The Hidden Engineering of Wildlife Crossings

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

This is the Wallis Annenberg Wildlife Crossing under construction over the 101 just outside Los Angeles, California. When it’s finished in a few years, it will be the largest wildlife crossing (*of its kind) on the planet. The bridge is 210 feet (64 meters) long and 174 feet (53 meters) wide, roughly the same breadth as the ten-lane superhighway it crosses. Needless to say, a crossing like this isn’t cheap. The project is estimated to cost about $92 million dollars; it’s a major infrastructure project on par with similar investments in highway work. And it’s not the only example. The Federal Highway Administration recently set aside $350 million federal dollar to fund projects like this. The reasons we’re willing to invest so much into wildlife crossings aren’t as obvious as you might think, and there are some really interesting technical challenges when you’re designing infrastructure for animals. I’m Grady, and this is Practical Engineering.

Roads fundamentally change the environments they cross through. And while on its face, it might seem that it’s always a disaster for wildlife, there are actually some winners amongst the losers. For vultures, crows, coyotes, raccoons, insects, and other decomposers, roads provide a buffet for nature’s scavengers. And they sometimes make for pretty good housing too, at least if you’re a swallow or a bat. In fact, cliff swallows are now so famous for nesting on the underside of highway overpasses that they’re often referred to as bridge swallows. The sides of highways have clear zones kept free from trees and similar obstacles for vehicle safety, but the lack of shade allows tender greens to thrive, creating a salad bar for species from monarch butterfly caterpillars to white-tailed deer.

Of course, especially in the case of deer, this can attract animals into spending time eating dinner in danger. And the truth is that roads mostly range from a mild inconvenience to totally catastrophic for wildlife. In the battle between the two, wildlife usually loses, and in more ways than just getting squished. The ecological impacts of roads extend beyond the guardrails. Habitat loss and fragmentation, noise pollution, runoff, and of course, injecting humans into otherwise wild places are all elements of the environmental challenges caused by roads. It’s actually a pretty complicated subject, and there are even road ecologists whose entire careers are dedicated to the problem. And it’s not just wildlife that’s affected.

According to the Federal Highway Administration, there are over 1,000,000 wildlife-vehicle collisions annually on US roadways. That results in tens of thousands of injuries, about 200 human fatalities, and over 8 billion dollars of damages per year. Even if you haven’t personally been involved in a collision like this, there’s a good chance that you know somebody who has. Along with the astronomical numbers reported by the FHA, it’s likely that a huge portion of wildlife collisions go unreported. There are lots cases that just don’t get counted, like if an animal is too small to notice, or if it survives the impact and escapes, or is collected by somebody practicing the dubious art of roadkill cuisine (yes, that’s a real thing and there are multiple cookbooks out there for it).

There’s a wide range of consequences from animal collisions, from minor vehicle damage to human fatalities. When you average them out, researchers estimate that in 2021, the average cost of hitting a deer was $9,100. Of course, the bigger the animal, the bigger the economic loss. For a moose, that number is over $40,000 per collision. Regardless of how you might feel about environmental issues and wildlife, the economic impacts alone can justify the sometimes enormous costs required to let them safely cross our roadways.

Luckily for the animal and human populations alike, there’s been increasing interest in reducing the negative impacts roads have on wildlife over the past few decades. I’m no stranger to infrastructure built for animals. It is fairly unusual for fish to get hit by cars, but they have their own manmade barriers to overcome, and I released a series of videos on fish passage facilities for dams you can check out after this if you want to learn more. Like aquatic species, there is a lot of engineering involved in getting terrestrial animals across a barrier. But fortunately, a lot of that research and guidance has been summarized in a detailed manual. I may not be a road ecologist, but I am an engineer, and I love a good Federal Highway Administration handbook!

One of the most important decisions about building a wildlife crossing is where to put one. You might imagine that the busiest roads are where most of the collisions occur. And it’s true up to a point. As the number of cars on a road increases, the percentage of wildlife crossing attempts that end in a safe critter on the other side drops, and the fraction that are killed grows. But, if we keep increasing the daily traffic numbers, something unexpected happens: the number of “killed” animals declines! Eagle-eyed viewers may realize that so far, this graph is incomplete; these percentages don’t add up to 100%. That’s because there’s a third category: “repelled” animals. As highway traffic increases, you reach a point where the vehicles form a kind of moving fence, and all but the most brazen bucks will turn away.

Road ecologists sometimes struggle to drum up support for wildlife crossings at high-traffic freeways (like the Annenberg crossing in LA) because of this effect. For some people, if they don’t see actual road kills on the shoulder, they struggle to accept the greater impact on wildlife populations. Habitat fragmentation caused by roads can be difficult for any species, but it’s especially hard-hitting for migratory species who HAVE to cross in order to survive and reproduce. For example, following the opening of I-84 in Idaho, biologists recorded the starvation of hundreds of mule deer mired in the snow, unable to cross to food sources.

And it’s not quite as simple as the graph makes it seem. A study by Sandra Jacobsen breaks down animals into four categories of crossing style. Some animals, like frogs, are non-responders who cross roads as if they aren’t there at all. Their wild instincts compel these animals to cross without regard for their own safety, and they’re often too small for most motorists to notice.

Next, you have the pausers, like turtles. These creatures, when spooked on the road or elsewhere, instinctively hunker down and stay put. While the shell of a box turtle might be impenetrable to a curious coyote, it is, sadly, no match for a box truck. Then you’ve got avoiders. This group often includes the most intelligent members of the local fauna. Grizzly bears, cougars, and other carnivores often fall into this category. For them, even low-traffic rural backroads can cause significant issues with habitat fragmentation, leading to poor genetic diversity. The small gene pool of a number of southern California cougars is one of the major drivers of the construction of the Annenberg bridge. Deer fall into the last category, speeders. As the name implies, these are fast, alert animals who, given the chance, will burst across a road to get to the other side.

But even these categories have their exceptions. The poster-cat of the US-101 project, a cougar called P-22, famously crossed the 10-lane highway and took up residence in the shadow of the Hollywood sign. There just is no one-size-fits-all approach for getting animals across roads. Engineers and ecologists use a wide variety of mapping, including aerial photography, land cover, topography, habitat, plus ecological field data and even roadkill statistics to choose the most appropriate locations for new wildlife crossings. And in many cases, what works for one species may be completely ineffective for another. So most designs are made for a so-called “focal species,” with the hope that it works well for others too.

But before you have a crossing, you have to get the animals to it. In most cases, that means fences, and even that is complicated. Do the focal species have a habit of digging under fences like badgers or bears? Well, then you’ll want to bury a few feet of fence to maintain its integrity. And where do they start and stop? Ideally, fences will terminate in areas that are intentionally hard to cross so animals don’t end up in a concentrated path across roadways. Sometimes boulders will be placed at the end of a wildlife fence to make it less likely that animals will choose to wander on the wrong side. But, inevitably, it happens. You don’t want to trap animals on the highway side of a fence, so many feature ramps or “jumpouts” that act almost like one-way valves for animals. There are even hinged doors for moderate-sized animals that allow wayward creatures to escape through fences.

Once you’ve got a site selected, the next big choice is over or under. It turns out that going under a road is often the easiest option. In fact, in many cases, existing bridges and viaducts can naturally create opportunities for wildlife to get across our roadways. Sometimes it’s as simple as building fencing to funnel animals into existing underpasses.

Another option for small animals is to use culverts as crossings. The engineering and materials for culverts are pretty well established since they’re used so much for getting drainage across roadways, so it’s not a big leap to do it with animals too. But it can be tricky getting them to use it. Since amphibians are also pretty lousy at walking long distances, it’s common to have many small tunnels installed near one another with special fencing to maximize survival. In some cases, they’re combined with buried collection buckets. During peak migration periods, the buckets are checked, and collected amphibians are manually transported across the road!

Larger animals won’t fit in a culvert (or a bucket), but there are some special considerations to getting them to travel beneath a highway bridge. Many animals are hesitant about dark areas during the daytime, so it's important to get as much natural light in as possible. Lighting also affects the vegetation that grows under a bridge. More light means more natural-feeling areas, which means more animals will be willing to cross under. And of course, keeping people out is important too. Disturbance from the public can really affect animals' willingness to incorporate a new, unusual route into their routine. Many crossings are designed with cover objects like logs, rocks, and brush that can help encourage a wider variety of wildlife to take advantage of the intended path.

But, for some species, underpasses just don’t work at all. You can’t FORCE a moose to do anything really, especially something like walking through a tunnel it doesn’t trust. In certain instances, the only effective way to allow safe passage across a road is over the top. For some particular focal species, an overpass might not need to be that grand. Canopy bridges just connect trees on either side of a road so primates and other tree-living creatures can get across. In Longview, Washington, there’s even a series of tiny bridges for squirrels, like the famous “Nutty Narrows” bridge.

Of course, the most impressive, usually the most effective, and often the most expensive wildlife crossings are designed as overpass bridges. Examples include the famous ecoducts of the Netherlands, overpasses of the Canadian Rockies in Banff National Park, and American structures like the Wallis Annenberg Wildlife Crossing. I actually have one of these nearby. Opened in 2020, the Robert LB Tobin Land Bridge crosses the six-lane Wurzbach Parkway in San Antonio, Texas. These are full-on bridges designed specifically for the use of animals.

Structures like these have all the same design issues as regular bridges for humans, plus their own engineering challenges as well. They have to hold up their own weight with a significant margin of safety, be designed to weather the elements for decades, and be inspected just like other bridges. They ALSO have to be engineered to be covered in thick layers of soil and vegetation (sometimes including trees), and be sized appropriately to accommodate focal species that might travel in huge herds or be wary of tight spaces. They have to be built to provide appropriate lines of sight for nervous crossers and often have walls that shield wildlife from the noise and light of the traffic below. One fun upside is that, at least in mountainous areas, the approaches can be a lot steeper than you might use for a vehicular bridge. An elk is pretty well suited for off-roading after all.

As for the design of the bridges themselves, they’re built a lot like highway bridges, usually beam bridges or arches, just with dirt instead of concrete for the deck. While the distance across a highway is long for a wandering moose, it’s not generally enough to require a structure of more heroic engineering like cable-stayed or suspension bridges. Unlike vehicular bridges, the approaches often flare out when viewed from above, making it easier for animals to locate the bridge and for better sight lines across it. This, plus the fact that they are usually covered in native vegetation, means that wildlife overpasses are among the most striking bridges you can see. It also means that from the perspective of the wildlife crossing them, these bridges can blend into the scenery. Ideally, a herd of pronghorn wouldn’t even realize they’re on a bridge at all.

It’s hard to think of any humanmade structures that have transformed the landscape more than modern roadways. They have an enormous impact on so many aspects of our lives, and it's easy to forget the impact they have on everything else that we share the landscape with. Sometimes when it comes to mitigating the negative impacts of roads on wildlife, the best thing is to just be more careful about where or IF we build a road at all. But for many of the roads we already have and the ones we might build in the future, it just makes sense - for safety, the economic benefits, and just being good stewards of the earth - to make sure that our engineering lets animals get around as easily as we can.

What’s the Deal with Base Plates?

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

A lot of engineering focuses on structural members. How wide is this beam? How tall is this column? But some of the most important engineering decisions are in how to connect those members together. Take a column, for example. You can’t just set it directly on a foundation, at least not if you want it to stay up. It needs a way to physically attach to the foundation. This may seem self-evident, maybe even completely obvious to most. But in that humble connection that’s so ubiquitous you rarely even notice it, there is so much complexity. Baseplates are the structural shoreline of the built environment: where superstructure meets substructure. And even understanding just a little bit of the engineering behind them can tell you a lot of interesting things about the structures you see in your everyday life. I’m Grady, and this is Practical Engineering.

Let me start us out with a little demonstration. If you’re a regular viewer, you know how much you can learn from our old friends: some concrete and a benchtop hydraulic press. I cast two cylinders of concrete about a week ago, and now it’s time to break them for science. These were cast from the exact same batch of concrete at the exact same time. For this first one, I’m pushing with a fairly narrow tool. I slowly ramp up the force until eventually… it breaks. I had a load cell below the cylinder, so we can see the force required to break this concrete. This scale isn’t calibrated, so let’s say it broke at 1400 arbitrary Practical Engineering units of force. Practicanewtons? KiloGradys? What would you call them? Now let’s do the same thing with a wider tool. At that same loading, this concrete cylinder is holding steady. In fact, it didn’t break until 3100 units. Here’s a trick question. Was the second cylinder stronger than the first one? Hopefully it’s obvious that the answer is no.

Most materials don’t care about force. I mean, in the strictest sense, most materials don’t care about anything. But what I mean is that the performance of a material against a loading condition usually depends not on the total force, but how that force is distributed over an area. It’s pressure; force divided by area. Increase the area, lower the pressure. And pressure is what breaks stuff. So that’s what a lot of baseplates do. They transfer the vertical forces of a column to the foundation over a larger area, reducing the pressure to a level that the concrete can withstand.

And that’s the first engineering decision when designing a baseplate. How big does it need to be? If you know the force in the column and the allowable pressure on the foundation, you can just divide them to get the minimum area of the plate. That’s the easy part. Because steel isn’t infinitely stiff. If I put this column on a sheet of paper, I think it’s clear that there’s no real load distribution happening here. The outside edges of the paper aren’t applying any of the column’s force into the table; I can just lift them. But this can be true for steel too. I filled up an acrylic box with layers of sand to make this clearer. If I use a thin base plate, the forces from my column don’t distribute evenly into the foundation. You can see that the baseplate flexes and the sand directly below the column displaces a lot more. I can try this with a thicker, more rigid baseplate, and the results are a lot different. Much more even distribution of pressure. So the second engineering decision when designing a baseplate is the stiffness of the plate, usually determined by the thickness of the steel, based on the loads you expect and how far the plate extends beyond the edges of the column. And in heavy-duty applications like steel bridge supports, vertical stiffeners can be included to make the connection even more rigid.

So far, though, the baseplate isn’t really much of a connection. That’s the thing about compressive loads: gravity holds them together automatically. There are no bolts in the Great Pyramid of Giza. The blocks just sit on top of each other. And that could be true for some columns too. The main load they see is axial, along their length, pressing the plate to the ground. But we know there are other loading conditions too. A perfect example is a sign. Billboards and highway signs are essentially gigantic wind sails. They don’t actually weigh all that much, so the compressive force on their base isn’t a lot, but the horizontal forces from the wind can be significantly higher than that. Those horizontal forces can increase the compression force on one side of the base plate, so you have to account for that in the design. But they also can result in shear and tension forces between the baseplate and foundation, so you’ve got to have something in place to resist those forces too. That’s where anchors come in.

There are a lot of ways to attach stuff to concrete. There are anchors that epoxy into holes, screw into place, or use wedges to expand into the hole. And of course, if you’re extra careful and precise, you can even embed anchor rods or bolts into the concrete while it’s still wet. There’s a huge variety of styles and materials that offer different advantages depending on your needs. Here’s just one manufacturer’s selection guide for the anchors and epoxies they provide. But like third year engineering students, all of those anchors can fail if they’re overloaded. And they can fail in a lot of different ways under tension or shear forces. The anchor rod itself can fracture or deform. It can lose its bond with the concrete and pull out. It can break out the surrounding concrete. Or if it’s too close to the edge, it can blow out the side. Calculating the strength of the anchor bolt and concrete connection against each of these failure modes is a lot more complicated than just dividing a force by a pressure to determine the baseplate area. So most engineers use software that can do the calculations automatically.

But, there’s another challenge about baseplates I haven’t mentioned yet, and it has to do with tolerances. Concrete foundations can be pretty precise. As long as you set the forms accurately and make them strong enough to avoid deflection while the concrete is being placed, you can feel confident in the dimensions of the structure that comes out of them. But there’s usually one surface that isn’t formed: the top. Instead, we use screeds and trowels and floats to put a nice finish on the top surface of a concrete slab or pier. But it’s rarely perfect enough to put a column directly on top. That’s not to say it can’t be done. I’ve seen concrete finishing crews do amazing work. But it’s usually not worth the effort to get a concrete surface perfectly level at the exact elevation needed for every column, especially when you have the time pressure of concrete setting up. And those tolerances matter. Just one degree off of level will put a 16-foot or 5-meter column out of plumb by more than 3 inches or 80 millimeters. Unless you’re in certain parts of Tuscany, that’s not gonna work. It’s more than enough to misalign some bolt holes. And that only magnifies for taller columns like signpoles. So, we usually need some adjustability between the plate and the concrete.

Sometimes that means shimming the baseplate to get it perfectly level. And the other primary option is to use leveling nuts underneath the plate. I welded up a custom-branded column and and baseplate that was laser-cut by my friends at Send-Cut-Send to show you how this works. These parts turned out so nice. By adjusting these nuts up or down, I can get the column to point in the exact direction required. And I can get it to the exact right elevation too. But maybe you see the problem here. All the work we did to make sure the baseplate distributes the vertical load even across its area is lost. Now the vertical loads are just being transferred through some shims or through the bolts directly into the anchors. So, in a lot of cases, we add grout between the plate and the concrete to bridge the gap. Grout is basically concrete without the large aggregate, mixed with a low viscosity so it flows more easily into gaps. And it often includes additives to prevent it from shrinking as it cures, making sure it doesn’t pull away from the surfaces above and below. When it hardens, the grout can transfer and distribute the loads into the foundation. So if you pay attention to baseplates you see out in the built environment, you’ll notice it’s pretty common that they sit on a little pedestal of grout and not directly on the concrete below. But even this comes with a few problems.

First is load transfer. Even with the grout, some of the vertical loads are still going into the anchor bolts that might not have been designed for compression. So now we’ve added a few more new potential failure modes to the laundry list: punching through the bottom of a slab, and buckling of the rod itself. Sometimes contractors will use plastic leveling nuts that can hold the column during construction, but will yield after the column’s loaded so the grout supports all the weight. Second is fatigue. Especially for outdoor structures that see wind and vibrations, the grout under the baseplate might not hold up to repeated cycles of loading. Third is moisture. Grout can trap water, leading to problems with corrosion, especially for hollow columns like sign poles where condensation needs a way out. And the grout can hide that corrosion, making it difficult to inspect the structure. And fourth, adding grout below a baseplate is just an extra step. It’s kind of fiddly work to do it right, and it costs time and resources that might otherwise be spent somewhere else. In fact, there are a lot of cases where it’s an extra step worth skipping.

You can design anchor bolts strong enough to withstand all the forces a column will apply, including the compressive forces downward. And you can design a base plate stiff enough that those forces don’t have to be distributed evenly across the entire area. And if you do, you have a standoff base plate. It just floats above the concrete with only the anchors in between. It looks like a counterintuitive design. We think of a baseplate as kind of a shoe, so it should be sitting on the ground. And a lot of them are designed that way. But for other structures, a baseplate is really just a way to connect a foundation to a column through an anchor. So if you pay attention, you’ll see these standoff baseplates everywhere. A lot of state highway departments have moved away from using grout to make signs and light poles easier to inspect. And they often install wire mesh to keep animals out from hollow masts.

Clearly there’s a lot more to baseplates than meets the eye, and that means there’s also a few myths going around grout there. A common misconception is that standoff baseplates are meant to break away in the event of a collision. And I totally understand why. If an errant vehicle hits a signpost, a relatively minor deviation from the road can turn into a deadly crash. Smaller signs installed near roadways often do use breakaway hardware or features. You’ll often see holes drilled in wooden posts, bolts with narrow necks meant to snap easily, or slip bases like this one to make sure a sign gives way. But for larger structures like overhead signs and light poles, that’s generally not the case. Having one of these break away and fall across a highway could create an even bigger danger than having it stay upright. So, even though they might look similar, standoff baseplates are distinct from sign mounts designed to break loose in a collision. Instead, larger structures installed in the clear zones of highways are protected from crashes using a guardrail, barrier, or cushion.

Baseplates are like bass parts in music, it’s easy to overlook them at first, but once you notice them, you can’t stop paying attention to how important a role they play. And just like bass lines, they might seem simple at first, but the deeper you dig, the more you realize how complex they really are.

Which Power Plant Does My Electricity Come From?

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

In June of 2000, the power shut off across much of the San Francisco Bay area. There simply wasn’t enough electricity to meet demands, so more than a million customers were disconnected in California's largest load shed event since World War II. It was just one of the many rolling blackouts that hit the state in the early 2000s. Known as the Western Energy Crisis, the shortages resulted in blackouts, soaring electricity prices, and ultimately around 40 billion dollars in economic losses. But this time, the major cause of the issues had nothing to do with engineering. There were some outages and a lack of capacity from hydroelectric plants due to drought, but the primary cause of the disaster was economic. Power brokers (mainly Enron) were manipulating the newly de-regulated market for bulk electricity, forcing prices to skyrocket. Utilities were having to buy electricity at crazy prices, but there was a cap on how much they could charge their customers for the power. One utility, PG&E, lost so much money, it had to file for bankruptcy. And Southern California Edison almost met the same fate.

Most of us pay an electric bill every month. It’s usually full of cryptic line items that have no meaning to us. The grid is not only mechanically and electrically complicated; it’s financially complicated, too. We don’t really participate in all that complexity - we just pay our bill at the end of every month. But it does affect us in big ways, so I think it’s important at least to understand the basics, especially because, if you’re like me, it’s really interesting stuff. I’m an engineer, I’m not an economist or finance expert. But, at least in the US, if you really want to understand how the power grid works, you can’t just focus on the volts and watts. You have to look at the dollars too. I’m Grady, and this is Practical Engineering.

Electricity is not like any normal commodity we buy and sell. You can’t really go to the store and pick up a case of kilowatt-hours. It can’t really be stored or stockpiled on an industrial scale, which means it has to be created at essentially the exact instant it's needed. And the demand is fairly inelastic. We want our lights, stoves, air conditioners, and devices to turn on no matter the time of day. That requires the supply side to handle incredible volatility, ramping up or down to meet demands in real-time. And the whole business is incredibly capital-intensive: you need very expensive infrastructure for pretty much every step of the process. The only reason it can work is that we all share that infrastructure, spreading out the costs. Call me a nerd, but I think all of this creates some fascinating challenges, both on the technical side for engineers and the organizational side for the policymakers, regulators, and all the companies that participate in the electric power industry.

It wasn’t that long ago that the electric utilities did it all. As the pros say, they were “vertically integrated.” Each utility owned and controlled the three major pieces of the grid within their service areas: generation (or power plants), transmission lines (which carry electricity at high voltage across long distances), and a distribution system (which delivers electricity to most customers at lower voltages). That meant they had a monopoly. Customers couldn’t choose where their power came from or how they got it. And that meant that electric utilities had to be carefully regulated to make sure that, without any competition, they were still offering customers a reasonable price for power.

Over time, utilities realized the value of interconnecting so they could help each other in times of need. Electricity is a true commodity, even if it has some unusual properties. For the job it does, it mostly doesn’t really matter who made it - a kilowatt is a kilowatt, no matter where it came from. If one utility’s power plant went down or bad weather hit, they could work out a deal to share power with a neighbor and keep demand satisfied. As the practice grew more common, power pools developed where multiple utilities would interconnect and agree to share power. Every system is different; subject to different risks, different weather conditions, and outages at different times. It just made economic sense to spread out that variability and risk. Eventually, huge parts of North America were interconnected by transmission lines, creating the “grids” we know today. The major interconnections in the US and Canada are the Western, Eastern, Quebec, and Texas.

Historically, the wholesale price one utility would pay another for power was regulated just like the rates utilities charged their retail customers. It was usually based on the actual cost of generating that power, so the big utilities couldn’t price gouge smaller companies. But a lot of that changed in the 1990s when the federal government opened the door for deregulation. The idea is simple on the surface: if power can move fairly freely on the grid, there’s no need for major utilities to be the only ones producing it, and there’s no need to regulate the prices for which it’s bought and sold. Let’s let market forces drive the decisions. It will increase competition and efficiency, driving down prices, and the investment risks will fall to the investors, not the customers.

Quite a few states took the opportunity to deregulate the production of power, and quite a few didn’t. In fact, right now, it’s roughly half and half, but there’s a lot of variety between states when it comes to who produces power and how it’s bought and sold between utilities and other companies, and even big differences within individual states. In truth, the process of deregulation has been anything but simple, and actually created a whole new set of interesting challenges. Companies trying to game the system, like what happened in California, is only part of it. In fact, power professionals often say that certain states aren’t deregulated; they’re just differently regulated. But how does all this really work in practice? Let’s set up an analogy.

Say I live on one side of a big lake. The water isn’t mine, but there is a water company on the other side. If I want to buy some water, they could load it on a truck and haul it to me. Or they could just put it in the lake, and I could take out the same amount. It’s probably not the same water, but it doesn’t really matter. In this analogy, water is water. Let’s scale it up. Now, hundreds of people need water, and hundreds of companies are selling it. Each person can hire any company they want to provide their water. The distance between buyer and seller doesn’t really matter. Every seller puts the amount of water they’ve sold in the lake, and each buyer takes as much out as they’ve bought. As long as everyone keeps track of how much they buy and sell, the lake stays full, and basic laws of physics will sort out how the actual water flows. In one way, you know exactly where you get your water: the company you paid to provide it. But in another way, you have no idea. All the water from all the companies is comingled in the lake. This is what happens on the grid. In a way, the power coming to your house comes from the power plant or plants that your utility paid to create it. But the electrons themselves probably didn’t. Just like the water in the lake, electricity flows according to the laws of physics from high potential to low, sloshing and flowing according to what everyone on the system is doing.

This is a lot like how a deregulated grid works. Utilities that supply electricity to their customers don’t generate the power themselves. They enter contracts to get it from wholesale power providers, separate companies who only generate electricity. But you can see a challenge here. If I’m on my big lake wanting some water, I may not want to coordinate with every water company to see who’s got the best price, especially if my need for water varies day by day or even second by second, and honestly, I’m not even sure exactly how thirsty I’m going to be. And if I’m a water company on the lake, it’s a lot of overhead work to deal with all these customers and their different needs. It makes a lot more sense if there’s a marketplace. So it is on the grid.

Like I mentioned, this varies quite a bit depending on where you are, so I’ll try to be as general as possible. Outside of those direct contracts between one buyer and one seller, most wholesale electricity in deregulated areas is bought and sold on the day-ahead market. Wholesale purchasers like utilities submit bids with estimates of how much electricity they’ll need for each hour of the next day. And generators submit their offers to sell a specific amount of electricity for a given price that’s based on their production costs, availability of fuel, and operational constraints. The facilitator of each wholesale market takes all the bids for every hour and matches the supply and demand to get the right amount of energy on the grid at the right times for the lowest cost. Here’s a basic example of a single hour of the auction:

Let’s say four generators submit bids to provide electricity during this hour: A nuclear plant bids 1200 MW for a price of $20/MWh. A natural gas peaker plant bids 400 MW for $100/MWh. A coal plant bids 500 MW for $30/MWh. And a wind farm bids 400 MW for $0. Wind and solar can submit very low bids because they have no fuel costs. There’s pretty much no way for them to lose money if they’re connected to the grid, especially because many get outside incentives for every megawatt-hour they generate. They even submit negative bids in some cases, meaning they’re willing to pay money to stay connected to the grid. In any case, electricity generators in our hypothetical hour have offered 2,500 megawatts of power to the market.

Let’s say purchasers submitted 2,000 megawatts of demand for this hour. We arrange the generation bids in order of least cost to satisfy demand. This concept is known as economic dispatch. We buy power at the lowest cost possible. We’re going to dispatch the wind farm and nuclear plant, dispatch the coal plant at 80% of the capacity they bid, and we don’t need the peaker plant at all. The clearing price is the cost of the last unit of supply to be dispatched. In this case, it’s $30 per megawatt-hour. Every producer gets paid that price for the power they put on the grid for that hour, even if they bid lower, and every buyer pays that price for wholesale electricity. This is why wind and solar are incentivized to bid 0 dollars. They essentially guarantee that they’ll make the cut.

It seems like a simple process in our hypothetical hour, but in reality there’s a lot more to it. For one, many types of power plants can’t just be toggled on and off with the flip of a switch. They need significant lead time to start up and shut down. They have minimum and maximum output levels. And their costs can vary a lot depending on how long they run. So the market has to take those factors into consideration. Also, we can’t perfectly predict the future, even for the next day. There are always going to be differences in the day-ahead forecasts. Demand varies, equipment has problems, and other unforeseen events like sabotage or solar storms happen all the time.

So another market runs in real-time, sometimes with auctions every 5 minutes, to make up those differences and keep the supply in check with demand. For example, if a wind farm overproduces what they bid into the day-ahead market, they can sell the extra on the real-time market. And if they underproduce, they may need to buy power in the real-time market to make up for the shortfall. And if things really get tight with not enough reserves, the real-time markets usually include a way to boost prices upward, even beyond what the clearing price would be, to make sure they’re more closely reflecting the actual value of electricity. That includes the cost to society if people lose power, or put another way, the cost they would be willing to pay to avoid a disruption in electrical service. This concept is called the value of lost load, and it’s something that the generators usually aren’t taking into account in their bids.

But that’s not all the markets. Many areas have a capacity market intended to make sure there are enough generators available to meet demands over the long term. These auctions happen only once a year or so, and generators bid to create new capacity within three years. All the generators that win in the auction are rewarded for adding capacity to the grid, no matter how much of that capacity actually gets used in the future. This doesn’t happen everywhere though. Texas doesn’t use a capacity market and instead relies on prices in the day-ahead and real-time markets to encourage generating companies to make long-term investments in capacity

Many areas also have markets for so-called ancillary services, basically services needed to keep the grid stable and reliable. There are auctions for regulation, which accounts for very short-term fluctuations in supply and demand to keep the frequency stable. There are also auctions for reserves that can keep plants ready to get on the grid quickly if another resource trips offline. Other services to keep the grid stable are often contracted directly instead of using auctions. Reliability-must-run contracts pay for power plants that are on the verge of retirement to stay in service until the capacity is replaced. Inertia services pay to keep a certain amount of rotating mass connected to the system. I have a video on that topic if you want to learn more. Black start contracts pay for some generators to have the ability to go from a total shutdown to operational without assistance from the grid. I also have a video on that topic. And reactive power contracts help maintain the stability of the voltage on the grid. And, I have a video on that one too.

A potentially surprising thing about many of these markets is that it doesn’t just have to be generation resources bidding into them. The overall goal is just to get the supply to meet demand, and there are two ways to do that. You can increase the supply or decrease the demand. I said earlier that electricity demand is fairly inelastic, but there are a lot of situations where customers can reduce demand, especially if they’re compensated for doing it. Large industrial power users like refineries can shift schedules around or even turn on their own generators if resources on the grid are getting scarce. This is how you get wacky news stories about cryptocurrency miners making more money participating in electricity markets than in Bitcoin. There are even companies that will gather up a bunch of smaller power users who have some flexibility in their demands, package them up, and sell that demand reduction as a service in the wholesale electricity market. And some utilities coordinate similar demand response programs with their customers, offering credits on your bill if you have a smart thermostat. Deregulation of wholesale electricity markets just opens up this world of possibilities in how we manage the grid.

But there is one big way my lake analogy from earlier breaks down. Because that lake symbolizes the transmission and distribution lines that carry power between the buyers and sellers. And in reality, they’re not really like a lake, but more like a series of interconnected canals. And they didn’t just appear. Someone has to build them and maintain them, often at great cost, so those costs need to be covered by the rates we pay for electricity on top of the generation. In this case, there’s really no way to deregulate those costs. It doesn’t make sense to build parallel, competing networks of transmission and distribution lines. It would cost too much, and we’d just have too many wires across the landscape. So regulators oversee the rates that transmission and distribution companies charge utilities to use their wires to move power between users and generators. And of course, there’s a whole host of complex financial systems in place to make this happen. Wholesale purchasers not only have to buy power they need and the power that will be lost along the way, but also reserve capacity on the transmission system for that power to travel, and pay the transmission and distribution system operators for the privilege.

Confusingly, the flow of power isn’t really controlled on a line-by-line basis or sometimes even on a system-by-system basis. Power flows where it flows once it’s released on the grid, and there’s no simple way to keep track of who made it or who bought it at individual points on the network. Transmission reservations and tariffs are the law of the land, but the actual electrical power follows the laws of physics. So unlike at your house where you pay one-to-one for the actual power that flows through your meter, payments to transmission operators aren’t always a perfect reflection of how each buyer’s power moves through their system. Still, it’s the best mechanism we have to ensure electricity moves reliably across the grid and that the owners of the transmission assets are fairly compensated.

The other thing is that those canals don’t have infinite capacity. They can only move so much water, just like the transmission system can only move so much power. So in managing the wholesale electricity market, you don’t only have to consider what’s the next cheapest source of power, but also whether you can actually get that power to where it needs to go. Grid operators have to account for congestion, like rush hour for electrons. They usually do this by allowing prices to vary from place to place, an idea called Locational Marginal Pricing. You can see on this map of Texas how significantly prices can vary across the state, reflecting a difference in where the demand is versus where the generators are and the congestion on the transmission system that results. And hopefully at this point you’re seeing how complicated all this really is. Grid operators have to take into consideration all these details - power flows, weather, limitations of every kind of generator, second-by-second changes in the system - in order to match supply with demand at the lowest cost possible. And it gets even more complicated when you add distributed generation sources, like home solar installations, that put energy on the grid from the other side of the meter.

And this is only on the wholesale side of the grid. Even though most of those dollars moving around came out of our pockets, the end-users of the electricity, you and I really don’t participate in this segment of the grid. For many of us, the company we pay for electricity (the retail provider) didn’t generate that electricity, and in many cases, doesn’t own the infrastructure that it traveled along to reach our house or place of work. And for around a quarter of the US, the retail market is deregulated to the point where you can choose which company you buy your power from. So what do they actually do?

In essence, retail providers just buy power on the wholesale market and sell it to you. They’re middlemen, the car dealerships of electricity. They navigate all that complexity we just discussed so you don’t have to. Retail providers all provide essentially the same thing, but they can differentiate themselves by offering different kinds of rates that suit their customers better. One provider in Texas, Griddy Energy, famously offered their customers the real-time wholesale price, exposing them to the incredible volatility of the market. Unsurprisingly, Griddy filed for bankruptcy after the winter storm in Texas when their customers couldn’t pay the exorbitant bills. The other thing retail providers can do is connect your dollars to specific sources of generation like renewables. Instead of buying power in the auction, where you have no control over the sources, they contract directly with wind, solar, and other generators to purchase it directly on your behalf.

So next time you get your power bill, take a look at those line items. Maybe there’s a base rate set by your provider that covers all the various costs of operating the grid from generation to transmission to distribution. Or maybe they’re broken out according to all the various costs that it actually takes to run the bulk power system. Do you pay a separate rate for the distribution service? Does your bill have an adjustment for the variability in the wholesale market? Is there a charge for the Public Utility Commission or whatever agency oversees this whole financial web of complexity? Every bill looks a little different, but I hope this video clears up some misconceptions and encourages you to think about what the price you pay for electricity actually accomplishes on the grid.

Why Are Cooling Towers Shaped Like That?

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

This is not smoke. And this isn’t a smoke stack (at least not the kind we normally think of). It serves a totally different purpose at a power plant than smoke stacks whose job is moving combustion products high into the air, allowing them to disperse away from populated spaces. Maybe you already knew that, or at least suspected it. After all, you saw the title of the video. Plus, this kind of tower is commonly associated with nuclear plants that don’t combust anything at all to create the heat that drives their generators. But that heat is the key. The largest class of power plants, called thermal power stations, use steam turbines (or tur-bines, depending on how you say it). But once that steam makes it through the turbine, it needs to be condensed back into liquid water. It’s kind of frustrating. You spend all those resources heating the water up, and then you spend even more resources to cool it back down. And a power plant isn’t much good if all the electricity it generates gets used up just trying to cool that steam back down. So, engineers have come up with some pretty creative ways to cool huge amounts of water, like millions of gallons or tens of thousands of cubic meters per hour, and do it relatively efficiently. Not all cooling towers look like this, but there are some really clever reasons for that iconic shape we all recognize, and I’m going to build one in the garage to show you how they work. I’m Grady, and this is Practical Engineering.

Power plants could just vent steam into the atmosphere, but generally, they don’t do that. For one, it wouldn’t be good for the environment. The heat, moisture, and noise would affect wildlife and the weather. For two, it would waste a lot of water. The feedwater for a boiler is often carefully treated to avoid corrosion and mineral buildup in the machinery. It’s expensive water, so it doesn’t make sense to set it free. And for three, it would waste a lot of energy. It’s generally less expensive to cool the steam down just enough to condense it back into liquid water so it can be reused as feedwater. But even that is an enormous challenge.

I talked a little bit about how power plants actually consume a lot of energy in a previous video. It’s a net positive, of course. But any energy you spend on all the industrial processes required to produce electricity at scale is energy not being sent out to the grid, and that includes cooling. So engineers want to do it efficiently. One simple option is to use a cooler stream of water that already exists, like a river, lake, or sea. And in fact, there are a lot of power plants near me that do exactly that. This plant draws water from the lake on the south side, sends it through condensers, and then releases it into a channel where it flows to the north side of the lake. The slow circulation gives that water time to cool down before it reaches the plant again. But it’s not feasible to have a lake or river for cooling water at every thermal power plant, and there are environmental impacts with the heat and intakes that require careful consideration. So instead, lots of power plants use cooling towers.

You might be familiar with the various machines humans have devised for cooling stuff down. You might even be enjoying the comfort of such a device right this very minute. But, like I mentioned, the simplest way to cool something is to simply let natural physical processes do the work, just wait for entropy to do its thing. After all, the temperature is usually less than boiling outside, so the heat from steam will naturally transfer to the ambient air if you let it. So that’s what many cooling towers do… kind of.

I designed a cooling tower in my garage so I can show you exactly how this works. This is made from laser-cut strips of acrylic with a carefully selected shape. And when I carefully tape these carefully sized strips together, I get a nice (somewhat transparent) cooling tower. This is a model of a natural draft tower. It’s not the most common type out there, but it is one of the simplest, and also the most iconic and recognizable, so it’s perfect for this demonstration. You may have noticed the holes I drilled in the bottom of each acrylic strip. This tower needs a way for air to get inside at the bottom. If you look closely at the real thing, you’ll see something similar. They aren’t continuous all the way down but actually open around the bottom.

Steam from the turbines doesn’t go to the cooling tower directly. Instead, there’s a separate stream of water, aptly called the cooling water. The steam is condensed into liquid water in a condenser that is cooled by cooling water, which then flows between the condenser and the tower. I’m simulating that here with a bucket of hot water and a beer brewing pump. That hot water gets pumped to sprayers inside the tower. If you know a little bit about thermodynamics, you know that we can only get this water as cool as the ambient air temperature. Heat naturally flows from hot to cold, so you can’t get any more heat transfer once the water reaches the outside temperature. But if you know a little more about thermodynamics, you know there’s a trick that can improve the performance of a system like this. And this layer of material below the sprayers is the key.

This is called fill. I’m just using a dehumidifier pad, but in an actual cooling tower, the fill is usually a layer of plastic, carefully designed to maximize the surface area of the water in the system. It does this by forcing the water to either splash into tiny drops or form thin sheets as it falls downward. The goal is to expose as much surface area of the hot water as possible to the air flowing through the tower. Water drips down. Air flows up. The pros call this counter-flow. And it’s the trick to this whole process. (Actually you can use cross-flow as well, but let’s jump over that rabbit hole for now.)

You might think that the outside air has just one temperature, but to cooling professionals, it has two. One we call the dry bulb temperature is what you normally encounter. That’s what shows up in the weather report. It’s what’s on the thermometer. But air also has a wet bulb temperature. If you soak the end of a thermometer and pass it through the air, that water will evaporate. The drier the air, the more easily water evaporates. This is why it feels so much hotter when it is also humid outside. It takes energy, called latent heat, to convert water from a liquid to a gas, and that energy is absorbed from the liquid water, cooling it down. So, as long as the ambient air isn’t already saturated (100% relative humidity), you can actually cool water below the dry bulb temperature using evaporation. And the lower the humidity of the air, the more evaporation can take place, so the bigger the difference in wet and dry bulb temperatures.

This isn’t anything revolutionary. Nature figured out evaporative cooling millions of years ago. It’s why we sweat when it’s hot. But using it at this scale is really impressive. And it’s not the only innovation in a natural draft cooling tower. For that, I need to show you a cool graph. This is a psychrometric chart. It looks pretty intimidating. You could spend an entire college course learning about this stuff, and there are probably a few HVAC professionals groaning at the screen right now. But I just want to use it to explain a few important things about the physical and thermal properties of air. First, as the temperature of air goes up, its capacity to hold water goes up too. Kind of like hot tea can dissolve more sugar than cold tea, hot air can hold more water than cold air. Next, as air temperature goes up, its density goes down. Confusingly, the psychrometric chart actually shows specific volume, which is the inverse of air density. So as you move up in temperature, these lines slope downward. Hot air rises. Most of us know that. But maybe less intuitively, it’s also true for humidity. If you hold the temperature constant, and just increase the amount of water in the air, its density goes down. Water molecules actually weigh less than the nitrogen or oxygen molecules in air. So, humid air is more buoyant than dry air. And this is the second key to a cooling tower: convection.

The hot water transfers its heat to the air. The warm air becomes buoyant, flowing upward in the tower and drawing fresh air in through the intakes. But some of that hot water is also evaporated, removing more heat from the water, and making the air even more buoyant. The process both cools the water down and creates a natural draft up through the tower, drawing in even more fresh, drier air as it does. Ignoring the pumps and minor control features, there are no moving parts. So for just the cost of spraying the water, you create this enormous natural convection in the tower, moving huge volumes of air into the bottom, up past the fill, and out at the top to reject large amounts of heat to the atmosphere. This is the “smoke” you sometimes see rising from a cooling tower. It’s not actual smoke; it’s just water vapor condensing into tiny droplets as the now-saturated air mixes with the cooler outside air at the top of the stack. It’s basically a cloud machine. That’s why the plume is usually more visible during the winter months.

I was really surprised at how well my little model tower worked. I was pumping water at about 120 F (50 C) and the water coming out was dropping by around 30 degrees F (17 degrees C). That’s like a perfect cup of coffee down to a lukewarm shower. The air coming out at the top was shockingly warm, and there was a lot of it. I was really surprised at how much airflow this thing could create just by spraying some hot water inside. I guess I just figured these processes wouldn’t scale well because of turbulence, but I was wrong. It was both pretty good at cooling the water down and looking cool on camera. And part of the reason this looks so cool, the shape of the tower itself, is also crucial to its function.

Natural draft cooling towers often feature this curved, swooping shape. The mathematicians call it a hyperboloid. You can actually make one yourself pretty easily. Put some sticks evenly spaced and connected in a circle around the top and bottom. Then twist. Actually the fact that it can be made from straight lines makes these easier to construct. But that’s not the only reason they’re built this way. After all, a cylinder has straight lines too. There are some aerodynamic benefits to using a hyperboloid as a chimney. The wide base provides more area for air to flow in at the bottom. The constricted center accelerates the flow upward. And the wider top helps promote mixing of the hot humid air with the cooler air outside. But, really, these are secondary benefits. The main reason for the shape is structural.

These towers are big. To get enough natural convection, you need a tall stack. The taller the tower, the more warm, humid air is contained inside, generating more buoyancy and more airflow. The largest natural draft towers are more than 650 feet or 200 meters tall, and more than 400 feet or 120 meters in diameter. And you want the walls to be as thin as possible. Less material means less cost and more area for airflow. But a really tall cylinder made of thin walls is not very strong. It’s basically a big empty coke can. But the double curvature of a hyperboloid stiffens the shell against vertical loads like the structure’s own weight and horizontal loads like wind. You can also try this yourself. A thin piece of paper has almost no stiffness. As soon as you put a curve in it, it’s much harder to bend perpendicular to the curve. And two curves are better than one. It’s the Pringle factor. My model shows this pretty well too. I started out with thin, floppy strips of acrylic. But even just taped together, this tower is really strong. Using a hyperboloid can cut the structural stresses in half compared to a cylindrical tower, making structures like this much more economical to build. So that’s why natural draft towers use that shape. But that definitely doesn’t mean this is the only kind of cooling tower.

In fact, hyperboloid natural draft towers are actually pretty rare in the same way that thermal power plants are pretty rare compared to large office buildings, hospitals, and schools that also often use cooling towers as part of the HVAC system. Those towers often use mechanical draft systems, basically using fans to create airflow instead of tall stacks. I talk a little bit more about this in my book. We still call them cooling towers, even though they usually aren’t too towery. And, in fact, lots of power plants, refineries, and chemical plants use mechanical draft cooling towers as well. They’re less dependent on ambient conditions to create the necessary airflow, they’re smaller, usually less expensive to build, and offer some flexibility if heat loads fluctuate.

And not all cooling towers use evaporative methods. Dry cooling towers just use heat exchangers inside, with the cooling water flowing in a closed loop. In dry systems, you’re limited by the higher dry bulb temperature instead of wet bulb, but you don’t lose any water to evaporation, and you don’t have to deal with the buildup of minerals that happens in wet systems as cooling water evaporates away.

The reason you see big natural draft towers at power plants has everything to do with scale. The long-term savings of not having to run big fans and maintain all the associated equipment outweigh the higher initial costs. Particularly at nuclear plants built with design lives of 50 years or more, you can amortize the cost over a longer duration. Also these facilities are usually already built in more remote locations where land is cheaper and height restrictions are less stringent, making it feasible to build such massive structures just for cooling. And they’re particularly common at nuclear plants for two reasons. Number one is reliability. Cooling is an essential part of safety at a nuclear plant. The fewer parts of a cooling system, like fans, that can go wrong in an emergency, the better. Number two is variability, or the lack of it. Nuclear facilities are usually baseload plants. Most of them run nearly nonstop at a constant output. So they can get away with a system that’s designed for a single heat load rather than mechanical cooling required to ramp up and down. But, even if the heat load doesn’t change at large baseload plants, the weather does, and not every climate is ideal for natural draft towers.

If you live in a dry place, you might be familiar with evaporative appliances that can cool and humidify the air. We called them swamp boxes when I was growing up. It makes sense that these work better in dry climates; there’s less moisture in the ambient air, so you get more evaporation, and thus more cooling potential. So, you might assume that natural draft towers work best in areas with low relative humidity, but that’s not necessarily the case. And this took me a little bit to wrap my head around. Let’s look back at that psychrometric chart. Say we’re in an area with a wet bulb temperature of 20 celsius, 70 fahrenheit. The water from our condenser comes in at 40 C, 100 F, so the air leaving the tower will be saturated at that temperature. And we’re trying to cool that water down to 30 C, 85 F.

If the ambient relative humidity is say, 20 percent, our air is starting here and going here. But it doesn’t go in a straight line. Since the air is coming in from the bottom, it’s not coming into contact with the warm water, but the coldest water first. So it actually heads toward the outlet temperature and gradually veers toward the water inlet temperature as it rises through the fill. If you look at the lines for specific volume you might see the problem. In the first part of the curve, the state of the air is moving parallel to the lines. In other words, it’s not gaining any buoyancy. It’s not going to rise up the stack. It might work right at startup, but as the water cools down, the airflow in the tower will slow down and stall, and you won’t be able to cool the water enough.

But watch what happens if you increase the relative humidity of the ambient air to 50 percent. The line still curves initially toward the outlet temperature before heading to the inlet temperature as it moves through the fill, but it decreases in density consistently along its entire path through the fill. So, cooling engineers say that, for a given wet bulb temperature, you get a better draft as relative humidity goes up. It seems counterintuitive, but another way to look at it makes more sense. Natural draft cooling towers just don’t work that well in hot climates. Even if the air is dry enough to evaporate a lot of water and create a lot of cooling, you just can’t get it to rise up a tower on its own. So if you pay attention, you’ll notice different types of cooling depending on where you are. There are two nuclear plants in Texas and both use reservoirs for cooling. That gives you a sense of the cost involved in cooling feedwater at a power plant. In both cases, it was cheaper to build and maintain a dam and huge lake than a cooling tower that would work well in our climate.

I know that’s a little in the weeds, but I think it’s so fascinating how much engineering goes into things like this, and I’ve just barely scratched the surface here. The economics of building large facilities like thermal power stations requires that we know for sure that each design is going to work before any construction starts, and that has driven a huge variety of types and styles of cooling towers. Engineers mix and match designs and styles according to what will work most efficiently for each application, so there’s practically no end to the designs you can spot if you keep an eye out. And actually, some newer cooling towers do put flue gas into the air stream, making the tower do double duty. So I kind of lied at the beginning of the video. Depending on the tower you’re looking at, there really might be some smoke in that plume coming from the top. But mostly, it’s just water. And in a world full of straight lines and right angles, I love that every once in a while, it just makes good engineering sense to use curvy shapes to accomplish a really important job.

The Wild Story of the Taum Sauk Dam Failure

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

Early in the morning of December 14, 2005, pumps were nearly finished filling the upper reservoir at the Taum Sauk power station, marking the end of the daily cycle. Water rose to the top of the rockfill embankment, reaching the concrete parapet wall that ran along the top of the dam. But the water didn’t stop. One of the two pumps shut off, but the other kept running, and soon, the water was lapping over the wall. Within minutes, those splashes turned into a steady stream cascading over the parapet, pouring against the embankment on the other side. The rockfill eroded slowly at first, but the hole grew deeper and wider. The pump finally shut off, but it was too late—the footing of the parapet wall had already been undermined. The wall tipped over, and a massive surge of water was unleashed down the mountainside headed directly toward a state park.

This award-winning pumped storage facility, considered a model of modern engineering, immediately became the center of intense scrutiny. And what the investigations found would change a lot about the field of dam safety. I’m Grady, and this is Practical Engineering.

When it was built in the 1960s, the Taum Sauk pumped storage plant was unlike really any other power plant in the world, at least in terms of size. South of St. Louis in the Ozark Mountains, it was designed to meet a very specific need. I’ve talked about pumped storage on the channel before, and Taum Sauk was one of the largest facilities of its time. Built by Union Electric, which eventually merged with Ameren, the whole plant is basically a battery. It’s actually a net consumer of electricity, which is normally not a good thing for a power plant. But managing the power grid isn’t only about how much electricity you can produce, but also when you can produce it. Large coal plants in the Missouri area could make lots of power, but they couldn’t ramp that production up and down to accommodate fluctuating demands throughout the day. So, Union Electric proposed a clever solution, one that’s pretty common today, but was innovative for its time. Two reservoirs were constructed: one low on the east fork of the Black River and another near the top of Proffitt Mountain, Missouri’s sixth-highest peak. Between them, a hydroelectric plant with two reversible turbines.

When electrical demands are low, rather than reducing the output of thermal power plants, that energy can go toward pumping water from the lower to the upper reservoir, usually overnight. Then when demand spikes during the day, all that stored potential energy created from cheap electricity can be harvested and put back on the grid by reversing the system to generate hydropower. Of course, you don’t get all the power out that you put in. Some of that water evaporates or leaks out, and there are losses of energy in the pumping and generation. But, with an overall efficiency of around 70%, it was more than enough to justify the enormous cost of building and operating two reservoirs and a power plant that doesn’t produce any of its own electricity.

The most striking part of the whole facility is the upper reservoir. It’s just such an unusual sight: a circular dam, sometimes called a ring dike or ring levee, perched on top of a mountain. This is not usually an efficient way to build a dam. We typically construct them across valleys so that the natural topography can form the sides and back of the reservoir. With a so-called “off-channel reservoir” you have to build the dam all the way around, increasing the costs and the engineering complexity. But there are no valleys at the tops of mountains, and that height is an essential part of a pumped storage facility. The power available from falling water is really simple to calculate: multiply gravitational acceleration, the density of the fluid, the volumetric flow rate, and the difference in height, called head. We can really only change two of these. So for a specific power output needed for a specific duration, you can trade height for flow. The greater the difference in height between the two reservoirs in a pumped storage facility, the less water you need to move, which reduces the size of all the infrastructure, and thus saves costs. The mountains in southeast Missouri provided a perfect location for the project, creating about 750 feet or 230 meters of height between the upper and lower reservoirs.

Actually the whole facility is named after the highest mountain in Missouri, Taum Sauk, which was the original site for the upper reservoir until there was too much pushback about building a project there, so they moved it to a slightly lower peak nearby. And they encountered some challenging conditions during construction, forcing the engineers to realign the dam to avoid an area of weak geology, giving it that unique kidney bean shape. The original dam was built as a rockfill embankment - basically just dumping a long pile of rocks around the perimeter of the reservoir. Rockfill usually works well as an embankment if you have a good source of material nearby. It’s really strong, doesn’t require a lot of compaction, and it doesn’t settle much over time like soil fills do. One thing rockfill doesn’t do well is hold back water. Too many spaces between the rocks. So concrete panels were installed all along the inside of the reservoir to make the embankment water-tight. A tunnel connected a morning glory inlet through the mountain to the generating plant. The inlet was set into a basin 20 feet or 6 meters below the bottom of the reservoir to suppress the potential for a vortex to form as it was drained each day. The whole project was designed to be operated remotely with no on-site technicians required, another innovation for the time, but one of the many decisions that would prove disastrous.

For most of its life, the Taum Sauk station operated on average around 100 days per year, usually during the hot summer months when electricity demands were more variable between night and day. Deregulation of electric power markets in the 1990s opened up the possibility of selling power to other utilities. Those 100 days per year went up to 300, meaning the upper reservoir cycled up and down, often twice per day, nearly every day of the year. And that was starting to cause some problems. The upper reservoir had dealt with leaks essentially since it started operating in the 1960s. Several projects were implemented throughout its life to deal with the issue, but the increased cycles of filling and draining were only making things worse. At one point, small ponds were built beside the reservoir to capture some of the leakage and pump it back inside. In the fall of 2004, Ameren decided to bring out the big guns and spent more than two million dollars to install a geomembrane liner to cover the entire reservoir. That essentially fixed the problem, but it caused a few new ones too.

About a year later, in September 2005, the Institute of Electrical and Electronics Engineers, or “I-triple-E” declared the plant an “Engineering Milestone” for its innovations in the world of electrical infrastructure. On the day before the ceremony, some of the participants took a tour of the upper reservoir and witnessed water pouring over the parapet wall on one side of the dam. The operators quickly switched from pumping mode to generation to get the water back down. They chalked up the issue to high winds from a remnant tropical storm that caused the overtopping, but just to be safe, they hired a dive inspection team to check on the level sensors. And what they saw was concerning.

When that geomembrane liner was installed in the reservoir, there was a valid concern that any penetrations might cause leaks in the future. But the reservoir needed level sensors installed for the control system to be operated remotely. So, instead of mounting those sensors directly to the concrete through the liner along their length, the engineers tried something different. Two cables were run between anchors at the top and bottom of the embankment slope. The conduits for the sensors were attached to those cables, minimizing the number of penetrations needed. Unfortunately, the mounting system was underdesigned. Those conduits were buoyant, and also subject to strong currents as the reservoir filled and emptied each day. Sometime after the spring of 2004, they had become dislodged and deflected, so the sensors inside were providing readings that were lower than the actual water level.

Based on those findings, the operators decided to reprogram the control system to subtract two feet from the upper set point on the pumps. The original design called for two feet or 600 millimeters of freeboard between the top of the wall and the maximum water surface. They figured that doubling that distance would be enough to avoid issues until permanent repairs could be made during the annual maintenance period when the reservoir was drained. Unfortunately, they would never get the chance.

Less than three months after that first time someone observed the reservoir overflowing, on December 14, it happened again, this time in the early dawn when no one was around to notice. Once the parapet wall collapsed, the water quickly eroded down through the dam, emptying roughly 6 billion liters or 200 million cubic feet of water down the steep mountainside straight toward Johnson’s Shut-Ins State Park, stripping away trees and rocks as it surged. By pure luck, the failure happened in the winter when the park was practically empty, but the park superintendent, his wife, and three kids (one of whom was only seven months old) were swept away when the water demolished their house. Incredibly, the entire family survived the event, but not without suffering from injuries and hypothermia. The wave of water flowed into the lower reservoir, where it would have gone anyway later that day, so there were no major downstream impacts.

The event was investigated by the Federal Energy Regulatory Commission, and the conclusions were surprising. Like most events of this kind, a series of small oversights, which on their own wouldn’t have resulted in a disaster, combined to cause hundreds of millions of dollars in damage, plus the effects on the family I mentioned. First was the embankment. That rockfill it was supposed to be built from wasn’t quite as rocky as the engineers who designed it intended. There was a lot more soil mixed into the fill, resulting in more settlement of the embankment over time. Unsound areas of soil in the embankment’s foundation were also not properly cleaned out, making the settlement even worse. From construction to failure, some parts of the parapet wall were a full two feet or 600 millimeters lower than where they started. That settlement wasn’t taken into consideration when the level sensors were replaced after the lining project in 2004. And with the sensors unattached and free to move around, there was no way for the logic controllers to know the actual elevation of the water in the reservoir.

Failsafe probes were installed on the parapet wall to provide a backup that would automatically shut off the pumps if the level got too high, but they were installed in a location that was actually higher than the top of settled parts of the embankment wall. If the water hit those sensors, it was already overtopping parts of the wall. And they were incorrectly programmed in a way that required both sensors to be activated before the pumps shut off. That first site visit when water was running over the wall didn’t trigger those failsafe sensors, but no one thought to check them. And rather than ground truth the important elevations like the top of wall and all sensor levels, they just decided to add a couple feet of margin and postpone a permanent fix. It would have been so easy to have someone on-site during those last few minutes of filling the reservoir each day to verify the levels against the electronic measurements, or even a closed circuit camera, especially after the enormous red flag of seeing it happen a few months earlier, but no one knew it was overtopping. And the owner hadn’t notified the regulator the first time it happened, so there was no oversight for how Ameren responded.

Probably the most significant error of all happened well before the facility was ever built. The design had no spillway. As an off-channel reservoir, there were only two ways water could get in: rain falling on top or water being pumped in. With enough freeboard for a rainstorm and the redundancies built into the control system, the designers never envisioned a need for a way to let water safely run over the top. Unfortunately, when you rely on complicated systems for safety, the likelihood for things to go wrong goes way up. These types of events are sometimes called “normal accidents,” a term coined by Charles Perrow. The idea is that, when systems are complicated, and especially when the safety measures themselves add to a project’s complexity, failures are much more likely, even expected. In other words, failure becomes normal. Compared to an industrial control system, a spillway is dead simple. Once the water gets to the crest, it just goes out. They’re not failproof - I’ve talked about several spillway failures in previous videos - but there are a lot fewer ways that things can go wrong.

FERC fined the owner 15 million dollars for the failure, the largest penalty they’ve ever issued. Five million of that went into a fund to improve the area around the project, although some recent reporting has alleged that those funds have been mismanaged. The State of Missouri also sued, and agreed to a 177 million dollar settlement, much of which went toward restorations at the state park, which held a reopening ceremony in 2010.

At the same time as Johnson’s Shut-Ins State Park was undergoing renovation, crews were working on rebuilding the upper reservoir at Taum Sauk. To avoid a relicensing process, the dam was built on the same alignment and to the same size as the original project. Rather than trying to repair the flawed rockfill embankment, Ameren and their consultants went with an innovative design. The new dam was built using roller-compacted concrete, a dry concrete mix that’s handled using earth-moving equipment and compacted into place using rollers. The new design would address both the settlement and leakage issues the original structure struggled with while still taking advantage of the material from the original embankment. That rock fill was crushed and processed into aggregate for the concrete, reducing the amount of material hauled into the remote site. Maybe most importantly, they included a spillway this time. The structure is the largest roller-compacted concrete dam in the United States. The plant reopened in 2010, was rededicated as an IEEE milestone, and the project won the US Society on Dams’ award of Excellence in the Constructed Project.

The failure at Taum Sauk was a wake-up call for the professional community. The regulator, FERC, implemented some big changes to its oversight of dam safety in the wake of the collapse. They put together a task force that issued a technical guidance document specifically addressing the challenges of pumped storage facilities that was circulated to the owners. They also updated rules for owners, requiring them to have an internal dam safety program and a Chief Dam Safety Engineer who is responsible for overseeing it, a role Ameren didn’t have at the time. The event triggered states as far as Hawaii to bolster their dam safety programs. And most importantly, the failure demonstrated the need for overflow spillways, even for off-channel reservoirs with redundant control systems meant to avoid overfilling.

If you’re paying attention to issues related to the electrical grid, you know the importance of storage. Particularly as intermittent sources of power become a large part of our portfolio, ways to balance out those mismatches in supply and demand are only becoming more important. Pumped storage has traditionally been the only large-scale way to do this economically, but obviously, it comes with a tradeoff. Dams are among the riskiest structures that humans build. They don’t fail very often, but when they do, those failures usually come with serious consequences to people, property, and the environment. And because they don’t fail that often, those lessons come slowly and tragically. But, with battery storage becoming cheaper and much more widespread, it will be interesting to see how the economics of pumped storage change. By 2030, some are predicting the US will have more than 400 gigawatt hours of battery storage on the grid; that’s more than 100 Taum Sauks. We’re right at the beginning of some major changes in how energy is stored. Those batteries have a lot of technical differences in how they interact with the grid, and they come with their own environmental challenges and safety considerations, but the risk profile is a lot different than building a major reservoir at the top of a mountain. As energy infrastructure keeps evolving, those differences in risks are probably going to shape the future of how we store power, and at what cost.

Is the World Really Running Out of Sand?

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

If you have to know the answer right away, it’s no; or at least, my goal with this video is to convince you that the world is not running out of sand. But if it were that simple, I wouldn’t be here (right?) and you probably wouldn’t be either. In fact, I was really surprised by some of the things I didn’t know as I dug deeper into the topic further, and how some of the most widely spread sand “facts” are dead wrong.

The wide world of sand is complicated, and not in a boring, pedantic kind of way. This simple material touches nearly every part of our lives, and the science and engineering behind it is rich and deep, and to me at least, hard not to become obsessed with. There’s a good chance you’ve seen articles or videos over the past few years with essentially the same story about sand. The “Sand Wars” documentary kind of kicked off the modern discussion, and then Vince Beiser wrote an excellent book on the topic, “The World in a Grain.” Of course, a lot of the book’s best reviews focus on the fact that sand is kind of a topic that’s “taken for granted” or “neglected.” But, at least in civil engineering, it is one of the most glected materials out there. And I’d like to give you a peek behind the curtain and show you how we think about this seemingly unlimited resource and why it’s worth knowing a little more about it. But first, I need to head out to the garage and put some sand in a rock tumbler, because I want to do a material property shootout, and this is going to take a little while. I have something really cool in the other barrel, and I’ll show it to you later on in the video. I’m Grady, and this is Practical Engineering.

What the heck is sand anyway? It’s kind of a “know it when you see it”-type material. If we use the US Department of Agriculture’s soil textural triangle, sand is any granular material that is at least 85% sand… So, what the heck is sand? For a better answer, we have the Unified Soil Classification System. Geotechnical engineers sometimes say that “dirt” is a four-letter word. Maybe because it undermines the importance of soil (which is also a four-letter word, by the way), but I like to think it’s because we have better names for all the dirts around the world, and here they are. In fact, there are four specific kinds of sand, but they all fit this one criterion, and it’s all about the size of the particles. At least half of those particles have to make it through a Number 4 sieve (about 5 millimeters), but no more than half can go through a Number 200 sieve (less than a tenth of a millimeter, or about 75 microns). That is a pretty wide range of materials, but I think when you picture sand in your mind, you probably imagine what the USCS would call “clean sand” where less than 12% pass the 200 sieve. So, just to make it simple, let’s say that sand is a material where the particles fit through this, but they don’t fit through this. But still, that encompasses a huge range of different dirts. And I hope, at this point, you’re asking, “Who Cares?” because I would love to answer that question.

In his book, Beiser calls sand “the most important solid substance on earth…the literal foundation of modern civilization…” We use it to make glass, semiconductors, fiber optics, filters, and abrasives, use it to texture surfaces, to play in, for beauty, and more. But, probably more than anything else, sand is an essential ingredient in concrete. And, you know, I’m a civil engineer; this is a channel about the built environment; so I wanna talk about concrete. And, in fact, if this video sparks your curiosity about one of my favorite materials, I have a whole playlist of topics I’ve covered in the past so you can learn more after this. You can’t really overstate how important concrete is and how much of it we use. There’s a bigger conversation to be had about its environmental impacts, but when compared to alternative building materials, it just has so much going for it. It is an extremely low-cost, durable substance that can be made into just about any shape you can imagine. Concrete has enabled us to build structures that last for generations from some very simple ingredients that are (mostly) available across the world: water, cement, gravel, and sand.

Most of those ingredients are mined and used directly as raw materials. And they’re usually mined close by. Transportation makes up a big part of the cost associated with sands used for construction, so the distance between where they’re found and where they need to go is highly correlated with how economical they can be. And that often leads to environmental impacts, some worse than others, depending on local regulations. It turns out that the best sand for concrete often comes from rivers, and mining in rivers can be particularly destructive because the impacts can spread upstream and downstream through changes in the nature of the channel. (I have a series of videos on that topic, too, by the way). Sand isn’t spread evenly throughout the world, and it’s a non-renewable resource. Geologic processes produce it a lot slower than we can use it. So, it makes some intuitive sense to say that we could eventually run out. But here’s a fact that is often overlooked in the discussion: we can make sand.

And it’s not that complicated, either. I talked about the definition of sand a little earlier, but here’s another one: it’s just very small rocks. And we have engineered machines that can transform big rocks into small ones. In fact, I have such a machine in my garage. It’s called a hammer. Some might argue that this isn’t the best use of my time, but I spent about an hour to artisanally manufacture a batch of sand just to hammer this point home. First, you crush the rocks. Then you put them through the sieves to remove the stuff that’s too big or too small. It takes a little extra processing, but this is not grain surgery. And this has a lot of benefits compared to sourcing natural sand. Hard rock quarries and crushing operations are already out there producing coarse aggregates like gravel, so sometimes the smaller stuff is a waste product anyway. It opens up possibilities when natural deposits aren’t available and can move mining operations upland, away from rivers, where the environmental impacts are less severe. And, it can make the concrete stronger. Let me show you what I mean.

I took some sand out of my kids’ sandbox and put it in a rock tumbler for a week to try and simulate the erosion it might see over years in windswept dunes of a desert. Obviously, this erosion reduced the overall size of the particles. So, I classified both materials with the sieves to make them a closer match for a fair comparison. And both batches are within the spec for concrete sand in the US. Looking in the microscope, you can clearly see the differences. The tumbled sand is rounded with roughly spherical, smooth grains. The manufactured sand is jagged with sharp, angular corners. And watch what happens when I pile them up. I filled up two pieces of pipe with the same amount of sand, and then pulled the pipe away. It’s not a huge difference, but you can see the rounded sand spreads out a little further because the particles have less friction. It makes intuitive sense that concrete made from this sand would be weaker than with this sand. Let’s see if that’s true.

I mixed up a simple batch of concrete from the crushed sand, and the tumbled sand keeping the weights of all the ingredients equal. Here’s my recipe if you want to try this experiment yourself. Then I molded some concrete cylinders and let them cure for a week. Most concrete mixes are meant to reach their design strength after 28 days, but concrete strength gain is fairly predictable, so the relative difference between the samples should be consistent in time. Most importantly, my little benchtop hydraulic press would not be able to break these samples if I waited that long. And this load cell is not calibrated, so I’m doing this test in arbitrary Practical Engineering units of force. The tumbled sand cylinder broke at around 2500 units. And the manufactured sand concrete broke at 7500 units. You can easily see the difference in the results. “Goodnight!” It was 3 times as strong. Of course, this is my garage, not a testing lab, and I only did one sample because my arm got tired of hammering rocks. Luckily, people much smarter than me have tested this out, and the results are pretty conclusive that, if you keep everything the same, the angularity of fine aggregate increases the strength of concrete. And that’s the story you probably know if you’ve read anything on this topic. It’s the common explanation for why we don’t use dune sand, the most visible of earth’s sand resources, in concrete. It’s intuitive. Rounded grains don’t lock together. Beiser makes the claim not once but three times in his book. But it turns out it’s not that simple, because strength isn’t the only property of concrete that we care about.

Before concrete has to be strong, it has to be placed. Ask anyone who’s done this kind of work, and they’ll tell you it’s hard. Well, it’s liquid at first, but it’s hard to work with. Concrete is about two-and-a-half times the density of water. It’s heavy stuff, and to get it into the forms often can require a lot of different tools: wheelbarrows, buggies, chutes, pumps, hoses, and more. The better the concrete flows, the easier it is to do a good job of placing it. And that matters. If a mix is too stiff, it can clog up hoses, trap air bubbles, and ultimately lead to poor quality in the installed product. This property of concrete is usually called workability. It’s often measured using a slump test. Fill up a cone of concrete, pull the cone away, and see how far the concrete slumps. But the problem with workability is that, in one big way, it works against strength. And it all has to do with water.

What happens when the concrete truck shows up to your job site and the mix is too stiff? Depends on if the engineer is there or not, but in a lot of cases, you just tell the driver to add a few gallons of water to the mix. More water; better flow; easier to place. It’s pretty straightforward, but there’s a reason you don’t want the engineer to know: water decreases concrete strength. I’ve done a whole video about this with some garage demos, so again, check that out if you want to learn more. The gist is that the ratio of water to cement is one of the most important factors determining concrete's strength when it cures. Cement isn’t like some types of glue that harden as the water or solvent evaporates. It goes through a chemical reaction, incorporating the water into the final product. That’s why we say concrete “cures” instead of “dries”. But, cement can only react with around 35 percent of its weight in water, so any more than that is just taking up volume in the mix that could be used by the stronger ingredients. More water; less strong. And here’s where the shape of the sand grains comes into play.

I did a little garage slump test to gauge the workability of those two mixes I made for the earlier demonstration. Here’s the rounded sand mix… and here’s the manufactured sand. “Haha, no slump at all.” Honestly I expected this to be a subtle difference, but it was like night and day. They weren’t even close. So I wondered, what would happen if, instead of holding ingredient ratios constant, I used the workability as the controlled variable? Let’s find out. First I used the manufactured sand with enough water to get it to a workable level. 100 milliliters got it to here, which is a bit better than the first one. Then I did a second mix with the tumbled sand, slowly adding water and running the slump test until it was pretty close. It took only 70 ml of water to make them match, 30% less than the first batch. After a week, I tested the samples. The tumbled sand sample broke at 4,800 units. And the manufactured sand broke at only 4,300 units. The tumbled sand with the rounder grains was stronger this time (by about ten percent), and it’s all due to the lower water content in the mix.

So yes, if you use the same amount of water, more angular sand like you might find from a river or manufactured sand is better, but that’s not what happens in real construction. I say this with many, many caveats, but very generally, you only add as much water as you need for workability. Rounded sand gives you better workability, so you can add less water, and thus get stronger concrete. This idea that we can’t use wind-blown sands in concrete because of their shape is a myth. In fact, the American Concrete Institute has a bulletin that says it better than I can:

“The influence of fine aggregate shape and texture on the strength of hardened concrete is almost entirely related to the resulting water-to-cement ratio of the concrete…”

I tried to track down the original source of this idea that we can’t use rounded grains in concrete, but got nowhere. Beiser cites an article from the UN, which itself cites a 2006 paper about using two types of desert sand from China in concrete. But that paper doesn’t mention the roundness of the particles at all. They didn’t include any measure of the shape of the grains in their study, and they didn’t make any suggestions about how that particular property of the desert sand may have affected the results of their tests. In fact, the conclusion of that paper includes saying that desert sand is a feasible alternative to other types of fine aggregates used in concrete. And the whole reason it was a subject of scientific study at all has to do with size, not shape. This is a widely used specification for the distribution of particle sizes of fine aggregate for concrete. Any sand in this area meets the spec. And here’s the soil used in that paper. Even if the conclusion was that it doesn’t work, I think this would have a lot more to do with the results than the shape of the particles.

And that really gets to the heart of this whole discussion. Fine aggregates are found throughout the world. We can even make our own. And concrete is like baking; different ingredients can change the end results. But just like regional bread recipes evolved based on the availability of local ingredients, the construction industry has developed a lot of ways to use different local materials to achieve good structural properties. The real challenge, like many things in engineering, is cost.

It can be more expensive to manufacture sand compared to mining raw materials that can be put directly in a mix, especially when you factor in the other ingredients, like chemical admixtures, that might be required to make it more workable without adding too much water. It’s expensive to transport better quality sand from far away, rather than finding it close to a job site or batch plant. It’s expensive to mine sand in adherence to environmental regulations that are becoming stricter worldwide. It’s catchy to say there’s a scarcity of fine aggregates on earth, but I think it’s misleading. “Sand is getting a lot more expensive than it used to be” just doesn’t make as nice of a headline. And the tricky part is that, in many ways, those costs have always been there; we’ve just externalized them onto the environment and our future.

All the ingredients in concrete are mined or harvested just like other natural resources. It’s just that concrete is made on a scale that blows most other materials out of the water. It’s a huge business, and there’s lots of money flowing, which means a lot of potential environmental harm and social conflict as a result. That’s especially true in places that don’t have robust oversight and enforcement of how sand is extracted. And I think it’s important to point out that the low-cost of sand, because of its simplicity as a material and its abundance, is a big part of why we use so much concrete in the first place, even in situations where it’s not necessarily the best material choice in other respects. Everything in engineering is a tradeoff, and if the economics around sand change, the engineering and construction industries can change with them. Look at other examples of this.

Diamond used to be exclusively a mined material, but now we can make it in a lab. Synthetic diamond gemstones used in jewelry are now less expensive than mined diamonds, but, admittedly, there’s a lot more to that economy than the costs to make them. What I think is more interesting is that 99 percent of diamond used in the world for industrial purposes is synthetic. It used to be a rare mineral, but now you can pick up a diamond drill bit or saw blade from the hardware store for a fairly small premium.

Timber is another example. Natural forests used to be the only source, but now plantations, trees planted specifically for harvest, now make up more than a third of the wood we use globally. And engineered lumber like plywood, OSB, and structural composites can make more efficient use of raw materials. I’m pointing out these examples, not to say they’re good or bad - there are pros and cons in both cases - but just to illustrate how our demand for materials in the construction industry changes with the supply, and how technology can have a huge impact on that. And there’s another parallel between timber and sand: they both can be renewable.

I had another barrel in the rock tumbler not going to use, so I broke up some chunks of concrete and threw them in. I ran these through the grits, just like you would with any other rock in a tumbler. Concrete is pretty soft compared to most natural rocks, so it didn’t polish up that nicely, but the result is still pretty cool. You can really see the constituent materials of the concrete after it spent so long rolling around in there: the small and large aggregates and the cement paste locking them together. But the point of this demo is that concrete is pretty much just rock, that’s mostly what it’s made of in the first place. And just like the rocks I crushed to create manufactured sand, concrete can be recycled into aggregates that get reused in the construction industry, either in new concrete or other materials, reducing demand for virgin sources.

There’s a lot changing in the construction industry, and a lot of growth in the need for materials like sand and gravel. But I don’t think it’s fair to say the world is running out of those materials. We’re just more aware of all the costs involved in procuring them, and hopefully taking more account for how they affect our future and the environment.

When Infrastructure Gets Hacked

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

This is a water tower, or as the pros would say, an elevated storage tank. Pretty common here in the US, especially in flatter areas where there’s no nearby hillside to build a ground-level tank. I have a whole video about how these work. In the most basic sense, a water tower is a buffer between the ever-changing demands for fresh water in a distribution system and the high-service pumps at the treatment plant that like to run at a constant rate. The level in the tank is a key measure of performance. If it’s high, pressure in the system is good, and the pumps can shut off… unless someone has messed with the computer system that controls that relationship.

In early 2024, that’s exactly what happened in Muleshoe, a small town in the Texas panhandle. A citizen noticed water spilling out of the elevated tank and reported it. When the city went to investigate the problem, they didn’t find a stuck valve, malfunctioning sensor, or broken pump contactor. The water tank was overflowing because of a deliberate attack by a hacking group linked to the Russian military

Some water was wasted, but ultimately, no one was hurt, and nothing was damaged in the attack. Muleshoe was probably just a victim of opportunity. Having grown up in the Texas panhandle, I think I’m safe saying that most towns there aren’t necessarily considered high value targets for international criminal campaigns. But that’s the thing with cyber security these days. It’s not just for the organizations with big secrets and lots of money. Even in tiny west Texas towns, critical pieces of infrastructure are run by computers, and a lot of them are connected to a network, making them vulnerable to bad actors. I’m not a security expert; I’m just a civil engineer. But, I’ve worked on a lot of projects where digital systems interact with infrastructure, and I’ve collected some really interesting stories about how that can go wrong that I thought would be fun to share. So let’s peek behind the control panel and talk about them. I’m Grady, and this is Practical Engineering.

Once upon a time, everything from the power grid… to drinking water distribution systems.. industrial manufacturing… to oil and gas…, dam operations, and more was run without the aid of computers. Calculations were done manually, and engineers carried slide rules. Decisions were made by skilled operators, valves were opened and closed by hand, wear and tear was measured by human eyes, and so on. It’s easy to see the opportunities for digitization. If you’re not relying on a person for everything, you can be more efficient, reduce the chance of error, and improve safety by not requiring workers to be so hands-on. And there are quite a few ways to computerize the control of industrial processes like operating a pipeline or a water system.

One of the most widely used is called SCADA or Supervisory Control and Data Acquisition. This is a fairly standardized architecture used in a wide variety of industries like manufacturing, oil and gas refining, and most of the utility systems we rely on like electricity, water, sewer, traffic lights, and more. Let’s look at an example of a basic municipal water system to see how it works.

Getting fresh water distributed to a city is a big undertaking that requires a lot of equipment, including valves, tanks, pipes, pumps, chemical systems, and more. Some of these will include sensors to take some kind of measurement, such as the water level inside a tank or the flow rate within a main line. Others will include actuators, devices that can do something like turn on a pump or open a valve. All the devices connect to one or more remote terminal units or RTUs. All the RTUs are then networked to a central supervisory computer that sends control commands and collects the data. This computer normally includes the Human Machine Interface or HMI. This is where an operator interacts with the system, and they’re usually set up as simplified diagrams of whatever’s being controlled.

Systems like this can be programmed to maintain certain conditions and automatically adjust equipment to keep everything running smoothly and as expected. Automated systems never get bored of doing the same thing over and over again, they don’t need to sleep, and they don’t mind being exposed to hazardous chemicals. For example, let’s look at the high service pumps and water tower. These are often configured in a lead-lag system with multiple pumps for redundancy and reliability. When the level in the tank drops below a set point, the lead pump turns on. With smaller demands, this will fill the tank to the upper set point, at which point the lead pump turns off. But under higher demands, the lead pump might not be enough. If the level continues dropping while the lead pump is running, a lag pump with a lower set point will kick on. With both pumps running, the tank will fill, eventually reaching the upper set point and kicking both pumps off. If you want to see an example of this in action, check out my Practical Construction series where I embedded on a construction site of a sewage pump station and documented the process from start to finish.

That’s a basic example, but you get a sense of how useful a SCADA system can be. You don’t have to manually control the pumps or be on site to check the tank level. And you can change those set points. Maybe during seasons when demand is low, you don’t want the tank full all the time, because the water spends too much time in there where its quality can degrade. You don’t have to hire an electrician to reconfigure a control panel or put a technician in a dangerous spot to adjust floats in the tank. Any trained operator can just change the values in the HMI. They’re designed for simplicity, and in fact, I’ve always thought they often look a lot like old video games. There’s something really nostalgic about the basic graphics HMI’s are often designed with. It’s easy to forget that they’re connected to real systems, often large and complex systems, where the stakes are high if something goes wrong, which is exactly what happened in Muleshoe.

According to security researchers, Muleshoe’s SCADA system was breached by a group called the Cyber Army of Russia Reborn with a portal set up so the city could have remote access. On January 18th, they posted this video supposedly showing them manipulating the HMI’s of two small Texas water systems. Judging by the haphazard clicking around, it seems that the hackers know a lot more about gaining access than they do about how water systems work. Most of the video seems to be someone clumsily navigating screens and changing values at random. Nevertheless, they managed to change a setpoint on one of Muleshoe’s booster pumps, leading it to stay on even after the water tower was full, and eventually causing it to overflow.

Ultimately the attack was pretty harmless, but it could have been worse. A similar event happened in Oldsmar, Florida in 2021 when a hacker reportedly changed the sodium hydroxide feed in the water treatment plant from 100 parts per million to 11,000. The event brought huge national attention to the issue of information security for critical infrastructure. Two years later, the FBI concluded it probably was an employee mistake and not an actual intrusion, but it was still a strong reminder of the type of havoc that could result from a SCADA system with poorly secured remote access.

Even further back than that, a SCADA system controlling sewer works in Maroochy Shire, Australia was hacked by a disgruntled ex-contractor, releasing thousands of gallons of untreated sewage into parks and waterways in 2001. And really, there’s no telling how many similar attacks have happened across the world. A lot of them don’t make the news, and even though they’re often investigated by authorities, the details aren’t released for fear of sharing potential vulnerabilities that aren’t patched up in other systems. It’s a constant arms race happening mostly behind the scenes. Hackers are constantly probing systems for vulnerabilities, especially ones that are previously unknown (called zero-days, because that’s how long they’ve been known about when exploited). But access to industrial control systems isn’t the only digital threat to infrastructure.

In May of 2021, the Colonial Pipeline Company, owners of the largest refined petroleum pipeline in the US, was attacked by another Russian group called Darkside. They didn’t gain access to any pumping or control systems. Instead, they installed ransomware on the billing computers, locking the company out, and stole sensitive information, threatening to release it if the company didn’t pay. Not knowing the extent of the threat, the company shut the pipeline down. Over the next six days, a gasoline panic struck the eastern US, with gas hoarding emptying out more than 12,000 filling stations. A state of emergency was declared, and rules governing tanker trucks were relaxed to allow for more fuel to travel by road.

With FBI oversight, Colonial paid the ransom, 75 bitcoins, or about 4.4 million dollars at the time, within hours of the attack. But the tool provided by the group to unlock the system was so slow, that they ended up using mostly their own backups to get the billing system back online. Some of that ransom was eventually recovered, but it took six days to get the pipeline started up again, and there’s still a ten million dollar reward out for information leading to key leaders of the group. So how did they do it?

Really it wasn’t very sophisticated. An employee was reusing a password that had been leaked in a database from a prior breach. They just logged into Colonial’s VPN with purchased credentials. That’s all it took to take down one of the US’s most important pipelines for six days. Again, with that kind of access, it could have been a lot worse. And one thing you’re probably thinking is, “why would you have the ability for remote access to critical systems like this at all?” Is it really worth exposing yourself to the entire world of nefarious actors, just to save a commute to the HMI? And, actually, a lot of critical systems don’t have an outside connection. They’re air-gapped. But even that’s not a foolproof system.

One of the first, and maybe well known, examples of infrastructure hacking, especially an example designed to cause permanent physical damage, is STUXNET. Although the details are pretty murky, STUXNET seems to have been developed by the US and Israel as a military-grade cyber weapon. It was a worm, first reported in 2010, that specifically targeted SCADA software on Windows computers. Stuxnet famously exploited four zero-day vulnerabilities to spread and interact with SCADA systems. If a computer didn’t have the target software, it would just do nothing except replicate. But when it found a computer with SCADA software and some very specific motor drives connected, it would send a command to rapidly speed up and slow down the motors while faking sensor data so that the SCADA system wouldn’t shut down or throw an alarm that something was awry. Those specific motor drives were pretty much only used in gas centrifuges used to enrich uranium so it could be used in nuclear weapons. It’s pretty clear what the worm was designed to target, and it did work. STUXNET reportedly destroyed around a fifth of Iran’s nuclear centrifuges, and probably shortened the lifespans of many more. And it was introduced to the facilities’ networks not through a remote connection (the system was airgapped) but from an infected USB drive.

And that really is the key to all this. Your cybersecurity is only as strong as one employee’s willingness to plug in a USB drive or reuse a personal password at work or click a deceptive link in an email or hold the door open for someone following behind them. And most of us are guilty of doing these things. At least, every once in a while… But, these days, no matter who you are or what you do, you probably use some kind of digital device in your life. And so whether you’re the operator of a tiny water system in rural Texas or manage the largest gasoline pipeline in the US, you also kind of have to be a cybersecurity expert too. The stakes are high. Digital systems interact with every aspect of our daily lives and basic needs: water, electricity, sanitation, public health, transportation, and more can all be seriously disrupted by someone or some group, anywhere in the world, if we let our guard down. With great computer power comes great computer responsibility. And just because many of these industrial control systems are only used or understood by a small group of people, security through obscurity just isn’t realistic anymore.

The Hidden Engineering of Landfills

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

This is the Puente Hills Landfill outside of Los Angeles, California. The first truckload of trash was dumped here in 1957, and the trucks just kept coming. For more than five decades, if you threw something away in LA County, there’s a good chance it’s buried somewhere inside this mountain of waste. At its peak, Puente Hills was accepting around four million tons of trash every year, making it one of the largest landfills in the country. It closed in 2013, creating a time capsule of everyday life and consumption patterns over a span of 56 years. But Puente Hills is also a time capsule of landfill engineering itself. In 1976, right in the middle of its lifespan, sweeping federal regulations changed how we deal with solid waste forever.

You probably don’t think too much about where your trash goes, and that’s kind of the whole point of the solid waste industry: to make sure you have the ability to throw something away without it having a serious negative consequence on the environment or public health. There’s a larger conversation to be had about the amount of waste we generate and how much of it can be recycled or reused, but there is always going to be stuff that just doesn’t hold enough value to be kept. Trash is an inescapable element of the human condition. And, I think you’re going to be surprised how complicated that really is. When Puente Hills opened in the 50s, a landfill was pretty much just a hole in the ground where trash was dumped. By the time it closed, landfills were highly engineered holes where trash gets dumped. And I have a scale model of a landfill in the garage to show you how it all works. I’m Grady, and this is Practical Engineering.

There are lots of kinds of waste in this crazy world, but one of the biggest sources is just you and me throwing stuff in the trash. The technical term is municipal solid waste, since its collection is usually coordinated at the city level. There are a lot of ways to manage it once collected, but the most common by far is disposal in a landfill. And, one of the biggest parts of landfill engineering is just deciding where to put one in the first place. The main goal of a landfill is to maximize the volume of waste that can be stored there while minimizing the cost and the environmental impacts too, which turns choosing a suitable site into a giant geometry problem.

Digging a hole sounds like an obvious choice, but consider this: digging a hole is expensive, and not digging a hole is free. There are costs of excavating tons and tons of soil just to get it out of the way so it can be replaced with trash and costs of hauling away all that soil (since your goal is to maximize the volume on the site). Plus, you have to avoid the water table, any unsuitable geology, and the challenges of building and working deep below the surface of the earth. That’s why most landfills mostly build up into what sanitation professionals call the “air space.” Looking upward, it may seem like the sky is the limit, but anyone who’s built a tower of anything, let alone trash, knows better. The waste pile gets less stable as its height increases, requiring shallower slopes. And the pressures at the bottom go up too, which can lead to settlement and damage of facilities. Plus, there are visual impacts. The bigger the garbage heap, the bigger the eyesore, and people are only willing to look at a landfill so tall.

They can’t be too close to airports, because they attract birds that can interfere with planes. And they can’t be too close to homes, parks, playgrounds, and other places people congregate for obvious reasons. Of course, there’s floodplains and wildlife habitat to avoid as well. And you don’t just need a place to put the trash. You also need a scale house to weigh the trucks coming in and out, a shop and storage for the equipment, and sometimes a place for ordinary citizens to drop stuff off. Finally, you need a spot that can handle the huge increase in truck traffic coming and going, practically nonstop. Pretty much, if you can get a college degree in it, it’s going to come into play when siting a landfill: geology, geography, politics, archaeology, public relations, biology, every kind of engineering, and lot more.

But once you have your landfill, you can’t just start dumping trash. Let me show you why with a demonstration with some help from my shop assistants. I have my hole dug, and we’ll start adding some trash. So far, no major problems. But eventually, it’s going to rain. And you can’t immediately see the issue. Granted, this is more of a flood than a drizzle, but it gets the point across. All that water is going to filter through the garbage to the bottom of the hole, and, eventually, into the underlying soil. It might go without saying, but I’m going to say it anyway: We really don’t want garbage juice percolating into our soils. Mainly because it can contaminate sources of groundwater, but also because it can migrate well beyond the limits of the landfill, causing all sorts of environmental troubles. So, modern landfills use a bottom liner to keep waste separate from the underlying soils. Often this consists of a thick sheet of plastic, carefully tested and welded together into an impermeable membrane. Even the area between the plastic welds is tested using air pressure to make sure there are no leaks. Another option is thick clay soil compacted to create a watertight layer. In many cases, the two options are combined, so you end up with this intricate structure of different impermeable layers stacked together.

Maybe you still see a problem with this solution on its own. Now when it rains, the landfill just fills up with water. This causes issues with stability and settlement. It causes garbage to decompose more quickly, leading to odor and temperature problems. Plus, you just can’t work on top. There’s no way for trucks to unload trash on top of a garbage swamp. So we need a way to get the garbage juice out, without letting it flow into the soil below. By the way, garbage juice isn’t a technical term. It’s actually called leachate, so I’ll use that from here on out. And all modern landfills have sophisticated leachate collection systems to keep the waste as dry as possible and avoid the issues I mentioned. Usually, this consists of a system of perforated pipes covered in a layer of sand, draining to sumps, and eventually leading out of the waste.

I built a little leachate collection system in my model landfill using a small tube so you can see this in action. Now my clay bathtub has a drain. When the rain comes, the water that makes its way into the waste is able to flow out of the landfill, keeping it from becoming a swampy mess. This is a little simplified compared to a real landfill. I’ve made a video all about French Drains, which is much closer to what a leachate collection system consists of if you want to learn more after this. Obviously, in my example, the leachate system has to penetrate the bottom liner, which can be a potential source of leaks. So these penetrations are sealed really carefully in the real world, or the collection system just uses pumps and risers that run up the slope of the landfill to the top, so no penetration system is necessary.

Of course, now you have a stream of leachate you have to deal with. Actually, leachate management is one of the biggest costs of running a facility like this. Some landfills send it off to a treatment plant that can clean it up. Some have ways to treat it on-site with settling ponds, evaporation, biological treatment, and even plants that can consume and convert landfill leachate into waste that’s easier to dispose of (maybe even back into the landfill itself).

Finally, the bottom of our landfill has all the necessary pieces, but the work doesn’t stop there. Remember that volume is everything in a landfill. For as much effort goes into finding a location and building the infrastructure, it’s essential that we get the most trash in here as possible. You probably know this, but municipal garbage just isn’t that dense. Maybe you’ve had to smash a few more bags in the can because you missed the collection one week. If so, you know there’s usually a lot of room for densification. The trucks that collect garbage usually have a way to compact it to make more room in the box before needing to be emptied. But once the trash is at the landfill, there’s still an opportunity for compaction. Landfills often use massive roller compactors with enormous teeth and giant blades to grade out and compress waste and get as much as possible into the site. It saves money, and it’s good stewardship of the space. But density isn’t the only challenge with day-to-day operations.

Despite what you’ve heard, landfills are kind of gross. I mean, that’s their whole point is to accept the stuff we don’t want to put anywhere else. But putting it all in one place creates a lot of problems: pests, odors, windblown waste, fires, birds, and more. So to mitigate some of that, most places require that the garbage be covered up at the end of every day. This “daily cover” can take a lot of forms. The basic approach is just to put a layer of soil over the top of the working face at the end of the day.

When I do this in my model, you get a sense of the problem. All that clean daily cover is taking up precious space in the landfill. One option is to trim it back off each morning before trucks start arriving, but that’s a sisyphean task of just moving tons and tons of soil around each day. Other alternatives for daily cover are tarps, or just holding back certain types of waste that are more inert like foundry sand, foam, paper, and shredded tires. They’re going in anyway, so you might as well use them on top to cover the more disagreeable stuff overnight. Those alternatives can also help avoid leachate getting perched within the waste, encouraging it to continue downward to the collection system.

Ideally, a landfill will last for decades, slowly filling up by packing as much waste as possible. Throughout the course of operating a landfill, there’s constant testing of groundwater, surface water, leachate, air quality and more to make sure they’re not exceeding limits. Landfills are usually built in smaller cells so you don’t have to manage this huge area of waste all at once. A cell fills up, you put soil over the top (called interim cover), and start a new one within the landfill. But eventually, you reach the top of the airspace, and the landfill reaches the end of its useful life. And closing a landfill is not an easy job. Of course, you have to cover all that waste up, creating a mountainous sealed tomb of garbage. That final cover has to keep water out, to reduce the volume of leachate you’re having to collect and treat over time. But it also has to keep the garbage in, and not just the garbage itself, but anything else that comes with it like smells and leachate and pests. And it has to do it basically forever. So, just like the bottom liner, the final cover over a landfill is usually a system of multiple layers, including compacted soil, membranes, and fabrics. And then you have to get the grass to grow, to protect the soil from erosion and damage over time. I don’t have time to wait for grass to grow in my demo, so I’m cheating a little bit.

But the fun isn’t quite over yet. The waste may be sealed up, but that doesn’t mean it’s inert. In fact, there’s a lot of chemistry and biology happening inside a landfill, and a lot of those reactions generate gases like methane and hydrogen sulfide that can create pressure, heat, smells, greenhouse effects in the atmosphere, and the potential for explosions. So, one of the steps in landfill closure is to install wells that can collect the gases from the waste. Usually, these consist of vertical pipes connected to a blower that constantly draws air to a collection point. There’s a lot that goes into these systems too. You can’t pull too hard, or you might draw oxygen into the landfill, changing the reactions and microbiological processes, and creating a potential for a fire within the waste. Plus the gas includes a lot of humidity, so managing condensation creates another liquid stream that has to be collected and treated. Once it’s collected, the landfill gas can be flared, combusting it into less environmentally harmful constituents. Another option is to put it to beneficial use to create heat or even electricity. The Puente Hills landfill I showed earlier has a gas-to-energy facility that’s been running since 1987, and even though the landfill is now closed, it currently provides enough electricity to power around 70,000 homes.

Once a landfill is closed, there’s not a lot you can do with it after that. It’s a big, sealed up, mountain of trash, after all. Owners are generally required to look after a closed landfill for at least 30 years afterwards, inspecting for leaks, monitoring the air and water, and repairing any damage. Those costs have to be built into the rates they charge, since there’s not a lot of benefit (or revenue) after closure. But, with all that open space and carefully-maintained landscaping, one option that many landfill operators are trying out is parks. And I love this idea. They say, “We’re willing to put our money where our mouth is and invite the public to spend time here, to enjoy this place that used to be, you know, one of the least enjoyable places you can imagine.” Puente Hills in California has big plans, including trails on the slopes, biking, slides, gardens and more. It looks like it will be a really nice place to visit when it’s done. And it also puts the whole concept of landfills in perspective.

Of course, we have a lot of room for improvement in how we think about and manage solid waste in this world. Landfills seem like an environmental blight, but really, properly designed ones play a huge role in making sure waste products don’t end up in our soil or air or water. It’s not possible to landfill waste everywhere. Many places are too densely populated or just don’t have enough space. But where they are, the environmental impacts are relatively small. Just consider the resources that go into them. I pay about 20 dollars a month, probably a little on the low end of the national average, and that buys me 64 gallons (about a quarter of a cubic meter) of space in a municipal landfill per week. Of course, I don’t fill the can every week, and that trash gets compacted. But still, do that for a decade, and your 20 bucks a month has paid for the volume of a modest apartment. It’s covered the cost of building the lining and collection systems, the environmental monitoring, the daily operations, the closure, the gas collection, and the maintenance for at least three decades afterwards and for your trash to stay there effectively forever. It’s (almost) free real estate, not that you’d want to live there. But my point is: landfills are a surprisingly low-impact way to manage solid waste in a lot of cases. I hope the future is a utopia where all the stuff we make maintains its beneficial value forever, but for now, I am thankful for the sanitary engineers and the other professions involved in safely and economically dealing with our trash so we don’t have to.

Why Are Texas Interchanges Texas So Tall?

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

This is the Dallas High Five, one of the tallest highway interchanges in the world. It gets its name from the fact that there are five different levels of roadways crossing each other in this one spot. In some ways, it’s kind of atrocious, right? It’s this enormous area of land dedicated to a complex spaghetti of concrete and steel; like the worst symbol of our car-obsessed culture. But in another way, it really is an impressive feat of engineering. 37 bridges and more than 700 columns are crammed into this one spot to keep the roughly half a million vehicles flowing in every direction each day.

They say everything’s bigger in Texas, but that’s not always true when it comes to engineering projects in the US. The tallest concrete dam is split between Arizona and Nevada. The longest bridge span is in New York. The longest road tunnel is in Alaska, and the longest water tunnel, not only in the US but the whole world, is the Delaware Aqueduct in New York. The largest hydroelectric plant is the Grand Coulee Dam in Washington, while the largest nuclear plant is in Georgia.

But one thing that Texas really does do bigger is highway interchanges. If you’ve driven from one major Texan highway onto or over another, you may have been astonished to find yourself and your vehicle well over a hundred feet or 30 meters above the ground. There’s no clearinghouse of data for flyover ramp heights, as far as I can find. Plus there’s the complexity of what a true height really means since many interchanges use excavation below grade for the lower level. Still, even the most conservative estimate puts the High Five taller than the Statue of Liberty from her feet to the top of her head. And if you do a little digging, you’ll find that many, if not most, of the tallest highway interchanges in the world are right here in the Lone Star State. Let’s talk about why. I’m Grady, and this is Practical Engineering.

The idea of a freeway really started in the 1920s with what’s now the Autostrada A8 in Italy: an automobile-only road with controlled access. Freeways are separated from local roads with limited ways to get on and off. And if you’ve driven a vehicle in the past century, the idea of a controlled-access freeway is pretty much taken for granted. Smooth curves and limited chances to enter or exit mean more speed and more capacity. But eventually, those big roads intersect other roads (sometimes other big roads) and that creates an obvious challenge.

Unlike most roads that cross at the same level on the ground, or as engineers say, “at grade,” freeways use grade separation at intersections. Roads go over or under one another. No traffic signals, stopping, or interruptions. Again, this is nothing groundbreaking. But what if you want to turn from one road onto the other? Just like that, we’ve gone from an intersection to an interchange. And this is where things get a lot more complicated. But we have to build up to it.

The diamond interchange is probably the simplest way to get grade separation because it kind of half doesn’t. Through traffic on the freeway flows right by, in most cases without any need to slow down. But that’s not true at the crossroad. Ramps enter and leave the highway at gentle angles and meet the crossroad nearly at right angles. Viewed from above, the ramps form a rough diamond shape, giving the interchange its name. The intersections of the ramps and the crossroad are just that: intersections. They are usually controlled by stop signs or traffic signals. Diamond interchanges can often get away with having just one bridge, a relatively small one carrying the crossroad over the highway. So, this can be the cheapest and easiest to build type of interchange to build. But, those intersections create limitations on how much traffic it can handle, so it’s really only used when the cross road is a minor one.

This kind of interchange is sometimes called a service interchange, in contrast to a system interchange, where two controlled access highways cross. As traffic increases, the only way to increase capacity is to eliminate at-grade intersections. So, the largest interchanges implement grade separation for every lane. The classic system interchange is the cloverleaf. Four ramps form a diamond, usually for the right-hand turns. These are directional ramps, that is, they curve toward the ultimate direction a traveler is trying to go. You exit right and end up driving to the right. The OTHER four ramps give the cloverleaf its name. The loop ramps, usually used for left-hand turns, curve around while ascending or descending so they can cross over themselves. So, you can get traffic flowing in any direction with no at-grade intersections and just one bridge.

The loop ramps make the whole thing look like a four-leafed clover, but finding yourself on this type of interchange doesn’t usually feel very lucky. For one, the loops are often pretty tight, requiring motorists to slow way down. And for two, there’s the weave. Consider traffic entering the highway from one of the loops. In the same place vehicles are trying to get back up to speed and merge left onto the freeway, drivers trying to exit the highway are slowing down and moving right. This inevitably creates traffic as people struggle to merge and cross paths with one another. Along with suboptimal traffic conditions, cloverleaf interchanges eat up a lot of of land. When cloverleaf interchanges were at their height of popularity in the mid-20th century, land was plentiful, and there were fewer cars, but as the volume of traffic increased AND the cost of land went up, engineers had to come up with new solutions to build better grade-separated highway crossings. And so they did.

Now, there’s such a huge variety of freeway interchange designs that it would be impossible to cover them all. The turbine, the windmill, the braided interchange, the ITL, mixes of various designs, and more. Each of these balances the constraints of a project like this in a different way: land requirements, cost, capacity, safety, et cetera. And the design that generally provides the most capacity, on the smallest footprint, (often for the highest cost), is the stack.

Like the cloverleaf, a stack has the four directional ramps, usually for the right-hand turns. But we move the exit for the left-hand turn off the main highway to avoid the weaving problem, and fly them over the middle of the intersection where they meet up with the opposite directional ramp. These ramps are often called flyovers, and it’s easy to see why. The gentle curves and elevation changes of the stack mean that drivers can safely maintain speed whether they’re going straight through the interchange or changing direction. The curved ramps often bank to the inside of the curve, called superelevation, making it even easier to maintain speed through the turn. This conventional configuration is called a four-level stack. There’s one level for the freeway, another for the crossing freeway to pass over, and two levels for the flyovers. It’s bridges on bridges, each one providing enough clearance underneath for large trucks. So these upper ramps end up pretty high off the ground. Four-level stacks are actually fairly ubiquitous in the US these days. These are impressive structures in their own right, but this is where Texas takes it to another level, literally. And it mostly has to do with feeder or frontage roads.

Lots of highways use frontage roads running parallel to connect areas alongside that would otherwise be cut off from the roadway network. They allow businesses to develop right up to and facing the freeway with easy access to those coming on and off it, basically keeping areas attached to the roadway network. Texas took the idea and ran with it. Apparently, they started as a way to reduce the cost of acquiring land for road projects. If you could promise the landowner access to a new highway along a frontage road, you're making their property more valuable, so they’re willing to sell a portion for the highway at a much lower cost. Now, Texas has over 6,400 miles (or 10,300 kilometers) of frontage roads. That’s almost the circumference of the moon, and as far as I can tell, way more than any other state in the US. I won’t go into the pros and cons of this approach here. Some research has shown pretty conclusively that the money saved on acquisition costs doesn’t make up for their many disadvantages. And Texas has since changed its policy to only include frontage roads on new freeways where necessary and justified. Although, from what I can tell seeing new construction these days, there don’t seem to be many projects where they’ve been left out. And one major effect of putting frontage roads alongside every highway happens at interchanges. Because these are more roads that need grade separation from all the others. So, at stack interchanges around the state, there aren’t just four levels but five.

In fact, this kind of interchange is often referred to as the Texas stack because it's so popular here. In a typical configuration, one freeway goes below grade at the bottom level. The frontage roads sit at grade. The crossing freeway is elevated. Then there are the two layers of flyovers. With a minimum vertical clearance of 16 feet or about 5 meters, plus the thickness of each bridge, vehicles on the highest flyovers are often more than a hundred feet or 30 meters above the ground. It’s a nice way to get a good look at the city, even if you only get to enjoy the view for a moment.

The Dallas High Five is probably the most famous interchange in Texas with its cool nickname, but it doesn’t stand alone. There are quite a few five-level stacks around the state and even a couple that qualify as six-level stacks with flyovers connecting to other highways. My friend Brian, better known as the Texas Highway Man, documents a lot of new construction in Texas, including this replacement of an old cloverleaf crossing with a five-level stack in San Antonio. These flyovers will be higher than a twelve-story building when they’re done. The frontage roads for this new interchange use a pretty innovative concept. Four partial roundabouts morph into one funny-shaped roundabout that’s been lovingly nicknamed the “fidget spinner.”

Of course, Texas stacks don’t exist only in the Lone Star State. The Big I is another famous interchange in Albuquerque decorated with a tumbleweed snowman each winter. The Judge Harry Pregerson Interchange in Los Angeles gets its fifth level not for frontage roads but the high occupancy lane. Plus, it has a railroad at the lowest level, which I always appreciate. Not just because I like trains, but also because it’s a reminder that these artfully sculpted ribbons of concrete carefully woven together represent a tremendous investment of public money, our money, into a way of getting people from A to B that has a lot of downsides. Everyone has different thoughts about what a city should look like, but there’s a growing recognition that the way we prioritize motor vehicle traffic in the US may not have been the best path forward. And so, I admit that my ideal city has a lot fewer of these towering interchanges that kind of stand as a testament to a transportation network that doesn’t necessarily reflect our highest values and aspirations. But, I still find them pretty impressive in their own right, and whenever I’m in a new city, I try to plan my driving to hit those tallest ramps at the top of the stack to get a bigger, if momentary, perspective on the built environment. It’s always a nice reminder of our capacity for grand designs and ambitious projects, even if they might not always be the best solutions.

How French Drains Work

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

In February of 2017, one of the largest spillways in the world, the one at Oroville Dam in northern California, was severely damaged during releases from heavy rain. You might remember this. I made a video about it, and then another one about the impressive feat of rebuilding the structure. In the forensic report following the incident, one of the contributing causes identified in the failure was the drainage system below the spillway. Rather than being installed below the concrete, each drain protruded into it, reducing the thickness of the concrete and making it more prone to cracking. But why do you need drains below a spillway in the first place? Put simply: water doesn’t just flow on the surface of earth. It also flows through the soil and rock below it. Water that gets underneath a structure creates pressure that can lift and move it. That’s especially true when the water is flowing. Dam Engineers deal with the challenge in two ways: make concrete structures like spillways massive (so gravity holds them in place) and use drains to relieve that pressure, giving the water a way out.

Even though we depend on it to live, water is the enemy of all kinds of structures. Pressure is far from the only problem it causes. Most of us have come face to face with it in some way or other. Water causes some soils to expand and contract. It freezes, promotes rot, erodes, and corrodes, wreaking all kinds of havoc on the things we build. On the surface, water is relatively easy to manage through channels and curbs and slopes. Below the ground, things get much more challenging. Subsurface drainage is a really interesting challenge, and it applies to everything from simple landscaping at your house to the biggest structures on Earth, and there are a lot of things that can go wrong if they’re not designed correctly. I’m Grady, and this is Practical Engineering. Today, we’re talking about French Drains.

The idea of a subsurface drain is really pretty simple. And I built a model here in the garage to show you how they work. This is just an acrylic box with a hole at the bottom. I filled the box with sand to simulate soil. And I left a small area of gravel in front of the hole. A few strategically-placed dye tablets will help with the visualization. When I turn on the rainfall simulator, watch what happens. Water percolating into the subsurface continues flowing within the sand. It moves toward the gravel, eventually flowing into the holes between the stones and out of the model. (Don’t pay attention to those dye traces on the left. Turns out there was a small leak in the box that was acting as a… secondary outlet to my drain). When the rain is over, the subsurface water continues to flow until the soil is mostly dries out.

This is a very simple model of what’s often referred to as a French drain. It’s not from France but named after an American farmer, lawyer, politician, and inventor Henry French whose 1846 book on Farm Drainage cataloged and described many of the practices being used around the world. Funny enough, he was explicit that he didn’t invent these drains, claiming “no great praise of originality in what is here offered to the public.” Still, I have to admit, after reading his book, I understand why he became the namesake of the drains he made famous. The man had a way with words:

“The art of removing superfluous water from land must be as ancient as the art of cultivation; and from the time when Noah and his family anxiously watched the subsiding of the waters into their appropriate channels to the present, men must have felt the ill effects of too much water, and adopted means, more or less effective, to remove it.”

Well before we worried about draining subsurface water to protect buildings and structures, farmers were doing it in one way or another to keep their fields from sogginess that affects the growth of crops and bogs down agricultural equipment. In fact, “tile drain” is another common term for subsurface drains because clay tiles were used to hold the drains open. And there are plenty of fields still drained using clay tiles today. But French pointed out that rocks sometimes work just as well:

“Providence has so liberally supplied the greater part of New England with stones, that it seems to the most inexperienced person to be a work of supererogation, almost, to manufacture tiles or any other draining material for our farms.”

He was mostly right, and gravel-filled trenches are used all over the place for simple and non-critical applications. The problem with rocks is that they clog up. You can kind of see how sand migrated into the spaces between the gravel in my demo. Since it’s sand, it’s not really a problem, but if this were a finer-grained soil, it would eventually reduce the drain’s ability to transport water, slowing down the drainage process. Tiles provided the benefit of holding open a clear space for water to flow. Over time, perforated or slotted pipes began to replace tiles for use in drains. You’ve probably seen these before; there are a hundred different styles and materials. Rather than flowing in through the joints between the tiles, the water just comes into the holes in the pipe. But which way should the holes face? Turns out it’s a debate as old as pipes themselves among engineers and contractors, and there are strong opinions on both sides.

If the holes are on the top, water has to fill the gravel to the top of the pipe before it can get in and be carried away. If the holes are on the bottom, the flow path isn’t smooth, so the water flows slower and is less likely to wash away any soil or debris that gets inside. From my research, it seems like most of the manufacturers recommend holes down so the gravel envelope doesn’t have to be completely saturated before water can enter the pipe. I think, in practice, it’s really not too important, and actually, a lot of perforated pipes you can buy for drainage have holes all the way around so you don’t even have to think about it. That’s the best kind of decision, in my book. But, if it seems counterintuitive to you to orient the holes downward, I can demonstrate it in my model.

With a pipe in the middle of the gravel layer, I can turn on the rain again. Just like before, water makes its way through the soil toward the drain, and eventually out of the model. Let’s watch that sped up. When the rain is off, the soil continues draining out until it’s no longer saturated. Hopefully it’s clear how beneficial this is. Without that drain, water will eventually dry out of the soil by flowing away or evaporating over time. But getting it out quickly, with a drain, gives it less opportunity to apply pressure to basement walls, freeze against a structure creating long-term movement, swell the soils, or cause rot and corrosion.

I’m using sand in my model to speed up these simulations, so this envelope of small gravel with a pipe inside is working pretty well to keep the soil in place. But, somewhat inconveniently, most places we want to drain aren’t overlain by playground sand. They have finer-grained soils, including silt and clay. These small stones are holding back the sand, but tinier particles would just flow right through the cracks. That can lead to erosion over time as water dislodges and carries soil particles away through the drain. Watch what happens when I try my French Drain model with large stones between the sand and the outlet. You can see the turbid water coming through the drain, indicating that soil particles are making their way out. And if you watch closely on the right side, you can see where they’re coming from. Eventually, enough sand washes through the rocks to create a sinkhole, and the rest of the water bursts through. Made a HECK of a mess (pardon my French drain). I’ve talked about internal erosion and sinkholes in a previous video, so check that one out if you want more details. This erosion can also result in clogging if the soil particles move into the gravel and pipe. In fact, clogging is the biggest problem with subsurface drains, so properly designed ones usually have some kind of filter.

The design you’re probably most familiar with if you’ve seen or installed a french drain yourself uses geotextile fabric. These are permeable sheets that have a wide variety of applications: separating different layers of soil or rock, protecting against erosion, adding reinforcement to backfill, and filtering soil particles out of flowing water. A typical french drain design uses geotextile fabric around the gravel envelope to keep the fines from migrating in. It’s sometimes known as a pipe-within-a-pipe. But geotextile has some limitations. It’s easy to damage during installation. It’s pretty much impossible to repair or replace once it’s in place. And it also gets clogged up. It’s just a thin mesh of fibers, after all, so once soil particles get stuck, they can quickly lead to a decrease in permeability and efficiency. But there is another option for filtration, and it’s most commonly used on dams.

It is hard to overstate the importance of properly filtered drains for dams. If you don’t believe me, take it from the Federal Emergency Management Agency in their 360-page report, Filters for Embankment Dams: Best Practices for Design and Construction. If that’s not enough, try the Bureau of Reclamation in their 400-page report, Drainage for Dams and Associated Structures. A civil engineer could spend an entire career just thinking about subsurface drains, and for good reason. Lots of high-profile dam failures have directly resulted from a lack of drains or ones that weren’t designed well, including the Oroville Spillway incident I mentioned. For embankment dams that are built from compacted soil, any movement of those soil particles can spell demise. And if you think about all the ways that water is terrible for structures, you can imagine how hard it is to design a structure whose literal job is to hold it back. That’s why they use filters of a different design. You can see it in bold right here in this FEMA status report: “It’s the policy of the National Dam Safety Review Board that geotextiles should not be used in locations that are critical to the safety of the dam.”

Instead, they use sand. Just like the gravel in my demonstration lets the water through while holding back the sand particles, sand can hold back smaller particles of silt and clay, acting as a filter. But it’s a little more complicated than that. Every soil consists of a variety of sizes of particles. I can show that pretty easily, again using sand as an example. I have a collection of sieves with different sizes of holes, each one finer than the one above. I put my sand in at the top. Then give it a little shake (a little razzle-dazzle). And when I open it back up, the sand is all sorted out. If you weigh out the fraction that got caught in each sieve and plot that on a graph, you get something like this: a grain size distribution curve, also called the soil’s gradation. Soils can have a wide variety of gradations. And it’s super important to understand in this case, because before you can design a filter, you have to know what you’re trying to filter out. Once you know the base soil’s grain size distribution, there are a number of engineering methods to find a material that will both allow water to flow while still holding the soil back. And in a lot of cases, that just happens to end up being some variation on the sand we’re used to using in concrete and sandboxes and demonstrations about french drains.

Actually, for dams, you often can get either the filtration you need or the capacity to let water through, but not both in the same material. So lots of dams use two-stage filters. The first stage filters the base soil material. The second stage filters the first stage, but lets water flow more freely. And then, you put a perforated pipe in the middle to get the water out of the drain as quickly as possible. So they look basically identical to the demonstration I built: sand, then gravel, then pipe.

As for dealing with the water once it’s out of the ground, there are really just two options. The easiest is to simply release it by gravity to the surface at some low point. But if you don’t have a low point on the surface nearby, the other alternative is to pump it. If you have a basement at your house, there’s a good chance you have a sump, which is just a low spot for drainage to collect, and if you have a sump, it’s a REALLY GOOD idea to have a sump pump, to move that water out and somewhere outside your house.

Of course, there’s a lot more to this. Dams have all kinds of drainage features depending on their design. Concrete dams often include a gallery or tunnel with vertical drains into the foundation. Embankment dams often feature a large internal drain called a chimney filter to keep water moving through cracks or pores from carrying soil along with it. And it’s not just dams. Plenty of structures, like retaining walls, rely on good subsurface drainage for protection against all the bad things that water does, not to mention their widespread use in agriculture. There are lots of interesting designs and maybe even more proprietary products on the market all trying to accomplish those two main tasks: get the water out without getting the soil out too. In the end, it’s all the same engineering whether you’re trying to protect a multi-million dollar structure or just keep your basement dry. I think Mr. French put it best:

“Indeed, the importance of this subject of drainage, seems all at once to have found universal acknowledgement throughout our country, not only from agriculturists, but from philosophers and men of general science.”

I don’t think anyone could reasonably call me a philosopher, but I do love drains, and I hope you agree that, from dams to fields to foundations of houses, they are pretty important.

French drains are one of those topics that be hard to sell in a pitch meeting, right? No studio executive would be like, “Yes, this is a million dollar idea!” But the thing I love about this channel is that it’s created a passionate community around seemingly mundane things like subsurface drains. TV used to be like that too: something for everyone. I loved the old History and Discovery channel shows. Now it’s all converged into reality shows and reruns, and I’ve found that pretty much everything I watch these days is done by passionate independent producers. If you feel the same way, I have a recommendation for you: The Getaway by my friend Sam at Wendover Productions.

It’s a gameshow with a hilarious premise, which is that all of the contestants (who are all big YouTubers, by the way) are snitches, but each one thinks they’re the only one. And it just leads to all these very funny situations where everyone is trying to secretly sabotage the contests. Plus the behind-the-scene cuts to the producers trying to keep all the confusion under control are wonderful. It’s such a great twist on a game show, and it’s one of those creative experiments that only works because it’s independently produced. The chaos of it is what makes it great, and that’s why it’s only available on Nebula.

I talk about Nebula a lot. It’s a streaming service built by and for independent creators, and it’s growing super fast. After the major overhaul of the home page, making it easier to find new stuff to love, we’ve leaned into producing really good original content, like The Getaway; basically allowing your favorite creators to make bigger budget videos without the fear of having it flop on YouTube’s algorithm. That means you get more creative, interesting, and thoughtful videos. My videos go live on Nebula before they come out here, and right now, a subscription is 40% off at the link in the description.

Plus if you already have a subscription, now you gift one to a friend. We have annual gift cards now. Give someone you love a year’s worth of thoughtful videos, podcasts, and classes from their favorite creators. It’s 40 percent off either way at nebula.tv/practicalengineering for yourself or gift.nebula.tv/practical-engineering for a friend. Thank you for watching, and let me know what you think!

When Natural Gas Had No Smell

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

Excitement and hope permeated the crowds gathered in a dusty farm carved from the piney woods in east Texas. The rumor was that Columbus Joiner had struck oil. At 70 years old, Joiner had already won and lost several fortunes in the oil business, but it seemed like, on that October afternoon in 1930, he might just have one more in him. As the congregation grew, Joiner and his crew slowly swabbed the water and mud up and out of the well, relieving the pressure at the bottom. Eventually, the ramshackle derrick began to rumble and shake. Suddenly, the Daisy Bradford No. 3 erupted, showering black oil on the cheering crowd. It was the “discovery” well for what would quickly become the largest and most prolific oil field in the continental United States at the time. Joiner would lose that fortune too, but the boom he kicked off would change the state forever.

The sudden inrush of oil workers and their families inundated the area, including the unincorporated town of New London. New families needed a new school, so one was built in 1932. But no one could have imagined that what created the town in the first place would ultimately rob it of a generation only a few years later in one of the worst school disasters in US history. I’m Grady, and this is Practical Engineering.

With all of the extra population and tax revenue pouring in, New London quickly became one of the wealthiest rural school districts in America, and its new school was correspondingly designed. The building sat on a gentle slope with a footprint shaped like a large capital letter E. Both the north and south wings projected out from the hillside, creating space for classrooms below the main floor. But the main part of the school had a mostly unused crawlspace below its first level. This crawlspace had just two doors into the basement wings and four small vents to the outside for circulation. The school was originally designed to be heated by a large central boiler, but the school board changed the plan during construction to install cheaper, individual gas-fired radiators throughout the building. Gas was supplied by a local utility company for the following years until an opportunity arose in January 1937.

The Parade Gasoline Company had constructed a condensate extraction plant not far from the school. Natural gas is an incredibly important resource today, but at the time, it was mostly considered a byproduct of oil drilling. The supply was just so much higher than the demand because gas was difficult to transport at large scales. Networks of long-distance natural gas pipelines wouldn’t arrive until after World War II. But it was possible to extract liquids from raw natural gas by cooling the vapor. The resulting condensate (often known as “drip gas”) had many uses and could even be used as a low-quality substitute for gasoline in older engines.

Parade’s plant was a simple operation. It accepted raw natural gas from nearby wells, extracted the condensates, and then sent the “residue” gas back to the oil fields in another pipeline, where it was mostly burned off in flares. Since it was already a kind of garbage gas, it was common practice at the time for homes, businesses, and public institutions within easy reach of the residue pipe to tap a line without explicit permission to get the free gas. Since the company was already getting rid of it, they were usually happy to look the other way.

This might seem like an outrageous and dangerous practice today, but it’s not hard to become accustomed to risk, especially when lots of people are doing the same thing and the benefit is so immediate and obvious. In New London, the school board saw an easy opportunity to save about $250 per month on their heating bill (several thousand dollars in today’s money). In January of 1937, the connection was made by two bus drivers, a janitor, and a local welder. A radiator salesman inspected the new line. They installed a regulator to reduce the sometimes erratic pressure from the residue pipe. From there, the gas would flow into the school building’s crawlspace along a 2-inch line suspended by straps. 96 individual connections tied the main gas line to the heaters and burners throughout the school. But the district would never get to see the savings of the supply switch.

On Thursday, March 18, 1937, near the end of the day, the school was full of students and teachers eager to be let out for a long weekend. In the basement wood shop, near the crawlspace, a teacher powered on an electric sander, flipping a rudimentary knife switch to complete the circuit. Unbeknownst to the shop teacher and everybody else in the building, the crawlspace had filled with an explosive mixture of residue gas and air. The spark from the knife switch was all it took to ignite the gas and set off a terrible explosion.

Except for the two doors and four small vents, the crawlspace was practically a sealed chamber of concrete. With nowhere to escape, the pressure of combustion lifted the first floor of the building upward, buckling the walls, and then collapsing the roof into the school. A large chunk of a concrete slab was blown over 200 feet from the building, crushing a car in the nearby parking lot. There were over 500 students, faculty and staff in the building at the time. Chaos ensued as many parents, who had been at the PTA meeting at an adjacent building, reacted to the thunderous sound and ran to the scene. Soon, the school was overwhelmed by residents, oil field workers, and emergency personnel, all doing whatever they could to rescue victims within the collapsed building.

The Texas Inspection Bureau report described the effort involved:

The removal of this debris is worthy of comment since it is probable that in no past disaster of record, has it been approached in speed or efficiency and nowhere but in the oil fields would it have been possible to find the necessary equipment, labor and experience immediately available. In the short space of 17 hours after the work was organized, some 2000 tons of debris were picked up piecemeal and hauled away during an all night rain storm; concrete slabs were broken up, tangled steel cut with torches and the smaller fragments that had to be shoveled, were carried off in small baskets and carefully emptied under flood lights to avoid overlooking a hand or foot or any torn portion of a body.

The story broke across the US, and was reported on by journalists from across the country, including a very young Walter Cronkite, working for the United Press in Dallas at the time. Later in life, he recounted: "I did nothing in my studies nor in my life to prepare me for a story of the magnitude of that New London tragedy, nor has any story since that awful day equaled it."

Mother Frances Hospital had just finished construction in Tyler, about 25 miles east of New London. An all-day ceremony, including a ribbon cutting and banquet, had been planned for Friday, March 19. But, when the staff received word of the explosion that Thursday afternoon they decided to open the hospital early. Medical facilities in the surrounding area were overwhelmed with victims, but also with donations and offers to help. In the end, the explosion killed 270 students and 24 adults.

The governor of Texas declared martial law in New London and appointed a team of officials to form a military court of inquiry, and investigators from the state and federal governments got involved as well. Their first job was to rule out potential causes of the explosion. There were rumors that the blast was a result of dynamite. Apparently, workers had been using it to construct a running track at the nearby athletic field. Eighteen sticks of dynamite were stored below the auditorium on the day of the explosion, but they were found intact afterward.

Investigators were confident that gas caused the explosion but were not yet sure of its source. They tested the school’s sewer system for combustible gases but found none. They also drilled more than 70 holes into the soil below and around the school to determine whether gases were seeping up from the ground. Their detectors found no meaningful traces of hydrocarbons. The only possible source was also the most obvious: the natural gas line traversing the crawlspace.

Looking back, there may very well have been warning signs. Students and teachers had been complaining about headaches in the week leading up to the explosion. The superintendent and several board members had, in fact, met the day of the explosion to search for potential sources of the complaints. A school janitor even searched the crawlspace that morning and, struggling to see, lit a match to get a better view. Either the leak didn’t start until later that day, or it hadn’t reached an explosive mixture yet. The residue gas was tested in a lab after the event and found to explode when mixed with air at proportions between around 4 and 13 percent.

All the investigation reports suggested that switching sources from the utility gas to the extraction plant residue line didn’t directly contribute to the explosion. Even though the gas's chemical composition was different, it wasn’t significantly more explosive than the previous supply. And, the regulator should have been able to manage the less reliable pressures coming from the residue line. The Bureau of Mines report concluded that “no appreciable difficulty should have resulted from its use.” But, all the work involved in changing the gas supply was performed by an unqualified crew rather than professional plumbers.

The Texas Inspection Bureau report noted that the school janitors were often “jacks of all trades and probably masters of none” and that they might not have tested for leaks or tightened joints, or they may have just knocked something loose while they were working.

We’ll never know for sure what caused the leak because all the plumbing was destroyed in the explosion. But, the investigations did cite a bunch of factors that magnified the likelihood and severity of the disaster. The crawlspace was large, spanning the entire length of the school, creating a huge volume for natural gas to accumulate. The robust concrete foundation and limited sources of ventilation left no easy paths for the pressure of combustion to escape, making the explosion extra powerful. And most importantly, natural gas is mostly odorless. There was no way to detect a leak. And even the smell of the less-refined residue gas from the extraction plant would have been nearly impossible to notice in a town surrounded by oil wells where the smell of petroleum was just a constant part of life.

But, ultimately, the inquiry found no grounds to charge anyone involved in the disaster. Although the dangers of natural gas were well-known by that time, safety regulations just hadn’t kept pace with its growing use in buildings. The explosion resulted from a number of profound misjudgments, but no laws were broken. Instead of charges, the court issued recommendations to lawmakers to prevent a similar tragedy in the future, many of which were echoed by the other investigative reports. And several of those proposals would forever affect the fields of engineering, plumbing, petroleum production, and more.

Within a few months, Texas passed two sweeping new laws. First, they joined the growing ranks of states requiring the registration of engineers. At the time, anyone could call themselves an engineer and offer services to the public, regardless of their experience or qualifications. The new law created a regulatory board to oversee the licensing process, helping build public trust in the profession and limiting the possibility of unqualified engineers being involved in decisions that affect public safety. Similar laws and licensing would come for the plumbing industry a decade later.

The second major law that resulted from the explosion established regulations for the odorization of natural gas. Many utilities across the country (and elsewhere in the world) were voluntarily adding chemicals to their gas to make it more detectable by smell, but the new law in Texas created standards that would quickly spread through the rest of the US. Only a few months after that law came into effect, Peerless Manufacturing began shipping an odorizer invented by two Texans, one of whom helped in the rescue effort at New London. The Type “M” Oderizer could precisely dispense liquid odorant into a gas stream, accounting for any changes in flow rate and pressure. The device was designated a Mechanical Engineering Landmark by the American Society of Mechanical Engineers in 1992, and odorizers of its kind have likely saved countless lives by making natural gas leaks easily detectable by smell. Odorization got its federal mandate with the passage of the Natural Gas Pipeline Safety Act of 1968, and is now a widely accepted and implemented safety measure across the world. Natural gas is so closely associated with the smell of ethanethiol (commonly known as ethyl mercaptan) that many never know that it is added artificially.

It’s interesting to look back on an event like this with a modern lens and see just how different our world is now. There are so many parts of the story that would have played out so differently within the current system of building codes and licensure and safety measures that we largely take for granted. And that’s a good thing, right? It means that, whether directly like those new laws, or indirectly in a wide variety of ways, we’ve learned from our mistakes. In the face of such a horrific tragedy, countless lives have been saved and accidents have been averted by our ability to reckon with errors and work hard to correct them. It gives me some comfort at least. Natural gas is one of the most important resources on the planet right now. That’s not to say there are no consequences that come with it, and hopefully, we’ll grow less dependent on it over time, but it’s driven countless innovations that benefit nearly everyone in a huge variety of ways. And so, even if you had never heard of New London, Texas before now, you can feel fairly confident that, all these decades later, you’ve also benefited in some way from the hard lessons learned there in 1937.

Why Bridges Don't Sink

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

The essence of a bridge is not just that it goes over something, but that there’s clear space underneath for a river, railway, or road. Maybe this is already obvious to you, but bridges present a unique structural challenge. In a regular road, the forces are transferred directly into the ground. On a bridge, all those forces on the span get concentrated into the piers or abutments on either side. Because of that, bridge substructures are among the strongest engineered systems on the planet. And yet, bridge foundations are built in some of the least ideal places for heavy loading. Rivers and oceans have soft, mucky soils that can’t hold much weight. Plus, obviously, a lot of them are underwater.

What happens when you overload soil with a weight it can’t handle? In engineering-speak, it’s called a bearing failure, but it’s as simple as stepping in the mud. The foundation just sinks into the ground. But, what if you just keep loading it and causing it to sink deeper and deeper? Congratulations! You just invented one of the most widely used structural members on earth: the humble foundation pile. How do they work, and how can you install them underwater? I’m Grady, and this is Practical Engineering. Today we’re having piles of fun talking about deep foundations.

I did a video all about the different types of foundations used in engineering, but I didn’t go too deep into piles. A pile is a fairly simple structural member, just a long pole driven or drilled into the ground. But, behind that simplicity is a lot of terrifically complex engineering. Volume 1 of the Federal Highway Administration’s manual on the Design and Construction of Driven Pile Foundations is over 500 pages long. There are 11 pages of symbols, 2 pages of acronyms, and you don’t even get to the introduction until page 46. And just a little further than that, you get some history of driven piles. Namely that the history has been lost to time. Humans have been hammering sticks into the ground since way before we knew how to write about it. And that’s pretty much all a driven pile is.

The first piles were made from timber, and wood is still used all these years around the world. Timber piles are cheap, resilient to driving forces, and easy to install. But, wood rots, it has an upper limit on length from the size of the tree, and it’s not that strong compared to the alternatives. Concrete piles solve a lot of those problems. They come in a variety of sizes and shapes, and again, are widely used for deep foundations. One disadvantage of concrete piles is that they have to be pretty big to withstand the force required to drive them into ground. Some concrete piles can be upwards of 30 inches or 75 centimeters wide. It is hard to hit something that big hard enough to drive it downward into soil, and a lot of ground has to either get out of the way or compress in place to make room. Steel piles solve that problem since they can be a lot more slender. Pipe piles are just what they sound like, and the other major alternative is an H-pile. Your guess is as good as mine why the same steel shape is an I-beam but an H-pile. But, no matter the material, all driven piles are installed in basically the same way. 

Newton’s third law applies to piles like everything else. To push one deep into the ground creates an equal and opposite reaction. You would need either an enormous weight to take advantage of gravity or some other strong structure attached to the ground to react against and develop the pushing force required to drive it downward. Instead of those two options, we usually just use a hammer. By dropping a comparatively small weight from a height, we convert the potential energy of the weight at that height into kinetic energy. The force required to stop the hammer as it falls gets transferred into the pile. Hopefully this is intuitive. It’s pretty hard to push a nail into wood, but it’s pretty easy to hammer it in... well, it’s a little bit easier to hammer it in. There are quite a few types of pile drivers, but most of them use a large hammer or vibratory head to create the forces required.

Maybe it goes without saying, but the main goal of a foundation is to not move. When you apply a load, you want it to stay put. Luckily, piles have two ways to do that (at least for vertical loads). The first is end-bearing. The end, or toe, of a pile can be driven down to a layer of strong soil or hard rock, making it able to withstand greater loads. But there’s not always a firm stratum at a reasonable depth below the ground. Quote-unquote “bedrock” is a simple idea, but in practice, geology is more complicated than that. Luckily, piles have a second type of resistance: skin friction, also known as shaft resistance. When you drive a pile, it compacts and densifies the surrounding soil, not only adding strength to the soil itself, but creating friction along the walls of the pile that hold it in place. The deeper you go, the more friction you get. Let me show you what I mean.

I have my own pipe pile in the backyard that I’ve marked with an arbitrary scale. When I drop the hammer at a prescribed height, the pile is driven a certain distance into the ground. Do this enough times, and eventually, you reach a point where the pile kind of stops moving with each successive hammer blow. In technical terms, the pile has reached refusal. I can graph the blow count required to drive the pile to each depth, and you get a pretty nice curve. It’s easy to see how it got stronger against vertical loads the deeper I drove it in. Toward the end, it barely moved with each hit. This is a really nice aspect of driven piles, you install them in a similar way to how they’ll be loaded by the final design. Of course, bridges and buildings don’t hammer on their foundations, but they do impose vertical loads. The tagline of the Pile Driving Contractors Association is “A Driven Pile is a Tested Pile” because, just by installing them, you’ve verified that they can withstand a certain amount of force. After all, you had to overcome that force to get them in the ground. And if you’re not seeing enough resistance, in most cases, you can just keep driving downward until you do!

But piles don’t just resist downward forces. Structures experience loads in other directions too. Buildings have horizontal, or lateral, loads from wind. Bridges see lateral loads from flowing water, and even ice or boats contacting the piers. Both can experience uplift forces that counteract gravity from floods due to buoyancy or strong winds. If you’ve ever hammered in a tent stake, you know that piles can withstand loading from all kinds of directions. And then there’s scour. The soil along a bridge might look like this right after the bridge is built, but after a few floods, it can look completely different. Engineers have to try and predict how the soil around a bridge will scour over time, from natural changes in the streambed and those created by the bridge itself. Then they make sure to design foundations that can accommodate those changes and stay strong over the long term. This is why bridge foundations sometimes look kind of funny. Loads transfer from the superstructure down into the piers. The piers sit on a pile cap that transfers and distributes loads into the piles themselves. Those piles can be vertical, but if the engineer is expecting serious lateral loads, some of the piles are often inclined, also called battered piles. Inclined piles take better advantage of the shaft resistance to make the foundation stronger against horizontal loads.

As important and beneficial as they are, driven piles have some limitations too. For one, they’re noisy and disruptive to install. Just last year, I had two friends on separate trips to Seattle who sent me a video of the exact same pile-driving operation. It’s good to have friends who know how much you like construction. But my point is, this type of construction is pretty much impossible to ignore. In dense urban areas, most people are just not willing to put up with the constant banging. Plus the vibrations from installing them can disrupt surrounding infrastructure. Pile driving is crude; in many cases, the piles aren’t designed to withstand the forces of the structure they’ll support but rather the forces they’ll have to experience during installation which are much higher. They can’t easily go through hard geological layers, cobbles, or boulders; they can wander off path, since you can’t really see where you’re going, and they can cause the ground to heave because you’re not removing any soil while you force them into the subsurface. The second major category of piles solves a lot of these problems.

And, wouldn’t you know it? There’s an FHWA manual that has all the juicy details - Drilled Shafts: Construction Procedures and Design Methods. This one a whopping 747 pages long. A drilled shaft is also exactly what it sounds like. The basic process is pretty simple. Drill a long hole into the ground. Place reinforcing steel in the hole. Then fill the whole thing with concrete. But, bridge piers are often, as you probably know, installed underwater. Pouring concrete underwater is a little tricky. Imagine trying to pour a smoothie at the bottom of a pool! Let me show you what I mean.

This is my garage-special bridge foundation simulator. It has transparent soil in the form of superabsorbent polymer beads… and you know we have to add some blue water too. You can probably imagine how easy it might be to drill a hole in this soil. It’s just going to collapse in on itself. We need a way to keep the hole open so the rebar and concrete can be installed. So, drilled shafts installed in soft soils or wet conditions usually rely on a casing to support the walls. Installing a casing usually happens while the hole is drilled, following the auger downward. I tried that myself, but I only have two hands, and it was pretty unwieldy. So, just for the sake of the demo, I’m advancing the casing into the soil ahead of time. Now I can drill out the soil to open the shaft. And now I’m realizing the limitations of my soil simulant. It was still pretty hard to do, even with the casing in place. It took a few tries, but I managed to get most of it out.

So now I have an open hole, but it’s still full of water. Even if your casing runs above the water surface, and you try to pump it out, you can still have water leaking in from the bottom. In ideal conditions, you can get a nice seal between the bottom of the casing and the soil, but even then, it’s pretty hard to keep water out of the hole, and luckily it doesn’t matter.

Instead of concrete, I’m using bentonite clay as a substitute. It’s got a similar density, and it’s perfect for this demo because you can push it through a small tube… if you get the proportions right. Ask me how I know. This is me pondering the life decisions that led up to me holding a gigantic syringe full of bentonite slurry in my garage. You can’t just drop this stuff through the water. It mixes and dilutes, just turning into a mess. Same is true for concrete. The ratio of water to cement in a concrete mix is essential to its strength and performance, so you can’t do anything that would add water to the mix. The trick is a little device called a tremie. Even though it has a funny name, it’s nothing more than a pipe that runs to the bottom of the hole. As long as you keep the end of the tremie below the surface of the concrete that you’re pumping in, or concrete simulant in my case, there’s no chance for it to mix with the water and dilute. I’m just pushing the clay into the casing with a big syringe, making sure to keep the end of the tube buried. Because concrete is a lot more dense than water, it just displaces it upward, out of the hole. 

In underwater installations, the casing is often left in place. One advantage is that you can build a floating pile cap. Instead of building a big cofferdam and drying out the work area to construct a big concrete structure, sometimes you can raise the pile cap into or above the water surface, reducing the complexity of its construction. These “high rise” pile caps are used a lot in offshore wind turbines. But, not all casings are permanent.

In some situations, it’s possible to pull the casing once the hole is full of concrete, saving the sometimes enormous cost of each gigantic steel tube. I tried to show this in my demo. It’s not beautiful, but it did work. Again, the concrete is dense, so the pressure it exerts on the walls of the hole is enough to keep the soil from collapsing. And because drilled shafts can be much larger than driven piles, sometimes you don’t even need a group of them. Lots of structures, including wind turbines, highway signs, and more, are built on mono-pile foundations. Just a single drilled shaft deep in the ground, eliminating the need for a pile cap altogether. Another interesting aspect of drilled shafts is that you can ream out the bottom, creating an enlarged base that increases the surface area at the toe. This helps reduce a pile’s tendency to sink, and it can help with uplift resistance too.

Driven piles and drilled shafts are far from the only types of deep foundation systems. There are tons of variations on the idea that have been developed over the years to solve specific challenges: Continuous flight auger piles do the drilling and concreting in essentially one step, using a hollow-stem auger to fill the hole as it’s removed. Then reinforcement is lowered into the wet concrete. You can fill a hole with compacted aggregate instead of concrete, called a stone column or tradename Geopier if you’re only worried about compressive loads. Helical or screw piles twist into the ground, instead of being hammered, reducing vibrations and disturbance. Micropiles are like tiny drilled shafts used when there are access restrictions or geologic constraints. And of course, there are sheet piles that aren’t really used for foundations but are driven piles meant to create a wall or barrier. Let me know if I forgot to mention your favorite flavor of pile.

Even though they’re usually much stronger than shallow foundations, piles can and do fail. We’ve talked about San Francisco’s famous Millennium Tower in a previous video. That’s a skyscraper on a pile foundation that sank into the ground, causing the building to tilt. It seems like they mostly have it fixed now, but it’s still in the news every so often, so only time will tell. In 2004, a bridge pier on the Lee Roy Selmon Expressway in Tampa, Florida sank 11 feet (more than 3 meters) while it was still under construction because of the complicated geology. It cost 90 million dollars to fix and delayed the project’s completion by a year. These case studies highlight the complexity of geotechnical engineering when we ask the ground to hold up heavier and heavier loads. The science and technology that goes into designing deep foundations are enough to spend an entire career studying, but hopefully, this video gives you a little insight into how they work.

This Bridge Should Have Been Closed Years Before It Collapsed

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

On January 28, 2022, about an hour before dawn, the four-lane Fern Hollow Bridge in Pittsburgh, Pennsylvania, collapsed without warning. Five vehicles, including an articulating bus, fell with the bridge, and another car drove off the abutment after the collapse, not realizing the bridge was gone. Although there were no fatalities, several people in the vehicles were seriously injured. And this bridge had been listed as being in ‘poor condition’ for over a decade. Anyone who walked by the supports in the park below would have had reason to question its safety, as seen in this sadly prophetic tweet from 2018, four years before the collapse. So, why was it left open in this state?

While some initial findings were released earlier, the official NTSB report was delivered to the public this year in February, more than two years after the collapse and over a year after the replacement bridge was built and open. The report included the use of some really cool technology for the forensic study of structures and revealed systemic flaws in how we inspect, analyze, and prioritize repairs for bridges. In fact, the NTSB issued recommendations to basically every organization involved in this bridge from the bottom to the top, and they referenced that tweet that got so much attention. This is a crazy case study of how common sense can fall through the cracks of strained budgets and rigid oversight from federal, state, and city staff. And the lessons that came out of it aren’t just relevant to people who work on bridges. It's a story of how numerous small mistakes by individuals can collectively lead to a tragedy. I’m Grady, and this is Practical Engineering.

The Fern Hollow Bridge was opened in 1973, replacing an aging arch bridge built at the beginning of the 20th century, crossing Frick Park and Fern Hollow Creek. The 1973 bridge used a K-Frame design with continuous steel girders supporting the deck, each with two angled steel legs supported on concrete thrust blocks. At the time, the bridge’s design was celebrated because it blended well into its settings. It was featured on the cover of the 1974 edition of Prize Bridges by the American Institute of Steel Construction. And a big part of why it looked so harmonious with the park below was the type of steel used for the design.

The Fern Hollow bridge was fabricated from weathering steel, sometimes referred to by its genericized trademark, COR-TEN. And it was developed commercially right there in Pittsburgh by US Steel in the 1930s. Unlike most types of steel, whose rust can continually flake away, exposing more material to corrosion, the oxide layer on weathering steel (called its patina) is more stable and protective, shielding the underlying material against exposure to the elements. In that way, weathering steel acts kind of like aluminum, which protects itself from corrosion in a similar way. And, architecturally, it’s a nice material. You get this rustic look that can give structures a more comfortable and less obtrusive appearance. But weathering steel has a limitation: For a stable patina to form, the material has to stay mostly dry. If water pools or the steel is kept damp for extended periods, that patina of rust isn’t enough to protect the underlying steel, and it will continue to corrode. Corrosion of structural steel is called section loss by engineers, and it's easy to see why in these photos of the bridge.

What’s more alarming than what’s in those photos is where they came from: inspection reports of the bridge. It’s not that this deterioration somehow went unnoticed. The bridge supports were clearly visible from a popular walking trail. Between 2005 and 2021, this bridge was inspected a total of 14 times! In those reports, you get a clear and vivid story of its decline. First were the drainage problems. You can see in these images from multiple previous bridge inspections there were drainage grates on the roadway that were 100 percent clogged. Rainwater, and even worse, salty meltwater from the frequent snow that Pittsburgh sees each winter couldn’t follow the prescribed drainage paths off the deck and into the creek below. Instead, that water would leak through the bridge deck, dribbling over the structural steel, and pooling in portions of the legs where webbing and tie-plates could catch puddles of water, leaves, and debris.

The City was aware of the section loss due to these drainage problems for many years before the collapse. Nearly every inspection report noted problems with the drains and the accelerating corrosion that was resulting. In fact, in 2009, the cross-braces connecting each pair of legs were found to be failing, and steel cables were installed as a temporary retrofit until the framing could be replaced. These cables were lightly tensioned to add structural integrity to the bridge but were never meant to be a permanent solution.

You can see the ends of two of these cables in this now-infamous tweet from 2018. Of course, the more notable feature of this image is the fully separated steel cross-brace! That photo was taken about nine years after the temporary cables were installed. And they remained in place for the rest of the bridge's life, which ended up being only four more years. But the cross-bracing between the legs wasn’t the only place where corrosion was an issue. The legs themselves were also fabricated from weathering steel, and that steel was suffering, too. Since 2005, inspection reports marked them in fair to poor condition with areas where the steel had completely rusted through. By 2019, all four legs were given the worst assessment possible for an individual bridge element. According to the code, that should trigger a structural review to check whether the integrity is affected by the poor condition of a structural element, but it was never done. And that’s not all.

An important part of inspecting steel bridges is identifying members that are “fracture critical.” That’s engineering jargon, but the idea is actually pretty straightforward. It’s any piece of steel under tension that lacks redundancy. If it breaks, the bridge collapses. And these types of members get special attention because of their importance, so inspection teams identify them ahead of time to make sure they get a proper look. This drawing shows in green which elements of the bridge were considered to be fracture-critical. Notice that while the girders crossing the span are identified, the legs are not. And, at first glance, that might match your intuitions. Bridge piers, columns, and vertical supports usually don’t experience tension forces, right? They’re in compression. So if there’s a crack or break, the forces just squeeze it together, generally not that big a deal. But K-frame bridges are different. By splaying out the legs, there are loading conditions that can apply bending forces, putting part of each beam in tension. And, this particular bridge had another feature that was absolutely essential to its performance.

Each leg of the bridge was essentially an I-beam: a central web with a top and bottom flange. To simplify the foundation design, each leg had a “shoe”: a tapered end that would connect to the concrete block. It’s clear that the narrower section would have less strength, so larger stiffeners were added to each shoe to strengthen that portion of the leg. These are just steel plates welded to the web and flanges to increase the leg’s rigidity. And I built a little cardboard model to clarify this point. This particular stiffener, called the transverse tie plate, bridges the two flanges right where they taper down. And if I apply a compressive force on the leg, it’s easy to see what kind of force that tie plate experiences as a result. It’s tension. These tie plates were fracture-critical members of the bridge, but never identified as such, and so, even though it was clear they were deteriorating quickly over time, the inspectors never elevated the concerns to a priority level that might have spurred a more immediate response. But there was another opportunity to catch the problem.

In 2013, an inspector was concerned enough about the bridge's safety to recommend that it be reviewed for a load rating. Most bridges are designed to allow any legal load to pass over, but sometimes, either because of an old design or poor structural conditions, it's necessary to limit the weight of vehicles allowed. Engineers analyzed the bridge in 2014 and decided it could only handle 26 tons per vehicle, just over half of its previous rating. When NTSB reviewed that decision in hindsight, they found some pretty serious errors.

For one, the load rating for the bridge was based on a layer of about 3 inches of asphalt paving on top of the concrete road deck. In reality, the bridge had about double that amount. The City’s records of the removal and addition of pavement were poorly kept, so the engineering firm doing the load rating had no idea there was so much extra weight. For two, the engineers didn’t fully account for all the corrosion on the legs. This was partly because inspectors hadn’t cleaned off the rust to measure and report the actual thickness of the remaining steel. Even so, the engineers used a method that distributed the section loss from corrosion evenly along the entire leg, instead of applying it where it actually was, at the shoes and tie plates. That’s a pretty commonly-used simplification that usually generates conservative results (since the worst corrosion is rarely located at the most critical part of a structure), but again, that wasn’t the case for the Fern Hollow Bridge, and no one had recognized how important those tie plates really were.

And for three, those engineers made an incorrect assumption about how the bridge’s legs would buckle. A structural member under compression will buckle at different loads depending on how much restraint it has at the ends. This is something you learn in year one of engineering. If you keep a column from rotating at its ends, you substantially increase the amount of force it can withstand, and with the original cross-bracing between the legs, that would have been the case. But I’m sure I don’t have to tell you that steel cables don’t provide the same support as rigid members. Again, this is engineering 101: “You can’t push a rope.” The cables provided some restraint, but not in the same way that the original bracing could, so the load rating applied to the bridge ended up significantly overestimating its actual capacity. In fact, when NTSB updated the load rating with these errors fixed, they found that the bridge should have been limited to 3 tons, basically nothing for a bridge. In effect, this load rating exercise should have closed the bridge to traffic entirely. These structural issues were exactly what the process was meant to identify. But instead, the bridge stayed open to everyone except the largest of trucks,and here’s what happened, courtesy of NTSB’s animation team.

The transverse tie plate on the southwest leg, weakened by corrosion, failed first under tension, separating the flanges on the leg, and ultimately causing it to buckle. With no redundancy in the supports, the loads had nowhere to go, and so the rest of the bridge fell into the valley below. The articulated bus had both rear-facing and forward-facing cameras on it, which captured some truly harrowing footage of the event. Looking at the rear-facing camera, you can see the western portion of the bridge begin to fail. Keep an eye on the railing, and you can see the exact moment it starts. Once it started, there was no stopping it. Within two seconds, the front-facing camera shows that the collapse had propagated all the way to the eastern abutment, and the bridge fully failed.

Thankfully, the collapse happened during particularly inclement weather. School delays and generally poor conditions meant that traffic was lighter than normal, and the weather also likely kept folks away from the trail underneath. On a fair weather day during rush hour, it wouldn’t be uncommon for the eastbound lanes of the bridge to be fully occupied by heavy traffic, and the trail underneath to be populated with dog walkers, families, or even classes on field trips from nearby schools. The bridge also carried a large gas line, which was severed during the collapse, leading to a major leak and some evacuations, but they got it shut off in time. It really is remarkable that nobody was killed in a failure of this magnitude. But there were still multiple victims needlessly affected for the rest of their lives by the collapse, not to mention the overall erosion of trust in the organizations and engineering systems meant to keep the public safe.

By pure happenstance, President Biden was set to arrive in Pittsburgh on the very day of the collapse to speak in support of the Infrastructure Investment and Jobs Act, so he rearranged his trip to make some remarks at the site. One of the entities supported by the act is the National Highway Performance Program, which ultimately funded the replacement of the collapsed bridge. But at that time, no one understood the full scope of neglect and bad assumptions that led to the gradual, and then sudden, demise of the bridge. In those two years after Fern Hollow Bridge fell, the NTSB conducted numerous interviews with those involved, from the paving contractors to the inspectors to the bridge rating engineers. They performed 3D laser scanning of the bridge components to compare them to as-built conditions. They tested sections of steel for strength and durability. They reviewed all the previous records of the design, construction, and repairs. And they built a detailed finite-element model of the bridge to confirm that the gradual corrosion of one small structural element, the transverse tie plate on the southwest leg, initiated the collapse. And then they documented why it got to that point: the City of Pittsburgh just didn’t fix it.

This figure in the NTSB report tells the story as clearly as I think is possible. From 2005 onward, recommendations from inspection reports to repair parts of the bridge didn’t fall off the list. They just kept being repeated by each new inspection, year after year. Since 2007, every single inspection report that included recommendations said to repair the stiffener plates in the legs that were heavily corroded. These were Priority 2 recommendations, which means the timeframe to complete them is before the next inspection. But it was never done. They didn’t fix the drainage problems that were accelerating the corrosion, they didn’t apply the protective coatings that might have slowed it down, and they never analyzed the capacity of the legs after they were rated the worst possible condition a structural element can have. And, apparently, there was no mechanism to follow up on those recommendations by the state charged with overseeing the bridge inspection program. When there was finally a chance to recognize how deficient the structure really was through a new load rating, the engineers made a few bad assumptions, missed it by a mile, and left the bridge open, a ticking time bomb, for years. (Years, by the way, in which the City still didn’t address the recommendations from inspection reports.)

Due to the nature of the emergency, the site was cleaned up quickly, with a huge crane brought in to remove the bus, and building of the replacement bridge happened on a fast-track schedule. The new bridge uses a more conventional design: pre-stressed concrete girders on vertical piers. The formed stone texture on the columns certainly doesn’t blend into the park as well as the graceful and patinaed K-frame once did, but I doubt anyone involved in the project could stomach another structure built from weathering steel, given the circumstances. The new bridge might not win any awards for beauty, but it could win some for speed. After a colossal effort from the design and construction teams, it opened to limited traffic less than a year after the collapse in December of 2022, and by July the next year, it was fully operational. It would be almost a year later before the NTSB concluded why the previous bridge collapsed, not for the purpose of blame, but to issue recommendations to prevent something like this from recurring in the future. And unlike the recommendations from those inspection reports, many of the NTSB recommendations have already been put into practice.

They published a special report on weathering steel bridges to highlight the specific challenges of keeping them in good condition, and they identified several similar bridges that needed a closer look. The City of Pittsburgh quadrupled their spending on inspection, maintenance, and repairs. And they redid the load ratings on all the bridges they owned, resulting in one bridge being closed until it can be rehabilitated and two more having lane restrictions imposed. PennDOT released a technical bulletin to shore up their bridge inspection program. And even the federal government has implemented a process to identify, prioritize, and follow up on recommendations related to weathering steel bridges.

But as I read through those recommendations from the NTSB, one thing struck me: They all add up to more paperwork. And this is just my own personal opinion as someone who did this kind of work for nearly a decade (not on bridges, but other large infrastructure projects). We have these national inspection standards and procedures - huge documents that you could spend an entire career understanding. We have the Federal Highway Administration overseeing the program, state DOTs charged with carrying it out, individual bridge owners, like Pittsburgh, responsible for inspecting their own bridges, and then the private contractors who do most of the actual work. We have this huge machine with thousands of people, federal, state, and local involvement, and millions of dollars meant to keep the traveling public safe. And what did it do for us when a photo like this is all it would take any reasonable person to say: “This bridge needs to be fixed”?

This big machine, in a lot of cases, has all the work sectionalized out. The inspectors see the bridge up close, but they have no autonomy to do anything but document and give recommendations. It’s not their bridge. But the owners who are charged with the safety of their bridges just see a piece of paper. Each recommendation is just another one on the list of sometimes hundreds of action items, to sort and prioritize and try to find the budget to cover. All the NTSB recommendations feel a little bit like bandaids if the real source of the problem was that no one person in this whole machine had both a full appreciation of the bridge's condition and the authority to do something about it. And if that’s the case, I’m not sure any of those recommendations really fixes that problem. I don’t know what the answer is, and I’m still wrestling with trying to understand how something like this can fall through the cracks of the enormous system we’ve built for the sole purpose of trying to prevent it.

If you take a walk on the Tranquil Trail through Frick Park beside Fern Hollow Creek and look carefully, you can still see remnants of the old bridge. And I’m glad the City left them, because they’re a good reminder that design and construction are two parts of a three-part system for keeping people safe. Maintaining infrastructure is thankless work. Don’t get me wrong, it can be a really rewarding career. Inspections involve a lot of time out in the field seeing cool structures up close. And repair projects are often interesting challenges for contractors. But they’re not rewarding in the same way that designing and building new stuff can be. No one holds a press conference and cuts a big ribbon at the end of a bridge inspection or structural retrofit. Building a new structure is not just an achievement in its own right; it’s a commitment to take good care of it for its entire design life, and then to rehabilitate, or replace, or even close it when it’s no longer safe for the public. And I think this is the perfect case study to show that there’s more we could do to encourage and celebrate that kind of work as well.

The Most Confusing Part of the Power Grid

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

In March of 1989, Earth experienced one of its strongest geomagnetic storms in modern history. It all started when scientists observed a cluster of sunspots—active, magnetic areas on the sun's surface—emerging on its horizon. Over the next few days, the sun slowly rotated until the region began to point directly at Earth. Just as it did, two solar flares erupted from the sunspots. Accompanying the flares were coronal mass ejections: huge bursts of solar wind, essentially charged particles from the sun. The coronal mass ejections eventually crashed into the earth’s magnetic field, causing it to squish and compress and ultimately induce electric currents at the surface.

In Quebec, Canada, the rapid changes in magnetic fields would have mostly gone unnoticed by people, but they didn’t go unnoticed by the power grid. The region’s unique geology, a shield of hard rock that is a poor conductor of electricity, kept these induced currents from dissipating into the ground. So they found another path: the electrical transmission lines. The geomagnetic storm ended up blacking out a large part of the Hydro-Quebec power grid for nine hours. And the first domino of the collapse (or rather the first seven) were pieces of equipment known as static compensators. But to understand how static compensators work and why a solar flare could trip them offline, you kind of have to start with the basics.

You might know that most parts of all modern power grids use alternating current or AC. The voltage and current on the lines slosh back and forth, 50 or 60 cycles per second, depending on where you live. If you love power electronics, that low, dull, AC hum might be music to your ears. But if this is kind of new to you, alternating current can be a little bit mysterious. What’s even weirder is that, even though the current constantly alternates its polarity, electrical power only moves in one direction… under ideal conditions. And geomagnetic storms aren’t the only thing that can make the grid behave in funny ways. There are devices even in your own home that force the grid to produce power and move it through the system, even though they aren’t even consuming it. Let’s go out to the garage, and I can show you what I mean. I’m Grady, and this is Practical Engineering. In today’s episode, we’re talking about how power actually flows on the grid.

I’ve built a model power grid here in the shop. (Not the first time I’ve said that, and it probably won’t be the last.) I’ll keep it simple at first and build up the complexity as I explain these concepts. And Zap McBody slam is back in the shop to help out. My grid has one power source, right now just a battery, a transmission line to carry the power, and a load (in this case, an incandescent light bulb). It’s probably not the most interesting circuit you’ve ever seen. But like I said, understanding the basics of power flow is essential to understanding the more complicated, and I think, the more interesting aspects of how it works on a huge scale. So, here’s a one-minute refresher on electrical circuits:

There are really only four numbers that matter the most in a circuit. First is voltage, the difference in electric potential between two locations. In the classic pipe analogy, voltage is the pressure that drives water to flow from one side to the other. In my circuit, the battery is supplying about 10 volts across the bulb. Next is current, the flow of electric charge. In the pipe analogy, this is the flow rate of the water. In my circuit, I can measure the current as 1.2 amps. Third is resistance, the opposition to the flow of current. It’s the size of the pipe. Incandescent bulbs actually change their resistance depending on voltage, so it can’t be measured directly with a meter. That’s okay, though, because all three of these values are related to each other. That relationship, called Ohm’s law, is about as simple as it gets. Voltage is equal to current times resistance. If you know two, you can find the other one with some basic math. For example, 10.1 volts flowing at 1.2 amps means the resistance of my lightbulb is around 8 ohms. The final number we care about is power, the transfer of energy. Power does the actual work, in this case, creating light and heat in the bulb. Calculating power is as simple as multiplying the voltage and the current together. 10 volts times 1.2 amps tells me that this bulb is dissipating 12 watts. That’s electrical engineering in a nutshell, and it’s relatively straightforward for a circuit like this that uses direct current or DC, because none of our important numbers change. But, as I mentioned, that’s not true on the grid.

Let me swap out the battery with a transformer plugged into an outlet and see what happens. At first glance, there’s no change. The bulb is still lighting, just like it did with the battery. I can measure the voltage by switching my meter to AC: 8.4 volts, not too far from the DC circuit. I can measure the current with this clamp over meter: 1.2 amps, same as before. But those are just simplifications of what’s really happening on the lines. To see that, we need a different piece of equipment. This oscilloscope measures voltage over time and plots it as a graph on the screen. And I can insert a resistor into the circuit and use a second probe to plot the voltage across that resistor as a simple way of measuring current. “So the yellow will be the voltage, and the green will be the current.” You can see that neither the voltage nor the current are constant… unless you trip over all the cords. They’re switching directions over and over again. This might not be too surprising to you yet, but watch what happens if I switch out the lightbulb with a different kind of load. Let’s try a capacitor.

This is a device that stores energy in an electric field between two plates. You see them everywhere in electrical circuits, and they do a funny thing on the grid. When I insert the capacitor to my circuit, the graph of voltage and current looks different because they’re no longer in phase. “Hey… that worked perfectly.” The current waveform is leading the voltage; the current peaks happen before the voltage ones. That’s because the current has to flow into the capacitor before the voltage between the plates rises. It takes time for the capacitor to charge and discharge, which results in a delayed response in the voltage. Now, let’s try another type of load called an inductor.

An inductor is basically a coil of wire. Like a capacitor, an inductor stores energy, but instead of an electric field, it stores that energy in a magnetic field. This is just an electromagnet like you might see in a scrapyard. If I swap in an air-core inductor, you can hear the screwdriver rattle against the table as the magnetic field rapidly changes direction. And, we get the opposite effect of the capacitor when the inductor is inserted into the circuit. This time, the current waveform is lagging the voltage. That magnetic field resists changes in current, so it creates a delay, this time in the current waveform. I can even vary this inductance and thus the lag in the current by moving this ferrite rod in and out of the core. All this is interesting on its own, but these little shifts in a graph have serious implications on the grid, and have even resulted in numerous blackouts across the world. Here’s why:

Remember, that the power consumed by an electrical load is just the voltage multiplied by the current. We can do that for any point in time across this graph. For a purely resistive load, like the lightbulb, the current and voltage are in sync. When one is positive, the other is positive. When one is negative, the other is too. So when you multiply them together at any point along the graph, you always get a positive number. The power fluctuates, but it’s always moving in one direction. For a reactive load (the term used for inductors and capacitors), that’s no longer the case. There are times in the cycle when the current and voltage are opposite polarity, meaning, instead of being consumed, power is actually flowing out of the load. In fact, for a purely capacitive or inductive load, there’s no power consumption at all - no work being done. It’s just stored in a magnetic or electric field and returned. But there’s still current flowing, and that matters.

Of course, most things connected to the power grid aren’t purely reactive. But lots of devices that we plug in have some amount of inductance. Look around your home for any big motors. Air conditioners, refrigerators, washers, dryers, large power tools, and more primarily use induction motors because they’re cheap, simple, and last a long time. And inside an induction motor is a series of wire coils used to create magnetic fields that spin the rotor, just like the inductor I used in the demo. Part of the power that flows into those coils just gets sent back out onto the grid. You might be thinking, “So What? Nothing wrong with storing a little bit of energy, as long as I give it back in less than a sixtieth of a second afterwards.” But, the grid still had to produce that power, and more importantly, deliver that power to your home and carry it back away.

The electric meter at your home, in most cases, only tracks the power you actually consume. So, you don’t pay for the reactive power that flows into your devices and back out again. But that doesn’t mean it doesn’t come at a cost. It still has to flow through the power network, where some gets lost as heat from resistance in the lines. So, the generators have to make, and the transmission lines have to move, more power, in some cases a lot more power, than is actually being used in the system. Reactive power can make up a big part of the total load on the system, even though it’s not doing any work. Just having the infrastructure in place to handle it is also costly. The conductors, transformers, and generators on the grid have to be sized for the total current that needs to move through the system, not just the current that does work. And that stuff is expensive. It’s like if you were a photographer and bought a bunch of props for a shoot from a company with a generous return policy. After you take your photos, you return everything back to the store. Those props were useful, even necessary to you, but only for a period of time. And there was a real cost to warehousing, transporting, and restocking them, even if you didn’t bear it. Imagine if there were a hundred photographers that did the same thing. It wouldn’t be long before such a store wasn’t very profitable. But unlike at your home, where the utility is generous in their return policy, lots of industrial and commercial customers do get charged for reactive power that uses up capacity on the grid without doing any real work.

Even though the oscilloscope graphs just show a shift between the two waveforms, with some clever math, you can actually separate the real power actually being used from the reactive power that oscillates on the grid into two parts, and treat them like they flow through the grid independently. I’m going to do my best to avoid that math here partly because it involves imaginary numbers but mostly because it’s not needed to understand the practical impacts. (This is already a lot to wrap your head around.) But out of that math comes this visualization: the power triangle. This leg is the real power that actually gets consumed, measured in watts or kilowatts that you’re probably used to. This leg is the reactive power that is returned instead of used, measured in volt-amps-reactive or VAR. By convention, we usually say that inductive devices “consume” reactive power and capacitive devices “supply” it. The hypotenuse of the triangle is the apparent power, the total amount of power that flows through the grid, measured in volt-amps. If you divide the real power by the apparent power, you get this ratio, called the power factor, a number that will be important in a minute.

Take a look at the distribution transformer that connects your home to the grid, and you might see a rating on the side. That number is not in watts or kilowatts like what you might see on a toaster or microwave, but in kilovolt-amps because it includes the flow of real and reactive power. Large users of electricity, like factories and refineries, usually have a low power factor because they use lots of big induction motors. They need comparatively robust and high-capacity connections to the grid, even if they actually consume only a portion of the energy that flows through. So the electric utility installs a meter that can track power factor, or they just come out every once in a while to measure it, so they can put it on the bill. Instead of free returns on reactive power, like we usually get at our homes, those customers have to pay a rental charge on the power they store, even though it goes right back out. But, it’s not just a matter of keeping track of costs. The stability of the entire grid depends on managing the flow of reactive power.

If you’ve watched some of my other videos on the power grid, you know how important it is to closely match power generation with demands as they go up and down. If not managed carefully, the frequency of the grid, which needs to stay within a very tight tolerance, can deviate. And if it goes too far, the whole thing can collapse. That’s what almost happened to Texas during the winter storm in 2021. But, it’s possible for the grid to collapse even if there’s enough generation to meet the demand because you still have to move that power to where it’s needed over transmission lines. Engineers use a PV curve to keep an eye on this challenge. It relates the power flowing to a load on the system to the voltage it sees. As you would expect, the more power that flows, the more the voltage drops, since more power is lost on the transmission lines on the way to the load. It’s the same reason the lights dim in old houses when the air conditioner kicks on: current goes up, voltage goes down. If you combine Ohm’s law and the power equation, you can see that the power lost on a transmission line is related to the current squared. Double the amps; quadruple the power lost as heat. But the further along this curve that the system operates, the more dangerous things get. There is a point, the nose of the curve, beyond which greater demand on the system actually reduces the amount of power that can be delivered, all while the voltage continues dropping. The generators may have the capacity to supply more power, but it can’t reach the load because of the limitations of the system. Operating below the nose is unstable because generators lose control of their speed, like a rubber tire losing its grip on a road.

Infrastructure is expensive, and building new power plants and transmission lines always comes with legal and environmental challenges too, so we’re often forced to use the grid to the very limits of its capacity. But, grid managers need to make sure to operate with enough margin that any contingency, like a generator going offline or a transmission line faulting, doesn’t push the system over the electrical cliff. Here’s where power factor comes in. Loads with lower power factor shift the nose of the PV curve down and to the left. That reduces the margin and lowers the voltages in the system for a given power demand, making a voltage collapse more likely if some part of the system goes down. So we use several ways to supply reactive power to provide voltage support and shift the curve back up.

Power plants can adjust their operating parameters to supply reactive power, but transmission lines have their own inductance that consumes the reactive power as it travels through. So, it is usually more efficient to address the problem on the load side, and there are several types of infrastructure that make this possible. Synchronous condensers are big motors that aren’t attached to anything. Instead of converting electrical power to mechanical power, they basically spin freely, but with some clever circuitry, they can generate or absorb reactive power from the grid. They can also help stabilize fluctuations in the grid with the inertia of their heavy rotating mass, something that is becoming increasingly important as we transition more to renewable sources that use inverters to connect to the network.

Another option, and one you’re more likely to spot, are shunt capacitor banks connected across the lines. Sometimes you can see them in substations, but many capacitor banks are installed on poles out in the open for anyone to have a look. Like the capacitor in my demo, they increase the power factor and boost the PV curve up. That can actually become a problem during off-peak hours by boosting the voltage above where it should be, so many capacitor banks are switched on or off depending on system conditions. Looking back at the PV curve, you can see how leaving the capacitors off during periods of low demand keeps voltage within limits, and having them on when demand is high provides more margin and more voltage. Some run on timers to come on during the highest demands of the day, and many are operated at a utility’s discretion to accommodate the varying conditions on the grid. They’re usually either all the way on or all the way off, so deciding when to throw the switch is an important one.

A third option for reactive power supply, called a static VAR compensator or SVC, addresses that challenge. These use electronics to rapidly switch inductors and capacitors on or off to constantly adjust to conditions in the system. That switching happens automatically and quickly, making them much better suited to the dynamic changes that happen on the grid.

That’s why Hydro-Quebec had them installed on their system in 1989. The long transmission lines between the hydroelectric power plants in the north and the load centers, like Montreal, in the south require careful control of the voltage to avoid instability. But the geomagnetic storm threw a wrench in the works. The induced currents in the transformers and along those transmission lines seriously increased the reactive power demand of the system. The resulting distortions in the voltage and current waveforms hadn’t been considered when the equipment was installed. The SVCs weren’t configured to handle the dynamic conditions affecting the system, so relays designed to protect them tripped, pulling the equipment out of service. Without the SVCs, the voltage on the grid dropped, the frequency increased, and chaos ensued. The grid operators couldn’t disconnect customers fast enough to keep things stable, and within seconds, the rest of the system collapsed. Lots of equipment was permanently damaged, and millions woke up that frigid morning with no real power, reactive power, or apparent power, shutting down basically the entire province for half-a-day and requiring costly and expensive repairs. They learned a lot of lessons that day, and adjusted a lot of relay settings since then. It’s just one of many case studies on the importance of understanding and managing this hopefully a little-less-perplexing idea of reactive power on the grid.

Every Kind of Bridge Explained in 15 Minutes

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

The Earth is pretty cool and all, but many of its most magnificent features make it tough for us to get around. When the topography is too wet, steep, treacherous, or prone to disaster, sometimes the only way forward is up: our roadways and walkways and railways break free from the surface using bridges. A lot of the infrastructure we rely on day to day isn’t necessarily picturesque. It’s not that we can’t build exquisite electrical transmission lines or stunning sanitary sewers. It’s just that we rarely want to bear the cost. But bridges are different. To an enthusiast of constructed works, many are downright breathtaking. There are so many ways to cross a gap, all kindred in function but contrary in form. And the typical way that engineers classify and name them is in how each design manages the incredible forces involved. Like everything in engineering, terminology and categories vary. As Alfred Korzybski said, “The map is not the territory.” But, trying to list them all is at least a chance to learn some new words and see some cool bridges. And honestly, I can hardly think of anything more worthwhile than that. I’m Grady, and this is Practical Engineering.

One of the simplest structural crossings is the beam bridge: just a horizontal member across two supports. That member can take a variety of forms, including a rolled steel beam (sometimes called a stringer) or a larger steel member fabricated from plates (often called a plate girder). Most modern bridges built as overpasses for grade separation between traffic are beam bridges that use concrete girders. And instead of a group of individual beams, many bridges use box girders, which are essentially closed structural tubes that use material more efficiently (but can be more complicated to construct). Beam bridges usually can’t span great distances because the girders required would be too large. At a certain distance, the beams become so heavy, they can hardly support their own weight, let alone the roadway and traffic on top.

One way around the challenge of the structural members’ self-weight is to use a truss instead of a girder. A truss is an assembly of smaller elements that creates a rigid and lightweight structure. Unlike a beam, the members of a truss don’t typically experience bending forces. The connections usually aren’t actual hinges that permit free rotation, but they are close enough. So, all the load is axial (along their length) in compression or tension. That simplifies the design process because it’s easier to predict the forces within each structural member. The weight reduction allows trusses to span greater distances than solid beams, and there are a wide variety of arrangements, many with their own specific names. In general, a through truss puts the deck on the bottom level, and a deck truss puts it on top, hiding the structural members below the road. A particularly photogenic type of truss is a lenticular truss bridge, named because they resemble lenses, which themselves are named because they resemble lentils! A Bailey bridge is a kind of temporary truss bridge that is designed to be portable and easy to assemble. They were designed during World War II, but Bailey bridges are still used today as temporary crossings when a bridge fails or gets closed for construction. Most covered bridges are timber truss bridges. Since wood is more susceptible to damage from exposure to the elements, the roof and siding are placed to keep the structural elements truss-worthy. A trestle bridge is superficially similar to a truss: a framework of smaller members. Trestle bridges don’t have long spans, but rather a continuous series of short spans with frequent supports which are individually called trestles, but sometimes the whole bridge is just called a trestle, so like so many other instances of structural terminology, it can be a little confusing.

This next bridge type uses a structural feature that’s been a favorite of builders for millennia: the arch. Instead of beams loaded perpendicularly or trusses that experience both compressive and tensile forces, arch bridges use a curved element to transfer the bridge’s weight to supports using compression forces alone. Many of the oldest bridges used arches because it was the only way to span a gap with materials available at the time (stone and mortar). Even now, with the convenience of modern steel and concrete, arches are a popular choice for bridges. They make efficient use of materials but can be challenging to construct because the arch can’t provide its support until it is complete. Temporary supports are often required during construction until the arch is connected at its apex from both sides. In stone arches, the topmost stone is key to keeping the whole thing standing, and, of course, it’s called the keystone. When the arch is below the roadway, we call it a deck arch bridge. Vertical supports transfer the load of the deck onto the arch. The area between the deck and arch has a great name: the spandrel. Open-spandrel bridges use columns to transfer loads, and closed-spandrel bridges use continuous walls. If part of the arch extends above the roadway with the deck suspended below, it’s called a through arch bridge. A moon bridge is kind of an exaggerated arch bridge, usually reserved for pedestrians over narrow canals where there’s not enough room for long approaches. They’re steep, so sometimes you have to use steps or ladders to get up to the top and back down.

One result of compressing an arch is that it creates horizontal forces called thrusts. Arch bridges usually need strong abutments at either side to push against that can withstand the extra horizontal loads. Alternatively, a tied arch bridge uses a chord to connect both sides of the arch like a bowstring, so it can resist the thrust forces. That means a tied arch is structurally more of a truss than an arch, and that provides a lot of opportunities for creativity. For just one example, a network arch bridge uses the tied arch design, plus criss-crossed suspension cables, to support the deck. To tell an arch from a tied arch by eye, it’s usually enough to look at the supports. If the end of each arch sits atop a spindly pier or some other structure that seems insubstantial against horizontal forces, you can probably bet that they are tied together and it’s not a true arch bridge. Similarly, a rigid-frame bridge integrates the superstructure and substructure (in other words, the deck, supports, and everything else) into a single unit. They don’t have to be arched, but many are.

Another way to increase the span of a beam bridge is to move the supports so that sections of the deck balance on their center instead of being supported at each end. A cantilever bridge uses beams or trusses that project horizontally, balancing most of the structure’s weight above the supports rather than in the center of the span. This is such an effective technique that the Forth Bridge crossing the Firth of Forth in Scotland took the title of longest span in the world away from the Brooklyn Bridge in 1890 and held the record for decades. This famous photograph demonstrates the principle of that bridge perfectly: The two central piers bear the compression loads from the bridge. And, the outer-most supports are anchors to provide the balancing force for each arm. This way, you can suspend a load in the middle.

The longest bridges take advantage of steel’s ability to withstand incredible tension forces using cable supports. Cable-stayed bridges support the deck from above through cables attached to tall towers or spars. The cables (also called stays) form a fan pattern, giving this type of bridge its unique appearance. Depending on the span, cable-stayed bridges can have one central tower or more. Their simplicity allows for a wide variety of configurations, giving rise to some dramatic (and often asymmetric) shapes.

For shorter spans, you can combine the benefits of a cable-stayed structure with girders to get an extradosed bridge. Imagine a concrete girder bridge that uses internal tendons to keep the concrete in compression, then just pull those tendons out of the girder and attach them to a short tower. Rather than holding the deck up vertically like a cable-stayed bridge, they’re acting more horizontally to hold the girders in compression, giving them the stiffness needed to support the deck. It’s a relatively new idea compared to most of the other designs I’ve listed, but there are quite a few cool examples of extradosed bridges across the globe.

Where a cable-stayed bridge attaches the deck directly to each tower, a suspension bridge uses cables or chains to dangle the deck below. In a simple suspension bridge, the cables follow the curve of the deck. This is your classic rope bridge. They’re not very stiff or strong, so simple suspension bridges are usually only for pedestrians. A stressed ribbon bridge takes the concept a step further by integrating the cables into the deck. The cables pull the deck into compression, providing stiffness and stability so it doesn’t sway and bounce. This design is also primarily used for smaller pedestrian bridges because it can’t span long distances and the deck sags in the middle.

Then you have the suspended deck bridge, the design we most associate with the category with the longest spans in the world. Massive main cables or chains dangle the road deck below with vertical hangers. Suspension bridges are iconic structures because of their enormous spans and slender, graceful appearance. Towers on either side prop up the main cables like broomsticks in a blanket fort. Most of the bridge’s weight is transferred into the foundation through these towers. The rest is transferred into the bridge’s abutments through immense anchorages keeping the cables from pulling out of the ground. Alternatively, self-anchored suspension bridges connect the main cables to the deck on either side, compressing it to resist the tension forces. Because they are so slender and lightweight, most suspension bridges require stiffening with girders or trusses along the deck to reduce movement from wind and traffic loads. These bridges are expensive to build and maintain, so they’re really only used when no other structure will suffice. But you can hardly look at a suspended deck bridge without being impressed.

Bridges have to support the vehicles and people that cross over the deck, but they often have to accommodate boats and ships passing underneath as well. If it’s not feasible to build the bridge and its approaches high enough, another option is just to have it get out of the way when a ship needs to pass. Moveable bridges come in all shapes and sizes. A lot of people call them drawbridges after their medieval brethren over castle moats. A bascule bridge is hinged so the deck can rotate upward. A swing bridge rotates horizontally so a ship can pass on either side. A vertical lift bridge raises the entire deck upward, keeping it horizontal like a table. A transporter bridge just has a small length of deck that is shuttled back and forth across a river. That’s just a few, and in fact, every moveable bridge is unique and customized for a specific location, so there are some truly interesting structures if you keep an eye out.

On the other hand, sometimes there’s no need for ship passage or a lot of space below, and in that case, you can just float the bridge right on the water. Floating bridges use buoyant supports, eliminating the need for a foundation. These are used in military applications, but there are permanent examples too. Many use hollow concrete structures as pontoons, with pumps inside to make sure they don’t fill up with water and sink. And actually, a lot of bridges take advantage of buoyancy in their design, even if it’s not the main source of support. A design like this presents a lot of interesting engineering challenges, so there aren’t too many of them. Similarly, the pedestrian bridge at Fort de Roovere in the Netherlands (probably pronounced that wrong) has its deck below the water, giving it the nickname of the Moses Bridge.

If space or funding is really tight, one option to span a small stream is a low-water crossing. Unlike bridges built above the typical flood level, low-water crossings are designed to be submerged when water levels rise. They are most common in areas prone to flash floods, where runoff in streams rises and falls quickly. Ideally, a crossing would be inaccessible only a few times per year during heavy rainstorms. However, low-water crossings have some disadvantages. For one, they can block the passage of fish just like a dam. And then there’s safety. A significant proportion of flood-related fatalities occur when someone tries to drive a car or truck through water overtopping a roadway. Water is heavy. It takes only a small but swift flow to push a vehicle down into a river or creek, which means at least some of the resources saved by avoiding the cost of a higher bridge are often spent to erect barricades during storms, install automatic flood warning systems, and run advertisement campaigns encouraging motorists never to drive through water overtopping a roadway.

You may have heard the term viaduct before. It’s not so much a specific type of bridge, but really about the length. Bridges that span a wide valley need multiple intermediate supports. So, a viaduct is really just a long bridge with multiple spans that are mostly above land. There’s really not a lot of agreement on what is one and what isn’t. Some are singular and impressive structures. But many modern cities have viaducts that are, although equally amazing from an engineering standpoint, a little less beautiful. So, you’re more likely to hear them called elevated expressways. And that gets to the heart of a topic like this: without listing every bridge, there’s no true way to list every type of bridge. There’s too much nuance, creativity, and mixing and matching designs. The Phyllis J. Tilly bridge in Fort Worth, Texas combines an arch and stressed ribbons. The Third Millennium Bridge in Spain uses a concrete tied arch with suspension cables holding up the deck which is stiffened with box girders. The Yavuz Sultan Selim Bridge in Turkey combines a cable-stayed and suspension design. In some parts of India and Indonesia, living tree roots are used as simple suspension bridges over rivers. There are bridges for pipelines, bridges for water, bridges for animals, and I could go on. But that’s part of the joy of paying attention to bridges. Once you understand the basics, you can start to puzzle out the more interesting details. Eventually, you’ll see the Akashi Kaikyo Bridge on a calendar in your accountant’s office, and let him know it’s a twin-hinged, three-span continuous, stiffened truss girder suspension bridge with a double-tower system. Or maybe that’s just me.

How Bridge Engineers Design Against Ship Collisions

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

On March 26, 2024 (just a few weeks ago, if you're watching this as it comes out), a large container ship struck one of the main support piers of the Francis Scott Key Bridge in Baltimore, Maryland, collapsing the bridge, killing six construction workers, injuring one more, and seriously disrupting both road and marine traffic in the area. There’s a good chance you saw this in the news, and hopefully you’ve seen some of the excellent content already released by independent creators providing additional context. I got a lot of requests to talk about the event, and I usually prefer to wait to discuss events like this until there are more details available from investigations, but I think it might be helpful to provide some context from an engineering perspective about how we consider vessel collisions in the design of bridges like this one, and why the Francis Scott Key bridge may have collapsed. I’m Grady, and this is Practical Engineering. Today we’re talking about vessel collision design for bridges.

The Francis Scott Key Bridge was a steel arch-shaped continuous through truss bridge. I’m working on a video that goes into a lot more detail about the different kinds of bridges and how they’re classified, but this bridge had kind of a medley of structural styles, so let me hand it off to our special guest correspondent, Road Guy Rob, to break that terminology down.

Well, Grady, I'm in Long Beach, California today, standing on top of this brand new bridge that replaced an old arch/truss bridge that used to be right there. It kind of looked like a baby Key Bridge, and the Port of Long Beach is happy that it's gone.

The Gerald Desmond Bridge was a truss bridge. Instead of having one big large beam, a truss has lots of smaller connected structural members all attached together.

This creates a rigid structure that's much lighter weight than a big heavy beam, and that makes trusses efficient and clever when they work. Both the Key bridge and the old bridge that used to be here were “through-truss” bridges. It's a sort of arch shape, and the driving deck is suspended below the truss, so you sort of drive through the arch, but it's not an actual arch with like a keystone and all the pieces pushing horizontally to hold each other together. No, this through-truss bridge has no hinges or joints at the main supports, nothing that breaks it up into sections. So that's why engineers called the Key bridge a continuous truss bridge. It's all one big piece, and it's all bolted and welded into a single rigid truss across its entire length. And then that load distributes across all three spans of the bridge.

Now, the approach roads on each side are entirely separate bridges, even though they link together. They just look like concrete roads sitting on top of simple girder spans.

Well, you ask, what happened to that baby Key bridge in Long Beach? Well, the only way you're going to see it now is to turn on Grand Theft Auto five and look at the fictionalized version of it immortalized in Los Santos for all time. Because when this bridge opened, the port of Long Beach demolished the old bridge and the last scraps of it got all the way back in October.

In its place, this new, fancier looking cable-stayed bridge, the Long Beach International Gateway. And what the Port of Long Beach did in studying to build this bridge and the list of improvements they came up with might give us some clues what Baltimore might want to end up doing when they replace the Key bridge down the line.

And we'll talk about that in just a moment.

When the Dali container ship lost power and drifted into the southwest pier, the support collapsed, and most of the truss and deck fell with it. Both the southwest and central spans fell roughly vertically with the loss of support from the damaged pier. Part of the truss on the northwest side separated from the unsupported section and rotated toward the northeast span, taking several of the approach spans with it. Thankfully, the ship had put out a mayday call before the impact, allowing police officers at either end of the bridge to close it to traffic. Tragically, it wasn’t enough time to get the crew of eight construction workers off the structure before it fell, six of whom lost their lives.

Just dealing with the salvage and removal of the steel and concrete debris left over from the collapse has been a massive undertaking. Within a week, engineering teams were on-site measuring, cutting, lifting, and floating away huge chunks of the wreckage in separate salvage operations for the main bridge, approaches, and the vessel. As of this writing, they’re still working hard on it. At least seven floating cranes were involved, including the famous Weeks 533 that pulled US Airways Flight 1549 from the Hudson River in 2009. This was essentially a massive Jenga tower: the order of operations and the precision of each cut and each lift mattered. With so much debris underwater, they had to map it out to understand how everything was stacked together. Access was a major challenge, and the stresses in the wreckage were hard to characterize, so it’s been a slow and deliberate process requiring careful planning and tons of skill to do safely. Fortunately for Baltimore, there are large industrial facilities in the port that can process the thousands of tons of material that will ultimately be removed. Of course, reopening the port to shipping traffic is a huge priority. A small channel was marked out under one of the approach bridges for smaller vessels like tug and barges traffic, and the Army Corps of Engineers is making good progress on opening up the main channel, but it isn’t clear when full-scale operations at the port will be able to resume. Shipping traffic isn’t the only traffic affected either, the bridge carried thousands of road vehicles per day that now have to be re-routed. There is a tunnel under the harbor that provides a decent alternate route, but trucks with hazardous materials aren’t allowed through, requiring an enormous detour around the city.

It’s been more than a month since the event, but it will likely be a year or more before we get an official report documenting the probable cause of the failure. In the US, events like this are investigated by the National Transportation Safety Board or NTSB. This independent government agency is extremely diligent. And often, diligent also means slow. But events like this are how the field of engineering evolves. Human imagination isn’t limited to past experiences, but in many senses, engineering is. We just don’t have the resources to answer the millions of “what ifs” that might coalesce into a tragedy, so we lean on the hard lessons learned from past failures. When something terrible happens, it’s really important that we collectively get to the bottom of why and then make whatever changes are appropriate to our engineering systems to prevent it from happening again.

But, at the risk of stating the obvious, the failure mode in this case is pretty clear. You probably don’t need an engineer to explain why a massive ship slamming into a bridge pier would cause that bridge to collapse. I think what’s less obvious is how engineers characterize situations like this so that bridges can be designed to withstand them. Collisions with bridges by barges and ships are not a modern problem. Technically they’re called “allisions” since a bridge isn’t moving, but that term is used more in the maritime industry than by bridge engineers. Between 1960 and 2014, there were 35 major bridge collapses resulting from vessel impacts. And, 18 of those were in the US. We just have such a big network of inland waterways, and that means we have a lot of bridges.

Two spans of the Queen Isabella Causeway Bridge in Texas collapsed in 2001 when barges collided with one of the piers. A year later, a bridge carrying I-40 over the Arkansas River in Oklahoma was hit by barges when the captain lost control, collapsing a major portion of the structure. In 2009, Popp’s Ferry Bridge in Mississippi collapsed after being struck by a group of barges. In 2012, the Eggner’s Ferry Bridge in Kentucky fell when a cargo ship tried to go through the wrong channel. Before any of those, though, the Sunshine Skyway Bridge in Florida put a major focus on the problem. In 1980, a bulk carrier ship lost control because of a storm, crashing into one of the piers and collapsing the entire main span of the southbound bridge, killing 35. The event brought a lot of new awareness and concern about the safety of bridges over navigable waterways. But piers aren’t the only parts of a bridge at risk from ships. I’ll let Rob explain.

The Key bridge got into trouble because of a horizontal allision. That's where a ship moves side to side in the wrong way and hits something it's not supposed to.

Here in Long Beach, that really wasn't their concern, primarily because the old bridge columns were way inland here, so there was no way for a ship to exit the waterway and hit the column because the column was in lots of dirt. And the new replacement bridge takes no chances at all. Look how much farther onshore those columns are now!

Now, the Port of Long Beach were far more worried of the old Gerald Desmond Bridge getting hit vertically. The old bridge was 155ft tall. That's like a 15 story building. And if that sounds pretty tall to you, it sounded pretty tall to them back in 1968 when they built the bridge. But as we now know, ships are getting bigger and fatter and taller and 155ft wasn't cutting it for some of the modern ships that were trying to get into the back part of the port, where there's a lot of cranes and action happening over there. So the new bridge adds another 50ft, takes it over 200ft. That's like a 20 story building to get up from the waterline to that new bridge.

And this new, taller, Long Beach International Gateway helps the port scratch off one designation they didn't want - having the shortest bridge over a port in the United States. Well, that's gone now, and thankfully in a less tragic manner than what's happening on the East Coast.

In the aftermath of the Sunshine Skyway collapse, the federal government and the professional community, both from the engineering and maritime sides, invested a serious amount of time and investigation into the issue. One culmination was updated bridge codes that included requirements for consideration of vessel collisions. For highway bridges in the US, those specifications are put out by an organization called the American Association of State Highway and Transportation Officials (or AASHTO), but there are similar requirements worldwide, including in the Eurocode.

A lot of infrastructure is designed for worst-case scenarios, but at a certain point, it just isn’t feasible. This is something I’ve talked a lot about in previous videos: you have to draw a line somewhere that balances the benefits and costs. If the code required us to design bridges with Armageddon meteorite or Godzilla protection, we just wouldn’t build any bridges. It would be too expensive. And that’s kind of true for ship collisions too. The mass and kinetic energy of the cargo vessels today is tough to even wrap your head around. We just couldn’t afford to build bridges if they all had to be capable of withstanding a worst-case collision. Instead, for what engineers call “high consequence, low probability” events, codes often set the standard as some acceptable amount of risk. There’s always going to be some possibility of an event like this, but how much risk are we as a society willing to bear for the benefit of having easy access across navigable waterways? In the U.S., that answer, at least according to AASHTO for critical structures like the Key Bridge, is 0.01 percent probability in a given year. For some perspective, it’s roughly the chance of rolling a Yahtzee (five-of-a-kind) in a single throw. But it’s an annual probability, so you have to roll the dice once every year. If you did it forever, it would average out to once every 10,000 years, but that doesn’t mean it couldn’t happen twice in a row. So an engineer’s job is to design the structure not to survive in a worst-case scenario but to have a very low probability of collapsing from a vessel impact. And there’s a lot that goes into figuring that out.

This is the general formula for the annual probability of bridge collapse due to a ship collision. You have all these factors that get multiplied together. The first one is just the number of ships that pass under the bridge in a year. And there’s a growth factor in there for how that number might change over time. Then there’s what’s called the probability of vessel aberrancy; basically, the chance that one of those ships loses control. AASHTO has some baseline numbers for this based on long-term accident statistics in the US, and the designer can apply some correction factors based on site-specific issues like water currents and navigation aids. Then, there’s the geometric probability of a collision if a ship does lose control. When a vessel is aberrant, you don’t know which way it’s going to head. This gets a little complicated, but if you’re familiar with normal distributions it will make perfect sense. You can plot a normal distribution curve centered on the transit path with one standard deviation equal to the length of the aberrant ship to give you an approximation of where it might end up. The area under that curve that intersects with the bridge piers is the probability that the ship will impact the bridge if it loses control. And this is really the first knob an engineer can turn to reduce the risk, because the farther the piers are from the transit path, the lower the geometric probability of a collision. And this factor can be modified if ships have tethered tugs to assist with staying in the channel, something that wasn’t required in Baltimore at the Key Bridge.

But, even if there is a collision, that doesn’t necessarily mean the bridge will collapse. This is where the structural engineering comes into play. The probability of collapse depends both on the lateral strength of the pier and the impact force from the collision. But, that force isn’t as simple as putting a weight on a scale. It’s time-dependent, and it varies according to the size and type of vessel, its speed, the amount of ballast, the angle of the collision, and a lot more. Usually, we boil that down to an equivalent static load. And based on some testing, this is the equation most engineers use. It’s just based on the deadweight tonnage (basically how much the ship can carry) and its velocity. It’s interesting that they settled on deadweight, which doesn’t include the weight of the ship itself. But again, this analysis is pretty complicated, especially because you have to do it for every discrete grouping of vessel size and bridge component, so some simplifications make sense, and since this one assumes every ship is fully loaded, it’s relatively conservative.

Just for illustration, the ship that hit the Sunshine Skyway Bridge had a deadweight of 34,000 tonnes. The NTSB report doesn’t estimate the speed at which it hit the bridge, but let’s say around 5 knots. That would be equivalent to a static force of around 56 meganewtons or 13 million pounds if the ship was fully loaded, which it wasn’t (but there’s no way to account for that in this equation). The Dali has a deadweight of 117,000 tonnes and was traveling at roughly 5 knots on impact. That’s equivalent to more than 100 meganewtons or 24 million pounds, again, assuming the ship was fully loaded (which, again, it wasn’t). But you can validate this with some back-of-the-envelope physics. Force is equal to mass times acceleration. We know the mass of the ship and its cargo from records: about 112,000 metric tons. To decelerate that mass from 5 knots to a standstill over the course of, let’s say, 4 seconds requires, roughly, a force of 72 meganewtons or 16 million pounds. Even as a rough guess, that is a staggering number. It’s 5 SpaceX Starships pointed at a single bridge pier.

Designing a bridge to handle these forces is obviously complicated. It’s not just the pier itself that has to survive, but every element of the bridge along the load path, including the foundation, and (assuming the pier isn’t perfectly rigid), the superstructure too. Again, it’s not impossible to design, but it gets pricey fast, which is why designers have more knobs to turn to meet the code than just the strength of the bridge itself. One of those knobs is pier protection systems. Fenders can be installed to soften the blow of a ship impact, but for ships of this size, they would have to be enormous. Islands can be built around piers to force ships aground before the hit the bridge. But islands create environmental problems because of the fill placed on the river bottom, plus they get really big for deeper channels, so the bridge span has to be wider to keep the channel from being blocked. Islands can even affect currents in the water and the bridge structure, creating additional load on the foundation as they settle after construction. Another commonly used protection structure is called a dolphin. This is usually a circular construction of driven sheet piles, filled with sand or concrete. Dolphins can slow a ship, stop it altogether, or redirect it away from critical bridge elements like piers. The new Sunshine Skyway Bridge used islands and dolphins to protect the rebuilt span, and actually, the Key Bridge had four dolphins, one on either side of each main support. Unfortunately, because it came at an angle, the Dali slipped past the protection when it lost control.

It’s important to point out that everything I’ve discussed is a modern look at how engineers consider vessel impacts to bridges. When the Francis Scott Key Bridge was finished in 1977, there were no requirements like this, and the bridge never had a major rehabilitation or repair that would have triggered adherence to the newer codes. Container ships the size of Dali didn’t even exist until around 2006. And we don’t know what the ships of the future will look like. It’s easy to say with hindsight that a bridge like this should have been better protected against errant ships, but if you say it for this one, you really have to say it for all the bridges that see similar maritime traffic. And that represents an enormous investment of resources for, potentially, not a lot of benefit to the public, given how rare these situations are. That’s not me saying it shouldn’t be done; it’s just me saying that a decision like that is a lot more complicated than it might seem. I don’t expect we’ll see bridge design code changes come out of this event, but vessel collisions will certainly be on the minds of the designers for the replacement in Baltimore. I’ll let Rob explain.

When you take a look at photos of the Key Bridge, it looks like Maryland was doing a good job taking care of their bridge. So if the NTSB report comes back and says the bridge was in good shape, it's 100% the ship that's at fault, well, I don't think any of us are going to be really that shocked.

But for the old Gerald Desmond Bridge here in Long Beach, that used to be right here, well, the environmental impact report, where they studied to build this new replacement bridge, the port staff really didn't seem too concerned about a maritime navigation failure. Of a structural failure? Let's just say engineers scored bridges out of 100 points. So you have a brand new bridge, it gets 100 points. On that scale, the old Gerald Desmond Bridge that was right here scored a 43. I mean, anything below 80 points, you get federal money to work on the bridge to try to rehab it and get it back into good shape. And anything, anything under 50 points, it's so bad the federal government starts throwing money in trying to help you replace the bridge.

That's how bad off the Gerald Desmond Bridge was. Salt from the air of the sea and decades of it sitting above sea water and all of that, just nice salt in the air, eating away at the paint. Well, that paint was rated very poor on the old Gerald Desmond Bridge. And, you know, paint protects all the bridge members, all the metal from rusting out. And as Grady points out, every single member of a truss is really important if you, you know, want the bridge to stay in good shape and not fall down, right?

Engineers also conducted a load analysis. They tested to see as trucks drove over it, how the bridge was holding up. And they found members of the arch main span were overstressed for all design trucks. So that didn't matter if you drove a big truck or a little truck. They were all causing problems with the bridge. And the concrete that those trucks drove on? It was all cracking up. It was rated critical. The port had to install big nets to catch big chunks of pavement that were falling off the bridge and could hit somebody on the head down here.

So, Long Beach had four objectives that this new bridge needed to meet in order to build it. And those goals may mirror some of the ones Baltimore may want to have when building their replacement bridge. 1. This bridge had to have a design life of 100 years. Say, stay structurally sound for that long. 2. Long Beach wanted to reduce the approach grades on both sides, even getting up to 155ft before. Sometimes you were driving up a 6% grade and now that this thing is over 200 ft high, that would be way too steep. So they instead built these huge freeway viaducts that go on and on for like a quarter of a mile to lift people and trucks gently up to that new bridge height. Baltimore's bridge already has some very long approaches to it, so I don't know whether they're going to replace the, uh, ramps approaching the bridge or not. It'll be interesting to see what they end up deciding to do. 3. Provide sufficient roadway capacity to handle future car and truck traffic. The old bridge here was two lanes in each direction. Four lanes. This widens it to six. The Key bridge in Baltimore was also only four lanes. But this bridge handles twice the traffic every day. You know, compared to the key bridge back when it was open, right?

And both the Key Bridge and the old Gerald Desmond Bridge had no shoulders for emergency vehicles and stalled cars to pull off to the side. And as you can see, the new bridge has these excellent shoulders on both the outside and the inside of travel lanes. So that makes the road a lot safer, because you're not going to run into the back of a stalled truck in the dark.

It's also a lot safer if you're not in your car, because this bridge has a way you can cross it without being in a car. They've added this 12ft wide pedestrian and bicycle pathway, which is about 12ft wider than what they had before. Used to be on the old bridge, The only way across was inside a car. It's a good start, certainly not perfect. Right now, the path just hits this gate and stops. The city of LA owns the next harbor bridge down that way. It's called the Vincent Thomas Bridge. It's also old, so it doesn't have a pedestrian walkway, so this pathway sort of just ends at the city limit because there's nowhere for it to go. But adding in a multi-use path like this one onto the new Key Bridge would be such a no-brainer. It could take a four mile bike ride from like, Hawkins Point to Sparrows Point, down from four miles with the bridge path from 22 miles right now, without it.

And finally the fourth goal: providing vertical clearance for new generation of larger vessels, which the new bridge certainly has. And that must make the port very happy. And I'm willing to bet that Baltimore will take that goal and maybe turn it on its side and talk about horizontal clearance and insist on a design that eliminates the risk of an allision, like what happened on March 26th from ever happening again, Grady.

Thanks Rob. If you love deep dives into transportation infrastructure, go check out his channel after this. But, it’s important to point out that this wasn’t just a bridge failure; it was a bridge failure precipitated by a maritime navigation failure. Obviously, engineers who design bridges don’t have a lot of say in the redundancies, safety standards, and navigation requirements of the vessels that pass underneath them. But if you look at the whole context of this tragedy and ask, “How can our resources best be used to prevent something like this from happening again?”, reducing the probability of a ship this size losing control has to be included with the structural solutions like pier protection systems. I don’t know a lot about that stuff, so I couldn’t tell you what that might include, but I’m sure NTSB will have some recommendations when their report eventually comes out. Having tugs accompany large ships while they traverse lightly protected bridges seems like a prudent risk reduction measure, but that’s just coming from a civil engineer.

And, speaking of risk reduction, I have to say that using risk analysis as a tool for design is really not that satisfying. We humans are notoriously bad at understanding probabilities and risks, and engineers are not that great at communicating what they mean to people who don’t speak that language. That’s how we get confusing terminology like the hundred-year flood. And it’s unsettling to come face-to-face with the idea that, even if our bridges are designed and built to code, there’s still a chance of something like this. Everything’s a tradeoff, but the people driving over the bridge (or working on it) had no direct say in where the line was drawn or whether it applied retroactively, even as ships got bigger and bigger. But I hope it’s clear why we do it this way. The question isn’t “Can we design bridges to be safer?” The answer to that is always “yes.” The real question is, “How much risk can we tolerate?” or put in a different way, “How much are we willing to spend on any incremental increase in safety?” Because the answers to those questions are much more complex and nuanced. And if all bridges were required to survive worst-case collisions with ships like Dali, we would just have a lot fewer bridges. But sometimes it takes an event like this to remind us that risks aren’t just small numbers on a piece of paper. They represent real consequences, and my heart goes out to the families of the victims affected by this event. I hope we can honor them by learning from it and making improvements, both to our infrastructure and our maritime systems, so that it doesn’t happen again.

Connecting Solar to the Grid is Harder Than You Think

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

On June 4, 2022, a small piece of equipment (called a lightning arrestor) at a power plant in Odessa, Texas failed, causing part of the plant to trip offline. It was a fairly typical fault that happens from time to time on the grid. There’s a lot of equipment involved in producing and delivering electricity over vast distances, and every once in a while, things break. Breakers isolate the problem, and we have reserves that can pick up the slack. But this fault was a little bit different.

Within seconds of that one little short circuit at a power plant in Odessa, the entire Texas grid unexpectedly lost 2,500 megawatts of generation capacity (roughly 5% of the total demand), mainly from solar plants spread throughout the state. For some reason, a single 300-megawatt fault at a single power plant magnified into a loss of two-and-a-half gigawatts, dropping the system frequency to 59.7 hertz. The event nearly exceeded Texas’s “Resource Loss Protection Criteria,” which is minimum loss of power that requires having redundancy measures in place. Another fault in the system could have required disconnecting customers to reduce demand. In other words, it was almost an emergency.

If you lived in Texas at the time, you probably didn’t notice anything unusual, but this relatively innocuous event sent alarm bells ringing through the power industry. Solar plants, large-scale batteries, and wind turbines don’t produce power like conventional thermal power plants that make up such a big part of the grid. The investigation into the 2022 Odessa disturbance found that it wasn’t equipment failures that caused all the solar plants to drop so much production all at once, at least not in the traditional sense. Instead, a wide variety of algorithms and configuration settings in the power conversion equipment reacted in unexpected ways when they sensed that initial disturbance.

The failure happened just before noon on a sunny summer day, so solar plants around the state were at peak output, representing about 16% of the total power generation on the grid. That might seem high, but there have already been times when solar was powering more than a third of Texas’s grid, and that number is only going up. The portion of the grid comprised of solar power is climbing rapidly every year, and not just in Texas, but worldwide. So the engineering challenges in getting these new sources of power to play nicely with the grid that wasn’t really built for them are only going to become more important. And, of course, I have some demos set up in the garage to help explain. I’m Grady and this is Practical Engineering. In today’s episode, we’re talking about inverter-based resources on the grid.

Solar panels and batteries work on direct current, DC. If you measure the voltage coming out, it’s a relatively constant number. This is actually kind of true for wind turbines as well. Of course, they are large spinning machines, similar to the generators in coal or natural gas plants. But unlike in thermal power plants that can provide a smooth and consistent source of power through a steam boiler, winds vary a lot. So, it’s usually more efficient to let the turbine speed vary to optimize the transfer of energy from the wind into the blades. There are quite a few ways to do this, but in most cases, you get a variable-speed alternating current from the turbine. Since this AC doesn’t match the grid, it’s easier to first convert it to DC. So you have this class of energy sources, mostly renewables, that output DC, but the grid doesn’t work on DC (at least not most of it).

Nearly all bulk power infrastructure, including the power that makes it into your house, uses an alternating current. I won’t go into the Tesla versus Edison debate here, but the biggest benefit of an AC grid is that we can use relatively simple and inexpensive equipment (transformers) to change the voltage along the way. That provides flexibility between insulation requirements and the efficiency of long-distance transmission. So we have to convert, or more specifically invert, the DC power from renewable sources onto the AC grid. In fact, batteries, solar panels, and most wind turbines are collectively known to power professionals as “inverter-based resources” because they are so different from their counterparts. Here’s why.

The oldest inverters were mechanical devices: a motor connected to a generator. This is pretty simple to show. I have a battery-powered drill coupled to a synchronous motor. When I pull the trigger, the drill motor spins the synchronous motor, generating a nice sine wave we can see on the oscilloscope. Maybe you can see the disadvantages here. For one, this is not very efficient. There are losses in each step of converting electricity to mechanical energy and then back into electrical energy on the other side. Also, the frequency depends on the speed of the motor, which is not always a simple matter to control. So these days, most inverters use solid-state electronic circuits, and look what I found in my garage.

These are practically ubiquitous these days, partly because cars use a DC system, and it’s convenient to power AC devices from them. I just hook it up to the battery, and get nice clean power from the other end… haha just kidding. These cheap inverters definitely output alternating current, but often in a way that barely resembles a sine wave. Connecting a load to the device smooths it out a bit, but not much. That’s because of what’s happening under the hood. In essence, switches in the inverter turn on and off, creating pulses of power. By controlling the timing of the pulses, you can adjust the average current flowing out of the inverter to swing up and down into an approximate sine wave. Cheaper inverters just use a few switches to create a roughly wave-like signal. More sophisticated inverters can flip the switches much more quickly, smoothing the curve into something closer to a sine wave. This is called pulse width modulation. Boost the voltage on the way in or the way out, add some filters to smooth out the choppiness of the pulses, and that’s how you get a battery to run an AC device… but it’s not quite how you get a solar panel to send power into the grid. There is a lot more to this equipment.

For one, look at the waveform of my inverter and the one from the grid. They’re similar enough, but they’re definitely not a match. Even the frequency is a little bit off. I will not be making an interconnection here, since I don’t have a permit from the utility, but even if I did, this inverter would let out the magic smoke. A grid-tie inverter has to be able to both synchronize with the phase and frequency of the grid and be able to vary the voltage of the waveform to control how much current is flowing into or out of the device. The synchronization part often involves a circuit called a phase-locked loop. The inverter senses the voltage of the grid and sets the timing of all those little switches accordingly to match what it sees. So, these are often called grid-following inverters. They synchronize to the grid frequency and phase and only vary the voltage to control the flow of power. And that hints at one of their challenges: they only work when the grid is up.

I’ve done a video all about black starts, so check that out after this if you want to learn more, but (in general), inverter-based resources like solar, wind, and batteries can only follow what’s already on the grid. When the system’s down, they are too, regardless of whether the sun’s shining or the wind’s blowing. That’s why most grid-tied solar systems on houses can’t give you power during an outage.

There’s another interesting thing that inverters do for solar panels, and I can show you how it works in my driveway. I have a solar panel hooked up to a variable resistor, and I’m measuring the voltage and current produced by the panel. You can see as I lower the resistance, the output voltage of the panel goes down and the current it supplies goes up. But this isn’t a linear effect. I recorded the voltage and current over the full range, and multiplied them together to get the power output. If you graph the power as a function of voltage, you get this shape. And you can see there’s an optimum resistance that gets you the most power out of the panel. It’s called the maximum power point. If you deviate on either side of it, you get less power out. In other words, you’re leaving power on the table. You’re not taking full advantage of the panel’s capacity.

What’s even more challenging is that point changes depending on the temperature of the panel and the amount of sun hitting it. I ran this test again with a few more clouds, and you can see how the graph changes. So nearly all large solar photovoltaic installations use what’s called a Maximum Power Point Tracker (or MPPT) that essentially adjusts the resistance to follow that point as it changes with sunniness and temperature. It’s really a separate device from the inverter, but often they’re located right next to each other or inside the same housing. Even this panel came with a charge controller that has this MPPT function, and you can see it adjusting the flow of current to constantly try and stay at the peak of the curve while it charges this battery. These can be used for an entire installation, but in many cases, each panel or group of panels gets its own MPPT because that curve is just a little bit different for each one. Tracking the peak power output individually can often squeeze a little more capacity out of the system.

Squeezing out capacity is essential to address another challenge associated with inverter-based resources on the grid: frequency. The rate at which the voltage and current on the grid swing back and forth is an important measure of how well generation and demand are balanced. If demand outstrips the generation capacity, the frequency of the grid slows down. Lots of equipment, both on the generation side and the stuff we plug in, is designed to rely on a stable grid frequency, so if it deviates too far, stuff goes wrong: Devices malfunction, motors can overheat, generators get out of sync, and more. It’s so important that rather than let the frequency get too far out of whack, grid operators will disconnect customers to get electrical demands back in balance with the available supply of power, called an under-frequency load shed. Things go wrong on the grid all the time, so generators have to be able to make up for contingencies to keep the frequency stable. Here’s the quintessential example: an unexpected loss of generation.

Say a generator trips offline, maybe because of a failed lighting arrestor like the Odessa example. The system frequency immediately starts dropping, since power demand now exceeds the generation. And the frequency will keep dropping unless we inject more power into the system. The first part of that, called Primary Frequency Response, usually comes from automatic governors in power plants. If we do it fast enough, the frequency will reach a low point, called the nadir (NAY-dur), and then recover to the nominal value. The nadir is a critical point, because if it gets too low, the grid will have to shed load in order to recover. The other important value is called the rate-of-change-of-frequency, basically the slope of this line. It determines how much time is available to get more power into the system before the frequency gets too low, and there are several factors that play into it: How much generation was lost in the first place, how quickly we can respond, and how much inertia there is on the grid. Thermal power plants that traditionally make up the bulk of generating capacity are gigantic spinning machines. They’re basically a bunch of synchronized flywheels. That kinetic energy helps keep them spinning during a disturbance, reducing the slope of the frequency during an unexpected loss.

Maybe you can see the problem with a simple grid-following inverter. It’s locked in phase with the frequency, even if that frequency is wrong. And it has no physical inertia to help arrest a deviation in frequency. If we keep everything the same and just increase the share of inverter-based resources, any loss of generation will mean a steeper slope, reducing the time available to get backup supplies onto the grid before it’s forced to shed load. Larger renewable plants, like solar and wind farms, are increasingly required to participate in primary frequency response, injecting power into the grid immediately when the frequency drops. And some inverters can even create synthetic inertia that mimics a turbine’s physical response to changes in frequency. But there’s another challenge to this.

Dealing with an over-frequency event is relatively straightforward: just reduce the amount of energy you’re sending into the grid. But, response to an under-frequency event requires you to have more energy to inject. In other words, you have to run the plant below its maximum capacity, just in case it gets called on during an unexpected loss somewhere else in the system. For a power company, that means leaving money on the table, so in most places, the energy markets are set up to pay power plants to maintain a certain level of reserve capacity, either through operating below maximum output or including battery storage in the plant.

The last big thing that inverter-based resources have to manage is faults. Of course, you need protective systems that can de-energize solar or wind resources when conditions on the grid could lead to damage. These are expensive projects, and there’s almost no limit to the things that can go wrong, requiring costly repairs or replacement. But, for the stability of the grid, you can’t have those protective systems being so sensitive that they trip at the hint of something unusual, like what happened in Odessa. This concept is usually referred to as “ride-through.” Especially for under-frequency events, you need inverters to continue supplying power to the grid to provide support. If they trip offline, or even reduce power, in response to a disturbance, it can lead to a cascading outage. This is kind of a tug of war between owners trying to protect their equipment and grid operators saying, “Hey, faults happen, and we need you not to shut the whole system down when they do.” And reliability requirements are getting more specific as the equipment evolves, because every manufacturer has their own flavor of protective settings and algorithms.

As inverter-based resources continue to grow rapidly in proportion to the overall generation portfolio, their engineering challenges are only becoming more important. We talked about a few of the big ones: lack of black start ability, low inertia, and performance during disturbances. And there are a lot more. But inverters also provide a lot of opportunities. They’re really powerful devices, and the technology is improving quickly. They can chop up power basically however you want, and they aren’t constrained by the physical limitations of large generating plants. So they can respond more quickly, and, unlike physical inertia that will eventually peter out, inverters can provide a sustained response. There are even grid-forming inverters that, unlike their grid-following brethren, can black start or support an isolated island without the need for a functioning grid to rely on. We’re in the growing pains stage right now, working out the bugs that these new types of energy generation create, but if you pay attention to what’s happening in the industry, it’s mostly good news. A lot of people from all sides of the industry are working really hard on these engineering challenges so that we’ll soon come out with a more reliable, sustainable, and resilient grid on the other end.

How Do Fish Ladders Work?

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

Building a dam imparts a stupendous change to the environment, and as with any change, there are winners and losers. The winners are usually us, people, through hydropower generation, protection from flooding, irrigation for farming, and a stable water supply for populated areas. But, we've known for a long time, probably since we started building dams in the first place, that many of the losers are fish (especially migratory fish) through fragmentation of their habitat. Even in 1890, the state of Washington in the US had laws on the books requiring consideration of fish when building dams. And, not just consideration, but specific infrastructure that would allow fish around a dam if they were quote-unquote “wont to ascend.” I recently took a tour of McNary Dam on the Columbia River in Washington (operated and maintained by the U.S. Army Corps of Engineers, Walla Walla District) and an aquatic research laboratory at the Pacific Northwest National Lab to learn more about the ways we balance our own needs with those of the aquatic wildlife impacted by the infrastructure we build. You should check out that video after this one if you want to see the whole tour. But one of the biggest pieces to that puzzle was the enormous fish ladders that allowed salmon and other migratory fish to swim up and over the dam. And it got me wondering: how do engineers design a structure like this? So this video is a follow-up, a chance to dive a little bit deeper into the intersection between engineering and wildlife. I'm Grady, and this is Practical Engineering. Today we're talking about fishways.

You've probably seen a fish ladder before, but if you haven't, this one on the Oregon side of McNary Dam is just one of many designs. The way it works is simple in practice: adult fish swim upstream toward the dam. The goal is that they don't even realize the dam is there. They simply continue upstream through the fishway and out into the forebay on the other side. Water flows in one direction—fish flow in the other. But, designing a fishway isn't simple at all. In a way, it's like engineering life support systems for manned space missions: all the design criteria are biological. How fast can fish swim, and for how long? How are they motivated? How high can they jump? What temperature, dissolved oxygen, pH, and salinity can they handle? And how do all these factors vary across seasons and species? These are difficult questions to answer, and in fact, a big part of my tour at the Pacific Northwest National Laboratory was all about how scientists study exactly the limits and preferences of migratory fish. I saw the wide variety of tracking systems they use to observe the behaviors of fish in the wild and lots of different ways they study fish in a lab as well. From research like that and decades of trial and error with the fish passage systems in the real world, engineers and biologists have started to zero in on a few designs that work best.

All fish are different, which means every fishway needs to be specially designed for the particular species that they handle. At McNary, that mainly means salmonids, a group of fish species that spend most of their adult lives in the ocean but return to shallow freshwater headstreams to reproduce. Fortunately, NOAA Fisheries has a detailed Anadromous Salmonid Design Manual that boils a lot of this knowledge down. And I’ve built a scale model of a few fish ladder designs in the garage to show you how they work.

Salmon encounter all kinds of obstacles in natural streams and rivers, even ignoring the human-made ones. They're quite capable of moving upstream in a wide variety of conditions like rapids, small waterfalls and even the presence of hungry bears. Their species literally depends on it. So, the goal of a salmon fish ladder is to mimic natural conditions, to trick the salmon into thinking they're simply making their way up a section of the river, if a somewhat steep and concrete one, without delay, stress, or injury. Part of that trick is in the flow rate. In fact, the flow of water through a fishway is one of the most essential parts of the design. And like every engineering decision, it’s a balance. Every drop of water that flows through a fish ladder is a drop that isn't stored behind the dam, so it can't be used for hydropower or water supply. But the flow of water is obviously important to the fish, too. If the flow velocity is too high, the fish struggle to swim against it. And if it’s too low, there might not be enough water to swim through. But, fish not only need specific flow to swim through; they also need it to navigate. If flows through a ladder are too low, fish can become disoriented trying to find which way is upstream. And that’s especially true at the entrance. A dam stretches the entire width of a channel, but the entrance to a fish ladder usually doesn’t, so there has to be some way to draw them to the entrance. That’s called attraction flow. Salmon use the sound and turbulence of flowing water to know which direction to swim, so a big part of fish ladder design is simply encouraging the fish inside. In fact, the flow of water through the ladder itself is often not enough for attraction, so many fish ladders have auxiliary water systems. At McNary, two enormous pumps draw water from the tail race of the dam up and into the entrance channel just so it can fall back down, creating the sound and hydraulic conditions required for salmon to find their way in. In addition, a huge valve and conduit system under the fish ladder pulls additional water from the forebay and releases it at an intermediate point down the ladder. Both of these systems provide some redundancy (since no piece of infrastructure can operate 24/7) and operators some control over the conditions along the entire length of the fishway, ensuring it can always mimic ideal conditions for the fish. But once they’re in, another challenge begins.

Dams are tall. At least, a lot of them are. And most fish can't climb actual ladders. They can't walk upstairs, and although there are some fish elevators, they’re a lot more complicated and usually less efficient than a system that allows fish to swim in a somewhat natural channel. So, the overall hydraulic design of most fishways is to break up that elevation into manageable “chunks” that fish can navigate at their own pace. They need to go kind of horizontal, but still make their way upward over the dam. A steeper channel is shorter, but it can make the water flow too quickly. A shallower channel has slower flow, but it’s a longer distance, increasing the cost and complexity of getting up to the top. So, for salmon, at least, the engineers and biologists have generally settled on something in between (usually a 10-15% slope) that breaks up the total height into passable increments with some kind of baffle.

The simplest and oldest fishways are called “pool and weir” designs. The idea here is that fish can use a burst of energy to swim up the fast moving flow over the weir and then rest in the pool above. When they’re ready, they swim up the next one, and so on. Nothing like a grown man playing with fish in his garage. Lots of fish can handle this no problem, but not all species can manage the challenge of swimming up a high velocity jet of water over and over again. Pool and weir designs are generally considered one of the less effective designs because they can limit the species and fitness of the fish that can ultimately make it through. Many of the newer fishways use more sophisticated geometry to try and address that shortcoming.

The fish ladder at McNary modifies the concept a little bit by breaking the weir into two parts with a non-overflow section in the center and including submerged holes through each baffle, called orifices. This design provides a wider variety of flow conditions, allowing more types of fish to find their way to the top. McNary even sees a good number of Lamprey, a jawless fish species with similar migratory behavior to salmon, pass through the ladder each year. Most of the salmon prefer to use the submerged orifices rather than jump over the top. My model isn’t quite scaled to my toy fish, so I’ll demonstrate that here with some movie magic.

This particular configuration is sometimes called an Ice Harbor design because it was first implemented at Ice Harbor Dam on the Snake River, just upstream from McNary. Both the pool-and-weir and Ice Harbor designs have a major limitation in that they’re sensitive to the water level above the dam. Small changes in the forebay can significantly alter the amount of flow passing through the ladder just due to the hydraulics of weirs and orifices. So the designs only work when the reservoir or forebay above a dam is regulated to a tight margin. McNary has several large crest gates that can be used to control this, but that’s not always feasible. One type of fish ladder solves it in an interesting way.

Vertical slot fishways are exactly what they sound like. Instead of a weir or orifice, they use a slot along the entire height of the baffle. That makes it possible for fish to move upstream under a wide variety of flow conditions. When I remove some of the stoplogs in my model to lower the level, the vertical slot baffles continue working in essentially the same manner. Plus, the velocity is fairly consistent from the top to the bottom of the slot, giving fish ample opportunity to pass through each one. The protrusion on the upstream face creates a gentle area for fish to rest if they need it.

The big question with these designs and all artificial fishways is this: how well do they work? Again, a difficult question to answer, especially considering that many are designed with a specific group of species in mind. The fish ladders along the Columbia and Snake rivers are surprisingly good for salmon. One study found that 97% of Chinook, Sockeye, and steelhead that entered a dam tailrace made it up the ladder and into the forebay. But, millions of dollars of engineering, research, and testing were put into those structures because of the huge cultural and economic value of these fish in the region. The results vary wildly for other species or other dams.

The primary way we know this is through tagging fish, introducing them downstream of a fishway, and measuring how many make it to the upstream side. That type of study comes with all kinds of complications, and of course, it’s impossible to compare those numbers to how many fish might make it upstream if there were no dam in the first place. The Pacific Northwest National Lab showed me some of the mind-bogglingly tiny tags that can be implanted into fish and the fascinating tools they use to track them, so again, check out that other video if you want to learn more about the process. One of the simplest ways we use to measure the effectiveness of fishways is to count the number and type of fish that pass through them. Many of the largest of these structures are equipped with counting stations, often a simple window into one side of the ladder where someone watches and marks the fish as they pass by. This is extremely useful data, not just used to measure effectiveness but to keep track of overall fish migration year over year, and a lot of it is available online. But even this requires a little ingenuity.

Most fishways are too wide to see from one side to the other in the sometimes murky water, so it’s necessary to funnel fish toward the window. Sloped grates called pickets allow water to flow through while the fish are corralled to the counting station. Even these pickets can discourage fish from continuing upstream, so fish ladder designs have to consider whether they’re really necessary. A trash rack at the upstream exit is usually installed to keep debris from getting into the ladder and clogging up the baffles. Fish swim through the rack and into the forebay to continue their upstream journey.

Of course, this is a drop in the bucket of all that it takes to manage fish passage. You won’t be surprised that adult salmon migrating upstream is a small subset of the vast array of challenges in getting fish around the barriers we build. Even McNary has juvenile passage facilities for younger fish traveling downstream and another fish ladder on the Washington side with a totally different design. Dams worldwide, particularly those installed where migratory fish species live, often have significantly different systems custom-designed for the species needing through. This is important. It’s not just for biodiversity’s sake, although that’s pretty important on its own. We depend on these fish for food, for recreation, for cultural identity, and more. So, we’re constantly innovating. I mentioned fish elevators and locks. In some cases, we do the migrating for the fish by barging up or downstream. You may have seen viral videos of the Whooshh fish cannon that aims to make fish passage possible where traditional ladders aren’t feasible. We’ve even tried dropping fish from airplanes. It didn’t work very well, by the way. All this to say, it’s something we care deeply about. For better or worse, a big part of engineering is fixing the problems we created through engineering of the past when we either didn’t know or didn’t care about the impacts our projects could cause. Everyone has a different perspective about what it means for humanity to live harmoniously with all the other life we share the planet with. I think it’s fascinating how those ideas and endeavors trickle down through engineering into the real world.

How the Hawaiian Power Grid Works

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

In January of 2024, right on the heels of a serious drought across the state, a major storm slammed into the Hawaiian islands of Oahu and Kauai. Severe winds caused damage to buildings, and heavy rain flooded roadways. At the Waiau Steam Turbine plant, the rain reached some of the generator unit controls, tripping two units and knocking 100 megawatts of power off the tiny grid (roughly 10% of demand). The overcast weather also meant solar panels weren’t producing much electricity, and the colossal battery systems at Kapolei and Waiawa were running out of juice. Other generating units were out of service due to maintenance scheduled during the cool winter months when power demands were lowest. Then, the H-POWER trash-to-energy plant tripped offline as well. By the evening of January 8th, all of Hawaiian Electric’s power reserves on Oahu were depleted, and it was clear that they weren’t going to have enough generation to meet all the needs. And if you can’t increase supply, the only other option is to force a reduction in demand.

At around 8:30 PM, the utility implemented rolling outages across the island of Oahu to bring power demands down to a manageable level. For about 2 hours, the utility blacked out different sections of the island for 30 minutes each to minimize the inconvenience. Twice since then, as of this writing, rolling outages have been forced on Hawaii Island from unexpected trips at generators and scheduled maintenance at backup facilities, making them unavailable to pick up the slack.

When we say “power grid” we’re used to imagining interconnections that cover huge areas and serve tens to hundreds of millions of people. But populated islands need a stable supply of electricity too. Those recent power disturbances highlight some really interesting challenges that come from building and operating a small power grid, so I thought it would be fun to use the 50th state as a case study to dive into those difficulties. I’m Grady, and this is Practical Engineering. Today we’re talking about the Hawaiian power grid.

Really, I should say Hawaiian power grids, because each populated island in the state has its own separate electrical system. Around 95% of customers are served by a single utility, Hawaiian Electric, which maintains grids on Oahu, Maui, Hawaii Island, Lanai, and Molokai. Kauai is the only island with its own electric cooperative. There have been a few proposals and false starts to connect the islands through undersea transmission cables and form a single grid. It is an enormous challenge to install and maintain cables of that depth and distance. When you add in the volcanic and seismic hazards of the area and the sensitive ecology of the surrounding ocean, so far, no one has figured out how to make it feasible. So, each island has its own power plants, high-voltage transmission lines, substations, and distribution system entirely disconnected from the others. And that makes for some interesting challenges.

“Reliability” is the name of the game when it comes to running an electrical grid. It’s not that complicated to build generators, transmission lines, transformers, et cetera. What’s hard is to keep them all running 99.9% of the time, day and night, rain or snow. Yeah, some parts of Hawaii occasionally get snow. This is a graph of a typical reliability curve that helps explain why it’s a challenge. At the left end of the curve, you can get big increases with a small investment. But the closer you get to 100 percent uptime, each increment gets a lot more expensive. It really boils down to the fact that, in many ways, reliability comes from redundancy. When something goes wrong, you need flexibility to keep the grid up. But, in practice, that means you have to pay for and maintain equipment and infrastructure that rarely gets used, or at least not to its full capacity.

Hopefully, it’s clear that the graph I showed is idealized. It’s much harder to put concrete numbers to the question. The random nature of problems that arise, our inability to predict the future, and the fact that everything in a bulk power system is interconnected all make it practically impossible to know how much investment is required to achieve any incremental improvement in reliability. But it’s useful anyway because the graph helps clarify the benefits of a large power grid, also known as a “wide area interconnection.”

For one, it smooths out demand. One part of a region may have storms while another has good weather. From east to west, the peak power demand comes at different times. Some areas get sun, some get shade. But overall, demands average out and become less volatile as the grid gets bigger geographically. Larger interconnections also have more redundant paths for energy to flow, reducing the impacts of major equipment problems like transmission line outages. They have more power plants, again creating redundancy and making it easier to schedule offline time to maintain those facilities. And, the power plants themselves can be bigger, taking advantage of the economies of scale to make energy less expensive and more environmentally beneficial. Finally, larger areas have more resources. Maybe it’s windy over here, so you can take advantage and build wind turbines. Maybe this area has lots of natural gas production, so you can produce power efficiently without having to pay for expensive fuel transportation. In general, a wide area interconnection allows the costs of equipment, infrastructure, resources, and operations to be shared, making it easier to keep things running reliably. Hawaii has none of that.

Roughly 75% of the electric power in the state currently comes from power plants that run on petroleum. There are no oil or natural gas reserves in Hawaii, which means the vast majority of power on the islands comes from fuel imported from foreign countries. That makes the state very susceptible to factors outside of its control, including international issues that affect the price of oil. Each island has only a handful of major power plants and transmission lines. And when storms happen, they often hit the entire place at once. It’s easy to see why retail energy costs in Hawaii are around 3 times the average price paid across the US. Every increment of reliability costs more than the one before it, and each island has no one else to share those costs with. So, they get passed down to consumers. But, it’s not just that the grids are small.

The bulk of the remaining roughly 25% of Hawaii’s electric power not produced in oil-fired power plants comes from renewable sources: wind, solar, and a single geothermal plant. This has the obvious benefit of reducing CO2 emissions, but it also reduces the state’s exposure to the complexities of the fuel supply chain and price volatility, taking advantage of resources that are actually available on the islands. But, renewable sources come with their own set of engineering challenges, particularly when they represent such a large percentage of the energy portfolio.

Of course, renewable sources are intermittent. You don’t get power when the wind doesn’t blow or the sun doesn’t shine. That sporadic nature necessitates options for storage or firm baseload to make up the difference between supply and demand. It also makes it more complicated to forecast the availability of power to plan ahead for maintenance, fuel needs, and so on. And, it requires those storage facilities or baseload plants to ramp down and up very quickly as the sun and wind come and go. But that’s not all. Solar and wind sources are also considered “low-inertia”. Thermal and hydroelectric power plants generally use enormous turbines to generate electricity. Those big machines have a lot of rotational inertia that stabilizes the AC frequency. The frequency of the alternating current on the grid is basically its heartbeat. It’s a measure of health, indicating whether supply and demand are properly balanced. If frequency starts to deviate too much, equipment on the grid will sense that something’s wrong and disconnect themselves to prevent damage. The same is true for lots of industrial equipment and even consumer devices. When conditions on the grid fluctuate - say a transmission line or generator suddenly trips offline - the rotational inertia in those big spinning turbines can absorb the changes and help the grid ride through with a stable frequency. Solar panels and most wind turbines connect to the grid through inverters. Instead of heavy spinning machines creating the alternating current, they’re basically just a bunch of little switches. That means disturbances can create a faster and more significant effect on the grid, reducing the quality of power and making it more difficult to keep things stable. I’m planning a deep dive into how inverter-based energy sources work, so stay tuned for that in a future video. But, it gets even more complicated than that.

Of all the renewable energy on the Hawaiian islands, about half currently comes from small-scale solar installations, like those on residential and commercial rooftops. They’re collectively known as “distributed energy resources.” This has the obvious benefit of bringing resources closer to the loads, reducing strain on transmission lines. It also takes advantage of space that is already developed and builds capacity on the grid without requiring the utility to invest in new facilities. But, distributed sources come with tradeoffs. Most parts of the grid are built for power to flow in one direction, so injecting electricity at the downstream end can create unexpected loads on circuits and equipment not designed to handle it. Distributed sources also affect voltage and frequency, since something as simple as a cloud passing over a neighborhood can dramatically swing the flow of power on the network. The inverters on small solar installations are generally dumb. And I’m using that as a technical term. They can’t communicate with the rest of the grid; they only respond based on what they can measure at the point of connection. The grid operator doesn’t get good data on how much power the distributed sources are putting into the grid, and they have little control over those inverters. They can’t tell them to reduce power if there’s too much on the grid already or increase power to provide support. And inverters, especially consumer-grade equipment, can behave in unexpected and unintended ways during faults and disturbances, magnifying small problems into larger ones.

Those inverters can also make the grid more vulnerable to cyberattacks since their security depends on individual owners. It’s not hard to imagine how someone nefarious could take advantage of a large number of distributed sources to sabotage parts of the grid. And finally, distributed resources affect the revenue that flows into the utility, and this can get pretty contentious. The rates a customer pays for electricity cover a lot of different costs, many of which don’t really evaporate on a kilowatt-per-kilowatt basis if you remove that demand from the grid. Fixed costs like maintenance of infrastructure still come due, even if that infrastructure is being used at a lower capacity on sunny days. With net metering, it gets even more complicated to figure out how much that power injected into the grid is really saving, not to mention how those savings should be distributed across the customer base.

And, these challenges are only becoming more immediate. Hawaii’s Clean Energy Initiative, launched in 2008, set a goal of meeting 70 percent of its energy needs through renewables and increased efficiencies by 2030. In 2014, they doubled down on the commitment, setting a goal of completely eliminating fossil fuel use by 2045. That would take them from one of the most fossil-fuel-dependent states in the US to the most energy-independent. And, they’ve taken some big steps toward that goal. Renewable generation has gone from less than 10% to about 25% of the total already, and a host of policies have been changed to create more opportunities for renewables on the grid. Solar water heaters are now required for most new homes. Rebates are available for solar installations. The only coal-fired plant in the state was controversially shut down in 2022. And, there is a big list of solar, battery storage, and biofuel turbine projects expected to come online in the near future.

For better or worse, Hawaii has become a full-scale test bed for renewables and the challenges involved as they become a larger and larger part of the grid. Many consider natural gas to be a bridge fuel to renewables, a firm resource that is generally cheaper, cleaner, and often more stable in price than other fossil fuels. But Hawaii is hoping to leapfrog the bridge. For the climate and their own energy security, they’ve gone all in on renewables, making them a leader in the world, but also forcing them to work out some of the bugs that inevitably arise when there’s no one ahead of you to work them out first. There are some really cool innovations on the horizon as Hawaii grows closer to its goal. Smart grid technologies will add sensors and communications tools to automate fault detection, recovery, and restoration, and enable power to flow more efficiently across distributed resources. Hawaiian Electric is also testing out time-of-use rates to encourage customers to shift their power use to off-peak hours, hopefully smoothing out demands and reducing the need for expensive generators that only get used for a few hours per day.

That idea really underscores the significant challenge Hawaii faces in keeping its grids operating. Improvements and capacity upgrades help everyone, but they cost everyone too, and they cost more for every additional increment of uptime. There’s no reliability menu, and kilowatt-hours don’t come a la carte. If you’re a self-sufficient minimalist or frequent nomad who isn’t bothered by the idea of intermittent power, you can’t pay a cheaper rate for less dependable service. And if you use a powered medical device or work a high-powered, always-connected job at home, you can’t pay extra for more reliability. In many ways, Hawaiians are all in it together. Drawing that line between what’s worth the investment and what’s just gilding the electric lily is tough already with such a diverse array of needs and opinions. Doing it on such a small scale, multiplied by several islands, and with such a quickly growing portfolio of renewable energy sources only magnifies the challenge. But it also creates opportunities for some really cool engineering to pave the way for a more resilient, secure, and flexible energy future, not just for Hawaii, but hopefully all the rest of us too.

How Fish Survive Hydro Turbines

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

Most of the largest dams in the US were built before we really understood the impacts they would have on river ecosystems. Or at least they were built before we were conscientious enough to weigh those impacts against the benefits of a dam. And, to be fair, it’s hard to overstate those benefits: flood control, agriculture, water supply for cities, and hydroelectric power. All of our lives benefit in some way from this enormous control over Earth’s freshwater resources.

But those benefits come at a cost, and the price isn’t just the dollars we’ve spent on the infrastructure but also the impacts dams have on the environment. So you have these two vastly important resources: the control of water to the benefit of humanity and aquatic ecosystems that we rely on, and in many ways these two are in direct competition with each other. But even though most of these big dams were built decades ago, the ways we manage that struggle are constantly evolving as the science and engineering improve. This is a controversial issue with perspectives that run the gamut. And I don’t think there’s one right answer, but I do know that an informed opinion is better than an oblivious one. So, I wanted to see for myself how we strike a balance between a dam’s benefits and environmental impacts, and how that’s changing over time. So, I partnered up with the folks at the Pacific Northwest National Laboratory (or PNNL) in Washington state to learn more. Just to be clear, they didn’t sponsor this video and had no control over its contents.They showed me so much, not just the incredible technology and research that goes on in their lab, but also how it is put into practice in real infrastructure in the field, all so I could share it with you. I’m Grady, and this is Practical Engineering. On today’s episode, we’re talking about hydropower!

This is McNary Dam, a nearly 1.5-mile-long hydroelectric dam across the Columbia River between Oregon and Washington state, just shy of 300 miles (or 470 km) upriver from the Pacific Ocean. And this is Tim Roberts, the dam’s Operations Project Manager and the best dam tour guide I’ve ever met.

“These are 1x4 hand-nailed forms that got built for the entire facility.”

But this was not just a little walkthrough. We went deep into every part of this facility to really understand how it works. McNary is one of the hydropower workhorses in the Columbia River system, a network of dams that provide electricity, irrigation water, flood control, and navigation to the region. It’s equipped with fourteen power-generating turbines, and these behemoths can generate nearly a gigawatt of power combined! That means this single facility can, very generally, power more than half-a-million homes. The powerhouse where those turbines live is nearly a quarter mile long (more than 350 meters)! It’s pretty hard to convey the scale of these units in a video, but Tim was gracious enough to take us down inside one to see and hear the enormous steel shaft spinning as it generates megawatts of electrical power. All that electricity flows out to the grid on these transmission lines to power the surrounding area.

McNary is a run-of-the-river dam, meaning it doesn’t maintain a large reservoir. It stores some water in the forebay to create the height needed to run the turbines, but water flows more or less at the rate it would without the dam. So, any extra water flowing into the forebay that can’t be used for hydro generation has to be passed downstream through one or more of these 22 enormous lift gates in the spillway beside the powerhouse.

As you can imagine, all this infrastructure is a lot to operate and maintain. But it’s not just hydrologic conditions like floods and droughts or human needs like hydropower demands and irrigation dictating how and when those gates open or when those turbines run; it’s biological criteria too. The Columbia and its tributaries are home to a huge, diverse population of migratory fish, including chinook, coho, sockeye, pink salmon, and lampreys, and over the years, through research, legislation, lawsuits, advocacy, and just plain good sense by the powers at be, we’ve steadily been improving the balance between impacts to that wildlife and the benefits of the infrastructure. In fact, just about every aspect of the operation of McNary Dam is driven by the Fish Passage Plan. This 500-page document, prepared each year in collaboration with a litany of partners, governs the operation of McNary and several other dams in the Columbia River system to improve the survival of fish along the river.

“It’s kind of a bible. It tells us how we operate. It tells us what turbine we can run, what order to run them in, what megawatts to run them at, what to do when a fish ladder or a fish pump goes out of service. So it’s a pretty good overall operating procedure for us.”

“So it’s the fish plan driving how you operate the dam?”

“Yeah, It dictates a lot of how we operate the powerhouse.”

This fish bible includes prescriptive details and schedules for just about every aspect of the dam, including the fish passage structures too. Usually, when we build infrastructure, the people who are going to use it are actual people. But in a very real sense, huge aspects of McNary and other similar dams are infrastructure for non-humans. On top of the hydropower plant and the spillway, McNary is equipped with a host of facilities meant to help wildlife get from one side to the other with as little stress or injury as possible. Let’s look at the fish ladders first. McNary has two of them, one on each side.

A big contingent of the fish needing past McNary dam are adult salmon and other species from the ocean trying to get upstream to reproduce in freshwater streams. They are biologically motivated to swim against the current, so a fish ladder is designed to encourage and allow them to do just that, and it starts with attraction water. Dams often slow down the flow of water, both upstream and downstream, which can be disorienting to fish trying to swim against a current. Also, dams are large, and fish generally don’t read signs, so we need an alternative way to show them how to get around. Luckily, in addition to a strong current, salmon are sensitive to the sound and motion of splashing water, so that’s just what we do. At McNary, huge electric pumps lift water from the tailrace below the dam and discharge it into a channel that runs along the powerhouse. As the water splashes back down, it draws fish toward the entrances so they can orient with the flow through the ladder. Some of this was a little tough to understand even seeing it in person, so I had a couple of the engineers at the dam explain it to me.

“So there’s water coming in the actual ladder and in the parallel conduit?”

“Right, right. So, it’s very complicated, huh? They’re going to approach the dam and enter from one of three spots on the Oregon side. There’s a north fish entrance on the north end of the powerhouse, south fish entrance on the south side of the powerhouse and there’s an adult collection channel that runs across the face.”

All these entrances provide options for the fish to come in, increasing the opportunity and likelihood that they will find their way.

“Between the regulating weirs on the north end, the regulating weirs on the south end and those floating orifices here, you back up that water. You need a massive amount of water to keep that step, that whole corridor.”

“I see.”

Once they’re in, they make their way upstream into the ladder itself. Concrete baffles break up the insurmountable height of the dam into manageable sections that fish can swim up at their own pace. Most of the fish go through holes in the baffles, but some jump over the weirs. There’s even a window near the top of the ladder where an expert counts the fish and identifies their species. This data is important to a wide variety of organizations, and it’s even posted online if you want to have a look. Once at the top, the fish pass through a trash rack that keeps debris out of the ladder and continue their journey to their spawning grounds.The goal is that they never even know they left the river at all, and it works. Every year hundreds of thousands of chinook, coho, steelhead, and sockeye make their way past McNary Dam. If you include the non-native shad, that number is in the millions.

“These pictures helps tremendously.”

And it’s not just bony fish that find their way through. Some of the latest updates are to help lamprey passage. These are really interesting creatures!

“I mean, in some parts of the country, they’re like, invasive. People want to get rid of them. Here, we’re trying to nurture them along because they’re a native uh, species, so there are some small changes we’ve been doing um, to try and make those make passage for lamprey more successful.”

I’m working on another video that will take a much deeper look at how this and other fish ladders work, so stay tuned for that one, but it’s not the only fish passage facility here. Because what goes up, must come down, or at least their offspring do (most adult salmon die after reproducing). So, McNary Dam needs a way to get those juvenile fish through as well. That might sound simple; thanks to gravity, it’s much simpler to go down than up. But at a dam, it’s anything but.

“And the way I explain to them is the adults are mission oriented. They’re coming back to spawn. The juveniles are just kinda dumb kids riding the wave of the ocean. I mean honestly, that’s what they’re doing. The main focus has been centered around the juveniles migrating out, right? How do we get the majority of them out? And so, when they’re coming down and they’re approaching the structure, uh, they got two basic paths to take, either the spillway or the powerhouse.”

I definitely wouldn’t want to pass through one of these, but juvenile fish can make it through the spillway mostly just fine. In fact, specialized structures are often installed during peak migration times to encourage fish to swim through the spillway. McNary Dam has lift gates where the water flows from lower in the water column. But salmon like to stay relatively close to the surface and they’re sensitive to the currents in the flow. Many dams on the Columbia system have some way to spill water over the top, called a weir, that is more conducive to getting the juveniles through the dam.

The other path for juveniles to take is to be drawn toward the turbines. But McNary and a lot of other dams are equipped with a sophisticated bypass system to divert the fish before they make it that far. and that all starts with the submersible screens. These enormous structures are specially designed with lots of narrow slots to let as much water through to the turbines while excluding juvenile fish. They are lowered into place with the huge gantry crane that rides along the top of the power house. Each submersible screen is installed in front of a turbine to redirect fish upwards while the water flows continues on. Brushes keep them clean of debris to make sure they fish don’t get trapped against the screen. They might look simple, but even a basic screen like this requires a huge investment of resources and maintenance, because they are absolutely critical to the operation of the dam.

“...incredibly labor intensive screens, we spend a lot of time cause, you know, you saw those brushes running up and down them. They’ve got submerged gearboxes, submerged motors, submerged electrical.”

“Oh my gosh.”

“Yeah, every December we pull them out for four months, we, we work on fish screens. Not to mention, so like, and if there’s a problem, these are a critical piece of equipment here, um, during fish passage season if that, if something goes wrong with that screen, this turbine has to shut down. You can’t run them without it.”

Once the fish have been diverted by the screens, they flow with some of the water upward into a massive collection channel. This was originally designed as a way to divert ice and debris, but now it’s basically a fish cathedral along the upstream face of the dam.

“Pretty cool huh?”

“That’s amazing!”

The juveniles come out in these conduits from below. Then they flow along the channel, while grates along the bottom concentrate them upward. Next they flow into a huge pipe that pops out on the downstream face of the dam. Along the way, the juveniles pass through electronic readers that scan any of the fish that have been equipped with tags and then into this maze of pipes and valves and pumps and flumes. In the past, this facility was used to store juveniles so they could be loaded up in barges and transported downstream. But over time, the science showed it was better to just release them downstream from the dam. Every once in a while, some of the juveniles are separated for counting so scientists can track them just like the adults in the ladder. Then the juveniles continue their journey in the pipe out to the middle of the river downstream.

Avian predation is a serious problem for juveniles. Pelicans, seagulls, and cormorants love salmon just like the rest of us. In many cases, most of the fish mortality caused by dams isn’t the stress of getting them through the various structures, but simply that birds take advantage of the fact that dams can slow down and concentrate migrating fish. This juvenile bypass pipe runs right out into the center of the downstream channel where flows are fastest to give the fish a fighting chance, and McNary is equipped with a lot of deterrents to try and keep the birds away.

All this infrastructure at McNary Dam to help fish get upstream and downstream has changed and evolved over time, and in fact, a lot of it wasn’t even conceived of when the dam was first built. And that’s one of the most important things I learned touring McNary Dam and the Pacific Northwest National Lab: the science is constantly improving. A ton of that science happens here at the PNNL Aquatics Research Laboratory. I spent an entire day just chatting with all the scientists and researchers here who are advancing the state of the art.

For example, not all the juvenile salmon get diverted away from those turbines. Some inevitably end up going right through. You might think that being hit by a spinning turbine is the worst thing that could happen to a fish, but actually the change in pressure is the main concern. A hydropower turbine’s job is to extract as much energy as possible from the flowing water. In practice, that means the pressure coming into each unit is much higher than going out, and that pressure drop happens rapidly. It doesn’t bother the lamprey at all, but that sudden change in pressure can affect the swim bladder that most fish use for buoyancy. So how do we know what that does to a fish and how newer designs can be safer? PNNL has developed sensor fish, electronic analogs to the real thing that they can send through turbines and get data out on the other side. Compare that data to what we already know about the limits fish can withstand (another area of research at PNNL), and you can quickly and safely evaluate the impacts a turbine can have.

What’s awesome is seeing how that research translates into actual investments in infrastructure that have a huge effect on survivability. New turbines recently installed at Ice Harbor Dam upstream were designed with fish passage in mind to reduce injury for any juveniles that find their way in. One study found that more than 98% of fish survived passing through the new turbines, and nearly all the large hydropower dams in the Columbia river system are slated to have them installed in the future. And it’s not just the turbines that are seeing improvements. I talked to researchers who study live fish, how they navigate different kinds of structures, and what they can withstand. Just the engineering in the water system to keep these fish happy is a feat in itself. I talked to a coatings expert about innovative ways to reduce biological buildup on nets and screens. I talked to an energy researcher about new ways to operate turbines to decrease impacts to fish from ramping them up and down in response to fluctuating grid demands.

“It doesn’t have to be that, you know, what’s good for the grid is necessarily bad for the fish.”

“Exactly.”

And I spent a lot of time learning about how we track and study the movement of fish as they interact with human made structures. Researchers at PNNL have developed a suite of sensors that can be implanted into fish for a variety of purposes. Some use acoustic signals picked up by nearby receivers that can precisely locate each fish like underwater GPS. Of course, if you want to study fish behavior accurately, you need the fish to behave like they would naturally, so those sensors have to be tiny. PNNL has developed miniscule devices, so small I could barely make out the details. You also want to make sure that inserting the tags doesn’t injure the fish, so researchers showed me how you do that and make sure they heal quickly. And of course, those acoustic tags require power, and tiny batteries (while extremely impressive in their own right) sometimes aren’t enough for long-term studies. So they’ve even come up with fish-powered generators that can keep the tags running for much longer periods of time. A piezoelectric device creates power as the fish swims… and they had some fun ways to test them out too.

Of course, migratory fish aren’t the only part of the environment impacted by hydropower, and with all the competing interests, I don’t think we’ll ever feel like the issue is fully solved. These are messy, muddy questions that take time, energy, and big investments in resources to get even the simplest answers.

“It’s really, it’s a complicated question. If you want to look at overall survivability from point A to point B, you can do that. But you’ve got to start talking about species. Is it a spring? Is it a fall? Is it a chinook? Is it a steelhead? Cause we have different models and studies that have been done. So it varies from species to species. People ask that question. I get really hesitant to respond, because I’m like, you don’t know how complicated a question you’re asking. You want to simplify it into one little number, and it’s not that simple.”

The salmon pink and blue paint in the powerhouse at McNary really sums it up well, with the blue symbolizing the water that drives the station, and the pink symbolizing the life within the water, and its environmental, economic, and cultural significance. This kind of balancing act is really at the heart of what a lot of engineering is all about. I’m so grateful for the opportunity to see and learn more about how energy researchers, biologists, ecologists, policy experts, regulators, activists, and engineers collaborate to make sure we’re being good stewards of the resources we depend on. I think Alison Colotelo, the Hydropower Program Lead at PNNL put it best:

“When you think about salmon and why we need to protect them, why we need to put all this money into understanding how do we, how do we coexist with our energy needs. It's because they're important from an ecological perspective, right?For the nutrients that they're bringing back, from an economic perspective, from a cultural perspective and if the salmon go away then so do a lot of other things.”