Applying Research Results to the Real World Adds up for Math Professor

Elisa Perrone is an assistant professor in the Department of Mathematical Sciences. She studies mathematical tools called copulas, which can make weather forecasts more accurate but have also been blamed for the 2008 global financial crisis. A native of Lecce in southern Italy, Perrone earned a Ph.D. in mathematics at Johannes Kepler University Linz in Austria. After a postdoctoral appointment at MIT, she joined the Kennedy College of Sciences faculty in 2019. 

A. My research is in a class of statistical models called copulas, which capture how various random phenomena interact with each other — for example, how rainfall totals of two different months are dependent on each other. The combination of mathematics research and the potential applications of my results is what I really, really like in my work. The popularity of copulas began when people started using them to model bank defaults and other things related to finance. There was a famous article in Wired magazine called “Recipe for Disaster: The Formula that Killed Wall Street” that states the reason why we had the big financial crisis in 2008 was because people chose the wrong copula to model joint defaults of banks. I fell in love with this class of functions. They are powerful tools, and this is a relatively new type of research in statistics — there’s so much that still can be done. 

A. Getting good weather predictions is challenging. People usually obtain good forecasts by using not just a single forecast; they assume the forecast is a random phenomenon, so they try to use a bunch of equations to get an ensemble forecast. The problem is that these are raw forecasts; they need to be statistically corrected. There are good correction methods that can be applied to single weather stations, for example, or to single weather variables, like the rainfall total in May or June. But there is always this problem — how do I join them together? We have infinite ways of putting together three individual weather variables. 

A. Exactly. Or they could be independent — maybe there’s no interaction at all. So copulas can be used to reconstruct a source of dependence between different weather variables. I can use them to correct multivariate weather forecasts.

 It’s interesting that in the realm of pure mathematics, you can say something about how the real world behaves. 

There’s so much overlap between different problems. They might be the same problem in the end, if we find a good way to communicate across fields. With these copulas, I found connections with statistics, but also other fields of math. It’s one of the reasons why I like my research, to explore these kinds of connections. 

 A. UMass Lowell has a very good Department of Environmental, Earth and Atmospheric Sciences. My next step will probably be to go from weather forecasting to climate, and my goal is to establish collaborations with them. I could easily connect with people who are working in finance as well. There are plenty of opportunities at UMass Lowell.