When Hurricane Harvey’s record-busting rains drenched Texas in August 2017, they triggered a cascade of chaos. Widespread flooding turned roads into rivers, impeding evacuations and access to emergency services. Stormwater swept up pathogens from wastewater treatment plants and toxins from Superfund sites, posing health threats. Phone and internet services failed in some areas, and 300,000 people in Texas lost power. Harvey also temporarily shut down a quarter of U.S. oil production in the Gulf of Mexico, raising gas prices.

Such scenarios—climatic events causing impacts that can themselves trigger still more chains of effects, like intersecting rows of toppling dominoes—are a key focus of the fourth National Climate Assessment (NCA), released by the U.S. federal government at the end of November. For the first time, the 300 government, academic and nonprofit experts who contribute to the report devoted an entire chapter to the under-studied but critical interaction between climate change and what are called complex systems.

The report emphasizes that scientists need to look not only at how global warming is changing natural systems but also how those changes will set off their own ripple effects through other areas—for example, how the increasing threat of drought harms agriculture, which in turn affects the economy and food availability. “Reality is complex. In a changing climate, nothing is being affected all by itself,” says Katharine Mach, a senior research scientist at Stanford University and one of the NCA authors. The complexity of these cascading effects means they can often be hard—or even impossible—to understand or predict in a meaningful way.  But that is exactly what scientists are now trying to figure out how to do.


Researchers typically study systems in relative isolation, deliberately overlooking convoluted interactions for the sake of scientific clarity. For instance, they might study what roads a heavy rain will flood—but not how a storm’s damage to communications or emergency services might interact with those road blockages, creating more knock-on effects. “You might ask what’s going on with, say, this one forest or this one agricultural crop. You draw boundaries around your system, and you’re just looking inside the box of your study,” Mach says.  That turns out to be a major weakness when scientists are trying to understand the potential risks and myriad impacts of climate change.

Here is a simple example: If researchers want to see how climate change will affect energy systems, they might simply model the effects of rising temperatures and heat waves on electricity demand. This could lead them to conclude that more power plants need to be built to keep up with higher energy demands for cooling. But a study with such a singular focus would overlook a critical pitfall: hotter temperatures also make drought conditions more likely in some areas, meaning there is less water to cool down power plants—and what water there is tends to be warmer, making it less effective for cooling. This complex interaction, which might typically be ignored, can significantly damage energy production. That is exactly what happened in Texas and the Southeast U.S. during recent droughts. Through such cases, “you realize that the problem you actually need to solve in the real world is more complex than the problem you thought you had,” says NCA co-author Anthony Janetos, director of the Pardee Center for the Study of the Longer-Range Future at Boston University. “You could spend a lot of money fixing the wrong thing.”

It is easier to see the dynamics between all these environmental and human systems in hindsight, because scientists can follow the threads of various interactions that have already occurred. But it is much trickier to anticipate them, making the task of helping communities prepare for future climate impacts that much more difficult. Research in this arena is growing, but “we’ve just begun to scratch the surface of understanding these complexities,” says Leon Clarke, a senior scientist at the Joint Global Change Research Institute in Maryland and another NCA author.

Some new studies are looking at issues such as how climate-induced shocks to agriculture affect global markets, food prices and land use; the relationship between flood risk and what flood protection measures societies decide to take; how expanding reservoirs can actually worsen water shortages during drought; and the link between rising temperatures and violence.

One such study, still under review, looks at how Hurricane Harvey affected Houston’s road networks and, consequently, people’s access to emergency services during the storm. For this study researchers used flood data and modeling to analyze whether people living in two different neighborhoods had road access to fire stations and hospitals at various times during the storm. They also looked at how much longer it would take people to travel to those emergency services if their normal routes were blocked, and whether one neighborhood—with a lower average income—saw greater disruption to its road access. “That's the chain of events that we wanted to highlight,” says Avantika Gori, a researcher of flood risk management at Princeton University. She hopes such studies will give first responders better information about where to devote their limited resources and will help city planners be proactive about events like Hurricane Harvey. They could, for example, improve evacuation plans for specific neighborhoods and give better route information to emergency vehicles when a big storm hits.


But experts note this type of research is incredibly challenging. It is hard enough to model one system on its own, let alone connect it with a series of others. Models needed to simulate various aspects of a problem (such as rainfall or how roads flood) may work at completely different time and spatial scales. Also, disparate data may need to be stitched together from different sources. And in general, the more complex a problem is, the more computer power and time it takes to run simulations. To complicate things further, a single research team may include social scientists, natural scientists, engineers and experts from other fields, all of whom use their own technical languages and methods for their work.

Constantly improving models, increasing computer power, and advances in techniques such as machine learning will help propel some progress, as will lessons learned from early studies now underway. Scientists may also need to combine simulations of future scenarios with expert interviews to help fill in forecasts with local, on-the-ground experience.

Beyond the research, scientists will also need to help decision-makers understand and make choices in light of all this complexity. For instance, since sea level rise predictions are uncertain, experts could assist a community in building an adjustable flood protection barrier. The good news is that some places, including New York City and Boston, have already begun considering complex systems as they prepare for climate change. “They’re doing a sophisticated job of modeling the interactions between sea level rise, storm surge, the particular geographies of their harbors and coastlines, and the infrastructure that is at risk,” Janetos says. “They’re considering lots of different dimensions as they think about how they’re going to respond in terms of adaptation.”

Still, experts acknowledge that ultimately they will never be able to put odds on every single possible set of interactions, Mach says. Rather it will be a matter of improving the available information to make better calls on what actions to take wherever possible. “We still need to make decisions,” she says. “We can’t let the complexity paralyze us.”