Climate and Weather Extremes

The largest negative impacts of climate change typically manifest in response to extreme events. However, the inherently rare nature of extremes raises a number of challenges in their study, including a potentially large role of natural variability and the quantification of the probability of never-before-seen events. Here, we first explore the extent to which observed changes in extreme summertime temperatures in the Northern Hemisphere can be explained by a “shift” in the distribution without changes in shape using quantile regression to allow for non-normality. The vast majority of the observed behavior is explained by the “shift” mode, and climate model ensembles suggest that many of the shape changes are due to sampling of internal variability. We then test this understanding for the recent record-smashing 2021 Pacific Northwest heatwave. Similar to the hemispheric-wide results, there was not evidence in advance of 2021 of significant greater warming of the upper tail compared to the mean, and a comparison to a range of analogs in a climate model large ensemble indicates that the 2021 Pacific Northwest heatwave can occur without the tails shifting more than the mean, although its probability is astonishingly small. Importantly, we find many cases in the climate model where traditional extreme value analysis based on fitting Generalized Extreme Value distributions would fail to predict events as large as simulated. Finally, we expand to compound heat/humidity extremes and present new methodologies to estimate changes in humidity conditional on temperature. In contrast to expectation from climate models, the results show widespread decreases in humidity on hot days across the globe’s semiarid regions. In the American Southwest, the decreases are likely explained by a lack of moisture available from the land surface, which deprives the atmosphere of moisture during the warm season.

An impact of the rise in global temperatures caused by climate change has been an increased amount of melt observed on the world’s ice sheets. The Greenland ice sheet has had an accelerating contribution to sea level rise as a result, including large spatial extreme temperature events that cause large areas of the ice sheet to melt simultaneously. However, standard extreme value analysis methods are unsuitable for modelling surface temperatures due to a soft upper limit in temperatures around 0◦C and a large mass of observations just below this limit. The limit – a result of the melting point of ice and the complexities of the melt process – results in a poorly defined tail that greatly varies in shape across the ice sheet. Using MODIS satellite data, we use a Gaussian mixture model fit to the full range of observations to improve the fit in the upper tail. This gives a more contextual definition of an extreme event in light of the data, allowing estimation of melt as a probability rather than using a fixed melt threshold. The mixture model is then used as a marginal model for a Gaussian process and later a Spatial Conditional extremes model to understand the dependence of surface temperatures and the nature of spatial melt events. We also use regression techniques to examine the relationship between the mixture model parameters and spatial covariates, allowing us to extend the marginal structure to new locations and informing our use of spatial dependence functions.

Temperature at any given location evolves with the horizontal transport of warmer or cooler air, the vertical transport of warmer or cooler air, adiabatic compression or expansion, radiation, and effects related to moisture. In the first part of this talk, I will describe a framework that enables an examination of how these contribute to temperatures at different percentiles. We use ERA-5 data to average the different processes conditioned on each temperature percentile and evaluate their relative contributions to making temperatures more or less extreme. Over much of the midlatitude land, horizontal transport of temperature is critical in driving temperatures towards extreme values. Other regions have other dependencies. Because of the dependence of midlatitude temperature extremes on horizontal transport, the idea that quasi-resonant amplification (QRA) of Rossby waves is important for temperature extremes appears plausible. In the second part of the talk, I will describe the proposed QRA mechanism and how it has been applied in the literature thus far. I will then show how we have explored this mechanism in an idealized climate model and found the theory to not actually hold even in this simple case. I will conclude with some thoughts about process-based understanding of extreme events. 

This paper studies big data streams with regional-temporal extreme event (REE) structures and solar flare detection. An autoregressive conditional Fr’echet model with time-varying parameters for regional and its adjacent regionals extremes (ACRAE) is proposed. The ACRAE model can quickly detect rare REEs (i.e., solar flares) in big data streams and predict solar activity. The ACRAE model, with some mild regularity conditions, is proved to be stationary and ergodic. The parameter estimators are derived through the conditional maximum likelihood method. The consistency and asymptotic normality of the estimators are established. Simulations are used to demonstrate the efficiency of the proposed parameter estimators. In real solar flare detection, with the new dynamic extreme value modeling, the occurrence and climax of solar activity can be detected earlier than existing algorithms. The empirical study shows that the ACRAE model outperforms the existing detection algorithms with sampling strategies. (Joint work with Steven Moen and Jili Wang).

The weather is said to have “a mind of its own”, as a chaotic system with limited predictability. Yet its fluctuations remain largely confined to a familiar probability distribution, to which the built and natural environment is well adapted. More impactful are the extreme, intermittent fluctuations of temperature, moisture, and wind, which can seriously threaten ill-equipped human and ecological communities. Extreme events have “dynamics of their own” which are difficult to quantify because they play out in a far tail of the probability distribution, and so sample sizes are limited in both observations and simulations.
Nonetheless, non-equilibrium statistical mechanics provide some general laws for the progression of rare events. In particular, transition path theory (TPT) formulates the ensemble of rare events as a flow of probability current through a high-dimensional phase space, climbing a landscape called the committor function which quantifies progress toward the event and distinguishes the dominant physical mechanisms at each step. In this talk I will present a TPT analysis of a particular atmospheric phenomenon, sudden stratospheric warming (SSW): a premature collapse of the stratospheric polar vortex whose impacts propagate down to the surface and manifest in midlatitude cold air outbreaks. In an idealized model setting, we use enhanced sampling and coarse-graining to piece together the probability current from a patchwork of short simulations. Equipped with physical features informed by wave-mean flow interaction theory, our data-driven procedure makes precise the role of stochastic forcing in driving the enstrophy transfer from mean flow to waves that destroys the polar vortex, and the altitude dependence of each process. Our computation and visualization methods, we believe, can enhance our understanding of many kinds of extreme events in the climate system.

Energy balance is the basis for studying climate and the same principles represent the foundation of understanding human heat stress, which is similarly governed by radiative and sensible and latent heat fluxes, as well as by metabolic heat generation.  Consequently, the basis exists for using between basic principles in atmospheric thermodynamics and dynamics to understand the prevalence and intensity of human heat stress in a changing climate.  Temperature is one key variable in determining human heat stress, but as is obvious from the general framing as an energy balance problem, atmospheric humidity, wind speed, and radiation loading are also important. I will provide some background on the historical development of these concepts and then bring them up to date and show how parts of the problem of understanding and predicting future moist heat waves and heat stress looks very much like the understanding the dynamics and predicting the future of monsoons.  Some parts do not. Reality is complex and we must embrace that too.  I will show how methodological choices have deeply biased much of the heat stress prediction literature and suggest some easy to implement best practices. I will then present some applications of future moist heat stress prediction to impacts on human health, labor capacity, and the associated global economic impacts with an emphasis on poverty and developing nations.