The statistics and physics of midlatitude heat extremes

Karen McKinnon
Thursday, April 22, 2021 · 11:00 am
ESB 5104-06
Hosted by
Rachel White

A clear consequence of climate change is increasing global temperatures, with greater warming over land than ocean. The greatest impacts from continental warming occur due to extreme events; however, we do not have a good understanding of how to map from this large-scale picture to regional extremes. In this talk, I will present two different studies focused on better understanding the physics and statistics of heat extremes. First, we will explore how daily summer temperature distributions have changed over the historical record in order to assess how the tails of the distribution are changing with respect to the middle. Across the Northern Hemisphere, we find that changes in summer temperatures can primarily be described by a simple shift, but that some regions such as eastern Europe show greater warming for hot extremes. Cognizant that these observed changes reflect both a deterministic ("forced") response to human influence on the climate and a random sampling of internal variability, we also quantify trends in a climate model large ensemble, which suggests the dominance of the simple shift behavior. While heat extremes have commonly been described primarily by temperature alone, their impacts are mediated by humidity, highlighting the importance of jointly modeling changes in temperature and humidity extremes. Using a newly developed semiparametric quantile smoothing splines model, we identify important dependencies between changes in temperature and humidity extremes, including a decrease in specific humidity on the hottest days in the American Southwest. This trend is in contrast to the general expectation of increases in specific humidity as global temperatures warm, and emerges due to decreases in soil moisture, which are an important local moisture source. While it is unclear whether this trend will continue due to climate model biases and divergent projections, the result underscores the importance of a process-level understanding of regional climate trends, particularly for high-impact extremes.