In 1979, the Army Corps of Engineers predicted that by 2014, Optima Lake, in the panhandle of Oklahoma, would have 600,000 visitors per year camping, fishing, boating, and swimming. Instead, the lake sat empty, a dry expanse of land about three miles long. Today, it is still abandoned.
The Optima Lake and Dam was originally intended to control flooding from Beaver Creek and the North Canadian River. After the $45 million it took to construct it was spent, though, the lake never filled up, and has never reached more than five percent of its capacity.
Ed Rossman, a planning branch chief for the Corps of Engineers’ Tulsa District, told a local NPR station in 2013 that the project had been led astray because planners used historical climate data to pick the spot where the dam would be built. That data was wrong by the time construction was completed, and the water no longer flowed there. “We know that the historic record may not be a good snapshot for the future,” Rossman said.
In other parts of the world, the opposite problem has occurred. A flooding barrier put up in the 1990s in the Netherlands to protect Rotterdam will likely fail 25 years sooner than was predicted when it was first built in the 1990s. Replacing the barrier will cost around a billion euros, 10 million of which would be just the cost for taking it down.
The issue here, in its different guises, is one that engineers and city planners continue to face, and which reveals an inherent problem with how we’ve planned, designed, built, and made predictions around all of our infrastructure. At its core, this problem revolves around a concept called stationarity.
Stationarity is the idea that, statistically, the past can help you predict and plan for the future—that the variations in climate, water flow, temperature, and storm severity have remained and will remain stationary, or constant.
Nearly all the infrastructure decisions with which we live have been made with the assumption of stationarity. Engineers make choices about stormwater drainage pipes based on past data of inches of rain. Bridge engineers design foundations that can withstand a certain intensity of water flow based on the severity a certain location has experienced in the past. Reservoirs are designed to hold water based on historical information about water flow, and the historical water needs of a community.
“Stationarity has been the foundational concept used in the design of water infrastructure for as long as knowledge of the past has been around,” said Chris Milly, a research hydrologist at the U.S. Geological Survey.
We need to undergo an ideological shift in how we think about infrastructure and how it interacts with the environment.
And it’s more than just water. Experts choose what materials to make power lines out of based on how hot it has been before in a given location; if the lines get too hot, they could sag or short circuit. Asphalt cracks at high temperatures, but you can design asphalt mixtures to withstand extreme heats; those mixture decisions are made based on past weather data. Train tracks, airport runways, power plants, sewage systems— they are all designed with the past climate in mind.
Yet the assumption of a stationarity world has not withstood the test of time, or of climate change. In 2019, the average temperature around the world was 1.7 degrees above the 20th century average; it was the second-warmest year ever to be measured. The planet’s temperature has been increasing steadily and the five warmest years since 1880 have all taken place since 2015. The increase in global temperature is causing temperature and weather extremes that past climate data can’t fully predict.
Experts say it means we need to undergo an ideological shift in how we think about infrastructure and how it interacts with the environment—discarding the notion that our history can dictate what we need in the future, and instead turn to more adaptable and flexible versions of infrastructure that embrace deep uncertainty.
Over the past two decades, many hydrologists and engineers have raised the alarm about how a stationary approach isn’t working anymore. In 2008, an international group of scientists (including Milly) announced in the journal Science that “stationarity is dead,” and that it was time we accept we’re living on a non-stationary planet.
How did stationarity die? We’ve changed the planet so much with carbon emissions and other human activity that the past can no longer reliably determine what will happen in the future, or guide decisions about what kind of infrastructure we will need, what size it needs to be, what material it needs to be made of, what kind of climate it needs to be able to withstand.
“When people say stationarity is dead they’re saying something pretty straightforward, which is the past is no longer a good guide to the future,” said Giulio Boccaletti, the chief strategy officer and global ambassador of water at The Nature Conservancy. “That’s pretty momentous for a sector like water, which for the last century has essentially based its designs on statistics from the past, rather than being able to predict what is going to happen. Then the issue becomes, well, OK, so how different will it be?”
Non-stationarity means that we live in a world where there is no such thing as “normal,” where every new year comes rife with uncertainty and the threat of extremes we’ve never seen before. And “stationarity cannot be revived,” the Science paper declared.
Watch more from VICE:
Engineers constantly have to make choices when designing infrastructure. If you are designing a road with a stormwater system—a pipe underneath the road that moves water away to prevent flooding—how big would you make the pipe? It would have to be a different size depending on whether you lived in San Diego, Minneapolis, or New York City.
Engineers don’t just pick a random diameter that they think sounds good—the process is systematic, said Mikhail Chester, an associate professor of civil, environmental, and sustainable engineering at Arizona State University.
“Often at the national level, there are engineering societies that will make recommendations that make their way into city codes,” Chester said. “And those say that when you design a stormwater pipe, it needs to be able to accommodate a 10-year-event, or something like that.”
A 10-year event isn’t when something happens every 10 years. It’s when an event, like an earthquake or flood, has a 10 percent chance of occurring each year. A 100-year event, rarely used when designing infrastructure, has a one percent chance of occurring each year.
If the future is going to be different, then the past isn’t so helpful.
Where do engineers get that probability? Weather stations have collected data on what happened in the past—rain, water flow, temperature—and we have records that go back 40 to 60 years. When trying to decide how big to design something, engineers will often go to the National Weather Service or the National Oceanic and Atmospheric Administration’s website. They plug in the location of the infrastructure they’re designing for, and see what the past has been like for that area. “This concept is everywhere in infrastructure,” Chester said.
It’s also linked intimately with stationarity. It’s only useful to use past data if you assume that the future will be similar to the past. “How big should my reservoir be, for example,” said Boccaletti. “The idea was, if I build a reservoir that can hold water for long enough to ride through whatever droughts there have been in the past, then surely that will also be adequate for the future.”
But if the future is going to be different, then the past isn’t so helpful. One paper on stationarity said designing infrastructure in this way was like “dancing on the tip of a needle.”
Twenty years ago, discussions about stationarity versus non-stationarity were non-existent in the context of infrastructure, Chester said. Figuring out how to design non-stationary infrastructure has only started to be discussed, researched, and published about in the last 10 years.
“Here I am today, in 2020, and I’m an engineer,” Chester said. “And I’ve literally had these conversations with engineers where they say to me, ‘Something’s not right. We’re seeing extremes happening at a greater intensity and more frequently than we’ve experienced in the past.”
To complicate matters further, infrastructure isn’t designed for only the next few years. Ideally, it will last decades. “If something is going to be there for a long time into the future, what should I be designing for exactly?” Chester said.
There’s not an easy answer to this question yet. Climate modeling is imperfect, but let’s say that one forecast estimates that flooding will get 12 percent worse. Should engineers make storm pipes 12 percent bigger?
“It doesn’t quite work that way,” Chester said. “Upsizing everything is not financially feasible. We don’t have enough money to upgrade the infrastructure that we have, let alone start making it even bigger for uncertainty in the future. This is where we stand as engineers, trying to reconcile what to do.”
When temperatures rise, there’s more evaporation, leading to more droughts in some areas, but also more moisture in the atmosphere, and so more rainfall in others. Sea levels have risen about 7 inches since the start of the 20th century. Summers are getting hotter. The rainfall during Hurricane Harvey was 15 percent more intense and three times more likely to occur because of the changes to the climate caused by human activities.
Our infrastructure is already inadequate in the face of these changes, and will, in the future, when the climate has changed even more, become even more inadequate. In 2014, for the first time, the National Climate Assessment report included a section about the ways that climate change threatened infrastructure. The authors highlighted how power, sewage, roads, drinking water are all linked to one another, and one extreme weather event could cause a cascading effect, like in 2011 when a heat wave led to 20 different infrastructure failures in only 11 minutes, affecting millions in Arizona, California, and Mexico.
“It’s not unreasonable to think that you are going to start to see more failures of assets under even normal conditions,” Chester said. “But the worst is going to be the extremes.”
These cascading events will become more likely if we continue with static infrastructure that’s designed from the past. And yet, the Trump administration’s infrastructure revitalization plan doesn’t mention the risk of climate change, nor does it discuss the necessity of anticipating the uncertainty that comes with designing infrastructure for a future that doesn’t resemble the past.
The solution is to develop infrastructure that is agile, flexible, and ultimately adaptable, rather than sturdy, unchanging, and permanent. The goal isn’t to make rickety bridges or weak pipes that need to be replaced all the time, but to shift from an ideology of rigidity to flexibility so that the infrastructure that surrounds us can be updated quickly to match our environments.
One way to achieve that is through modularity, inherently adaptable infrastructure, that can be changed more easily, and with compatibility of hardware—like being able to easily raise the height of bridges, or change the size or scope of drainage, or design infrastructure with multiple uses. One example of this is that in Kuala Lumpur, they can use their traffic tunnels as stormwater tunnels if there’s an extreme storm or heavy rainfall.
There’s a field of decision science called decision-making under “deep uncertainty,” and engineers have started their foray into this discipline; the collaboration will be crucial for the future. Simply taking the time to predict a myriad of future conditions and walking through the outcomes is helpful too, and isn’t something that reliably happens now with infrastructure design. Instead of planning funding until a project is complete, it could anticipate future changes and adaptations that need to be made in the budget from the start.
In the UK, the Thames Estuary 2100 Project was one of the first to rigorously consider deep uncertainty and climate change right from the beginning of the planning process. A large part of the plan was to upgrade or replace the Thames Barrier, which protects London from flooding due to a storm surge.
Instead of looking at what happened in the past to determine how to modify the barrier, the group planned out multiple possible options for the future, created an “adaption map” with different choices arising depending on what was going on with the climate, like a choose your own adventure book informed by the climate.
How do you plan for the future, when finally accepting that the future won’t look like the past?
Another approach is called “minimizing future regret.” “It’s a totally different way of looking at problems,” Chester said. “I don’t know how bad it’s going to be. What I do know, is that in making decisions, I want to minimize retreat about how I make those decisions. I essentially want to look at a number of scenarios of ways that I could adapt, and ask myself, ‘If I’m totally off in those scenarios, how bad is that? How much cost did I waste, how much resources did I waste, and what were the social, political, and capital concerns in doing that?’”
We need to build infrastructure that is safe to fail. Traditionally, infrastructure is designed to be fail-safe up to a certain point—those 10 or 100-year events. Anything over that, and severe consequences occur. At the moment we don’t design infrastructure with that failure in mind. Thinking about the consequences right at the beginning of the design process forces engineers and city planners to work to avoid them.
All of this doesn’t mean completely disregarding the past. “We need somehow to blend what we know from the past with what we can infer about the future,” Milly said. “Continuing to measure and observe the real system is a crucial part of this endeavor, because it gives us feedback about where the climate models might be on the mark or going astray.”
Non-stationary is a statistical problem that reflects a more existential one: How do you make decisions in the context of deep uncertainty? How do you plan for the future, when finally accepting that the future won’t look like the past?
Stationarity and non-stationarity can serve as apt metaphors to think about the climate crisis overall. We cannot continue to behave in the future as we have in the past, because we’ve changed the environment too much to do that. Even outside of infrastructure, we need to act differently than we did before. What Nietzsche wrote that “God is dead” he meant that the Enlightenment’s progression of science, philosophy, and society no longer needed to frame itself around the rules of religion and faith. Similarly, since “stationarity is dead,” we can no longer have faith in a constant, unchanging world, and the promise that a future will be just like the past.
This acceptance of uncertainty could prove useful in other domains too. Chester and his colleagues recently wrote that our inability to design adaptable and flexible infrastructure has been revealed in other sectors too, like during the COVID-19 pandemic. “With all types of infrastructure, including the medical infrastructure system, efficiency and resilience are in conflict,” they wrote.
Our medical infrastructure was only able to handle extremes it had dealt with in the past, but unable to adapt to a new, unprecedented extreme event. We weren’t able to easily scale up testing, contact tracing, or social and economic policies to help people as the country shut down. We ran out of essential medical equipment like PPE and ventilators.
“When the situation does change—and especially if the change is anomalous, high-impact, and rapid, allowing little time for adaptation—such a system will be very fragile, since the conditions to which it has been adapted no longer prevail,” the authors wrote.
Climate change, pandemics, population growth, resource demands—all of these factors reflect how fast and how much our societies are changing, and challenging us to innovate and develop along with them.
“The fact is that over the course of the next 100 years, everything is going to change because nature is changing quite significantly,” Boccaletti said. “And so we have to transition to a more adaptive and resilient system of managing.”
Follow Shayla Love on Twitter.