Research

I am broadly interested in applications of statistics, data science, and math to problems in the climate sciences, with a particular focus on extreme weather and climate events. Below are a few projects I’ve been working on.

Quantifying uncertainty in extreme bivariate events¶

James Butler (UC Berkeley), Jon McAuliffe (UC Berkeley, The Voleon Group), Michael Wehner (Lawrence Berkeley National Lab)

Many extreme weather and climate events are impactful not just due to extreme conditions in a single climate variable, but rather extreme conditions in multiple variables that co-occur and amplify the impact of any individual variable. For example, in heat waves, excessive temperatures are of course hazardous to human health, but simultaneously high levels of humidity can drastically increase human mortality. Further, wildfires are particularly dangerous when conditions are not just dry, but also simultaneously windy. Building on previous statistical work to estimate rare bivariate events given data from an unknown probability distribution, we propose a fully non-parametric uncertainty quantification procedure to construct confidence regions for such rare events, providing a more complete assessment of the risk of extreme bivariate events and informing decision-making surrounding climate risk.

• AGU23, poster

• Preprint coming Spring 2026.

Linking Antarctic Atmoshperic River Characteristics with Landfalling Impacts¶

James Butler (UC Berkeley), Michelle Maclennan (British Antarctic Survey), Jon McAuliffe (UC Berkeley, The Voleon Group), Fernando Pérez (UC Berkeley)

Atmospheric rivers are powerful storm systems characterized by long narrow bands of intense moisture transport that bring intense precipitation when they make landfall. In Antarctica, they are sort of a mixed bag: in the current climate, they bring intense snowfall which can offset sea level rise, yet they also bring warm air from the midlatitudes onto the interior of the ice sheet, causing melt events and the largest temperature anomaly ever recorded. We construct a catalog of individual landfalling atmospheric river events, develop an end-to-end cloud-based workflow to extract atmospheric quantities from the storms, and apply cutting-edge machine learning models to explore the associations between landfalling characteristics and average AR impacts, as well as extreme/rarely observed AR impacts.

• AGU24, poster

• 3rd Antarctic Atmospheric Rivers Group (AARG!) Workshop, oral presentation slides

• AGU25 (science), oral presentation slides

• AGU25 (cloud-based data workflows), oral presentation slides

• Preprint coming Spring 2026

Extreme Temperature Time Series and Ice Melt¶

James Butler (UC Berkeley), Fernando Pérez (UC Berkeley), Jon McAuliffe (UC Berkeley, The Voleon Group)

When communicating the potential impacts of global warming on ice melt and sea-level rise, much of the conversation is concerned with increases in the average global temperature. However, it’s not increases average temperature events that are devastating for ice sheets and glaciers, but rather rarely observed extreme temperature events. To demonstrate the importance of surrounding the conversation around changes in extremes versus the average, we simulate the melting of ice using a very simple melt model in response to temperature time series of varying extremal properties.

Other interests¶

I am also interested AI-based weather prediction models and their use for extreme weather events, having worked on a two-week long project with other researchers at Rossbypalooza 2024, a summer school hosted at UChicago.