AI shows climate change-driven sea-level rise could trigger mass migration to cities inland
- January 22, 2020
- University of Southern California
- A new study uses machine learning to project migration patterns resulting from sea-level rise. Researchers found the impact of rising oceans will ripple across the country, beyond coastal areas at risk of flooding, as affected people move inland. Popular relocation choices will include land-locked cities such as Atlanta, Houston, Dallas, Denver and Las Vegas. The model also predicts suburban and rural areas in the Midwest will experience disproportionately large influx of people relative to their smaller local populations.
When Hurricane Harvey slammed into the Texas coast in 2017, displaced residents flocked inland, trying to rebuild their lives in the disaster’s aftermath. Within decades, the same thing could happen at a much larger scale due to rising sea levels, says a new study led by USC Computer Science Assistant Professor Bistra Dilkina.
The study, published in PLOS ONE, Jan. 22, is the first to use machine learning to project migration patterns resulting from sea-level rise. The researchers found the impact of rising oceans will ripple across the country, beyond coastal areas at risk of flooding, as affected people move inland.
In the US alone, 13 million people could be forced to relocate due to rising sea levels by 2100. As a result, cities throughout the country will grapple with new populations. Effects could include more competition for jobs, increased housing prices, and more pressure on infrastructure networks.
“Sea level rise will affect every county in the US, including inland areas,” said Dilkina, the study’s corresponding author, a WiSE Gabilan Assistant Professor in computer science at USC and associate director of USC’s Center for AI for Society.
“We hope this research will empower urban planners and local decision-makers to prepare to accept populations displaced by sea-level rise. Our findings indicate that everybody should care about sea-level rise, whether they live on the coast or not. This is a global impact issue.”
According to the research team, most popular relocation choices will include land-locked cities such as Atlanta, Houston, Dallas, Denver and Las Vegas. The model also predicts suburban and rural areas in the Midwest will experience disproportionately large influx of people relative to their smaller local populations.
Predicting relocation areas
Sea-level rise is caused primarily by two factors related to global warming: added water from melting ice sheets and glaciers and the expansion of sea water as it warms. Within just a few decades, hundreds of thousands homes on the US coast will be flooded. In fact, by the end of the century, 6 feet of ocean-level rise would redraw the coastline of southern Florida, parts of North Carolina and Virginia and most of Boston and New Orleans.
To predict the trajectory of sea-level rise migration, the researchers took existing projections of rising sea levels and combined this with population projections. Based on migration patterns after Hurricane Katrina and Hurricane Rita, the team trained machine learning models — a subset of artificial intelligence — to predict where people would relocate.
Sea-level rise in the U.S. won’t only affect people living on the coasts—as homes flood in Florida and New Jersey, it may trigger mass migration inland, potentially making housing more expensive and jobs harder to find in other areas. A new study uses AI to map where people may go.
“We realized that while existing research has studied the effects of sea-level rise on coastal populations, it hasn’t considered the farther-reaching effects,” says lead author Caleb Robinson, a doctoral scholar at Georgia Tech who is currently doing research at USC. “In our study, we really aim to get at the indirect effects that sea-level rise can have through migration.”
Around 13 million Americans could be forced to move by the end of the century if the sea level rises six feet, headed in large numbers to the Midwest and cities like Atlanta, Houston, and Dallas. The study looked at projections for sea-level rise and population growth, and then trained a machine-learning model using data about where people moved after Hurricane Katrina and Hurricane Rita. While another previous study did also predict climate migration, it was based partly on how people move under normal circumstances. The new work is designed to come closer to what happens when people are fleeing the effects of climate change. “We really believe that dynamics when it’s forced migration, versus business as usual migration, will be different,” says Bistra Dilkina, a computer science professor at USC and one of the authors of the paper.
Of course, this only accounts for one impact from climate change. People may also move (and in different patterns) because of wildfires, long-term droughts, and other climate disasters. And they may overlap: Houston, one of the cities that the map suggests may see a large influx of climate migrants, saw more than 200,000 homes damaged from the flooding caused by Hurricane Harvey, a storm that saw record rainfall because of the changing climate. The framework created in the study, however, could also be used to look at where other types of disasters may force people to move. “It will basically be a matter of whether we have enough data to be able to say we have a good estimate of how people behave in these situations,” says Dilkina. “As we have seen, natural disasters are increasing in number and severity, and I think we will be getting more and more data, unfortunately.”
Cities that expect to see more climate migrants can use this type of research to begin to think about how to prepare for a growing population—something that few are doing now, although the mayor of Buffalo, New York, called it a “climate refuge” city in his state of the union address last year, and the local government thinks that potential migrants could help revive the city. Other cities may need to think about changes in infrastructure or housing.