between levees. When people on one side of the river raise their levees, they can push a wall of water onto their neighbors across the river. The people with the lower levee often have no recourse but to try to build up their own defenses.

What Happens in the Real World

The model we built in Minnesota showed a few simple examples of what could happen when people live near levees. But what do the effects look like in the real world?

Two university researchers did a study of the effects of levees in 2011. Reuben Heine of Augustana College in Rock Island, Illinois, and Nicholas Pinter at the University of California-Davis looked at 13 levees in Illinois and Iowa. Every levee they studied increased flooding across the river or upstream—sometimes by up to five feet.

Heine and Pinter proved what many scientists consider to be basic physics: during a flood, extra water always has to go somewhere. If a river has levees on only one side, some water is pushed across the river, flooding unprotected areas even more. But if a river has levees on both sides, the water between the levees piles up. In both cases, the water backs up, adding extra risk to nearby unprotected land upstream of the levee.

Scientists consider it a given that levees make flooding worse. It’s mentioned in Earth Science 101 textbooks, said Anne Jefferson, a geology professor at Kent State University. “It’s not terribly difficult to work out what happens when you’re squeezing things along a nozzle.”

Fredrik Huthoff, an engineering consultant in the Netherlands who specializes in flooding and river science, compared this to a traffic jam: River water backs up behind the bottleneck caused by the levees.

Heine and Pinter only looked at 13 levees because there isn’t enough good data. Heine told us that although they started with 203 possible sites to study, they had to throw out nearly all of them because the government-funded stream gauges necessary to study levees weren't collecting enough data, or weren't in place before the levees were built, or weren't in the right place in the river relative to levees. Still, in the 13 cases that were usable, the pattern was clear.

After Levees Were Built, Floods Got Worse on the Wabash River Near Mt. Carmel, Illinois

To rule out other reasons why water levels might rise — such as a landslide or extra sediment that might have raised the bottom of the river channel — Heine and Pinter also looked at gauge data from when the river is flowing, but not flooding. In these cases, they found no differences in pre- and post-levee water levels.

Jonathan Remo, a Southern Illinois University professor who studies flooding in the Midwest, told us he wasn't surprised by Heine and Pinter's study, since it confirms what's been shown through computer modeling, and echoes decades of research on how levees and other engineering projects have aggravated flooding, particularly along the Mississippi River.

Among the most striking examples Heine and Pinter found was in the gauge data near Mt. Carmel, a southern Illinois town just across the Wabash River from Indiana. The Army Corps built levees on both sides of the river near the town in the late 1960s. Heine and Pinter found that the two levees constricted the Wabash floodplain and forced water levels nearly three feet higher than they would have been for a minor flood, the kind of flood that has a 10 percent chance of occurring in a given year. Most of that water was pushed upstream of the pinch point into the area where the Wabash meets the White River nearby.

We asked Heine to map out roughly what that 10 percent chance flood would look like near Mt. Carmel before and after the levee: