CAREER: Attributing Extreme Weather to Stratospheric Variability Using Ensemble Forecasts

Cornell University, Ithaca NY

Investigators

Abstract

Accounts of extreme weather typically follow the logic of the "perfect storm", in which several factors come together to produce an event of exceptional severity. Following this logic the event occurred because all the factors aligned in just the right way, a rare occurrence that happened by chance. In that case we should not look for a single deterministic cause for the event but ask instead what factors contributed to it, in what ways, and to what extent did each factor increase the likelihood or severity of the event. This way of looking at contributions to an extreme event has been pioneered in "attribution" studies which quantify the extent to which climate change has increased the likelihood of warming-related extreme events. For instance one study provides evidence that the warming of the world over the past few decades has roughly tripled the odds of the extreme temperatures that occurred during the exceptionally deadly Russian heat wave of 2010. Work performed here applies the framework of extreme even attribution to events which evolve over the subseasonal timescale, meaning more than the one week timescale of typical frontal weather systems but less than a season. The focus is primarily on cold air outbreaks (CAOs), in which winds from the north cause extreme cold temperatures over the middle-latitude continents. Previous work suggests that CAOs are more likely when the polar vortex of the Northern Hemisphere stratosphere is weak, as is the case following sudden stratospheric warmings (SSWs). SSWs are in turn driven, at least in part, by the rapid amplification of upward-propagating planetary waves from the troposphere. Thus we can construct a "storyline" in which rapid amplification of planetary waves leads to an SSW which in turn leads to a CAO. The event attribution goal is then to determine how much the SSW increased the odds of the CAO and how much the planetary wave amplification increased the odds of the SSW. The project pursues these attribution questions using observations as well as simulations from models at varying levels of complexity, including a dry dynamical core model, a Linear Inverse Model (LIM), and the Community Earth System Model (CESM). A key challenge in the work is the rarity of the events of interest, which usually means that very long model simulations are needed to generate a statistically significant sample of extreme events. The project addresses this challenge using a rare-event sampling algorithm in which an ensemble of model simulations is steered toward an extreme event of interest by pruning ensemble members which are not evolving toward the extreme state and replacing them with new members spun off from ensemble members which are. The rare-event sampling algorithm dramatically reduces the amount of simulated time required to produce enough extreme events to draw reliable conclusions. The educational component of this CAREER proposal includes two activities, one of which is the development of a modular course intended to introduce statistical thinking to undergraduates. The course is motivated by the observation that most students are not familiar with the probabilistic conceptual framework that underpins much of geoscience. The course is intended for students pursuing majors in atmospheric science and statistics and follows a flipped classroom approach in which most of the class time is devoted to hands-on assignments addressed in small groups. The course uses real datasets to illustrate key aspects of probabilistic thinking without requiring high-level mathematics. Course materials are published online to facilitate their use by instructors worldwide. The second activity is a set of workshops that bring together users of weather and climate forecasts and undergraduate atmospheric science students at Cornell, carried out in collaboration with the Northeast Regional Climate Center (NRCC). Speakers are invited from regional organizations which represent a variety of sectors including renewable energy, emergency response, and water management. The speakers specifically address the use of weather forecasts and climate change projections in their decision making. Students are recruited from both Cornell and other colleges and universities in upstate New York, and research projects are developed in partnership with workshop speakers. The project has clear societal relevance given the hazards posed by severe weather. The work is of interest both for its specific results and for its development of event attribution and rare-event sampling methodologies that can be applied to subseasonal predictions of extreme events. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.