In 2011 and then 2012, Emma Levin’s hometown on Long Island was hit back to back with the one-two-punch succession of Hurricanes Irene and Sandy. The resultant damage and flooding left the then middle-school-aged Levin out of class for weeks. In the wake of such close-to-home impacts, Levin, who had always had a knack for math and hard physical sciences as a student, suddenly found herself drawn to climate and environmental sciences.

Now a second year graduate student in the Program in Atmospheric and Oceanic Sciences, Levin has dedicated herself to research in the pursuit of an increasingly urgent question: How will hurricane activity change in a warming world? 

She’s a member of The Vecchi Research Group – run by Gabe Vecchi, Princeton University Knox Taylor Professor of Geosciences, director of the High Meadows Environmental Institute and deputy director of the Cooperative Institute for Modeling the Earth System – whose members past and present have helped develop detailed climate model simulations which recreate historical hurricane activity. Using these simulations to peek at hurricanes of the past, she hopes to gain crucial insight into hurricanes of the future.

“I'm going back in time to understand past hurricane fluctuations in the context of trying to determine what will be different in the future,” said Levin.

Setting the hurricane record straight

Levin, who is pursuing a graduate certificate in Statistics and Machine Learning alongside a PhD, got her start working with climate models when she was just a high schooler. “I was in a high school research program, and I had to find a mentor outside of the school,” said Levin. “I sent 97 emails reaching out to anyone in the New York metro area studying climate because I thought it was just a really interesting topic and because I was personally affected by extreme weather.” 

The flurry of emails paid off. Levin secured a reply from Hiroyuki Murakami, a research physical scientist at NOAA’s Geophysical Fluid Dynamics Laboratory (GFDL), which is located on the Princeton University Forrestal Campus. Murakami welcomed the young Levin into his laboratory, where she got her first exposure to climate models, especially those dedicated to hurricane forecasting. “A lot of crucial data was coming from this hub in Princeton,” said Levin. “Just being able to tinker with United States climate models in high school was so cool and awesome.”

After high school, Levin earned a bachelor’s degree in applied mathematics with a certificate in energy studies at Yale, and after graduation returned to Princeton as a doctoral candidate in  the Program in Atmospheric and Oceanic Sciences (a collaboration between the University and GFDL). 

As a part of the Vecchi group, she is using models to understand how hurricane formation will change in the face of global warming. In particular, Levin is analyzing data on “seeds,” which is what scientists call storms which are the precursors to hurricanes. Seeds typically form off the coast of West Africa when the right combination of winds above warm ocean waters begin to swirl around a column of low pressure. 

Not every seed becomes a hurricane. But Levin wants to understand how climate change is impacting the different weather processes which lead to hurricane formation. “In doing so, we get a better sense of what exactly is changing hurricane frequency,” said Levin. 

So far, Levin’s results have suggested that in a warmer world, more and more of these seeds will turn into full-blown hurricanes. “In the future, the number of seeds will likely become more important in driving overall hurricane frequency,” said Levin. “So, it's important to consider seed development.”

To answer how hurricane activity will change in the future, first researchers must understand what was happening in the past. While there is satellite data dating back to the 1970s which unambiguously shows an uptick in hurricane activity in the Atlantic Ocean, fifty years is not a long enough time period to draw overarching conclusions about historical trends. Reconstructions of hurricane activity going back to the 1860s exist, but the data, which was drawn from ship tracking records, is flawed.

Levin’s work builds on that of her advisor Gabe Vecchi, which seeks to fill in the gaps in the historical data on hurricanes. “We used the 1800s ship track data along with satellite era storm data to try to map out what the likely number of missing storms was,” said Vecchi. “What we want to answer is: What is the context of this recent increase? Is it part of a longer term trend?” 

The adjustments to the historical record made by Vecchi and his colleagues revealed that the old record had severely undercounted past storms. Right now general consensus says the proportion of intense hurricanes is on the rise. But questions of how exactly frequency will change are still unanswered.

Vecchi describes the work Levin has contributed to his lab as creative, dynamic, and rigorous. But what really inspires him is Levin’s resilience as a researcher. As Vecchi tells it, Levin once spent several months trying to investigate East African rainfall before she finally had to concede the existing data wasn’t good enough and the models she was working with were wrong. “It was only after really exhaustively mapping out the fact that it was not a tractable question, given everything she knew and everything we had available at the time that she said, ‘All right, I'm going to wrap this up,’” said Vecchi. “She could have quit too early and not learned anything.” 

The inability to answer questions about East African rainfall led her to pursue her current research questions on seeds – a switch which Vecchi said has already yielded “really great, exciting answers.” 

For her part Levin said she’s thankful for “the extraordinary mentorship” of both Vecchi and Murakami, as well as now being a part of the Center for the Statistics and Machine Learning as a certificate student. Particularly since weather simulations are increasingly becoming based on machine learning. “A lot of my work is computational, and so there's a lot of statistics and analysis required,” she said. “It seems like machine learning is really the future of climate science, so that’s an avenue I hope to continue exploring in the future.”