Search NASA⌕ Search

Engineering topics

Stein, Michael L.

Publications and source records attributed to Stein, Michael L..

At least 19 records

Modeling Spatial Asymmetries in Teleconnected Extreme Temperatures

Abstract Combining strengths from deep learning and extreme value theory can help describe complex relationships between variables where extreme events have significant impacts (e.g., environmental or financial applications). Neural networks learn complicated nonlinear relationships from large datasets under limited parametric assumptions. By definition, the number of occurrences of extreme events is small, which limits the ability of the data-hungry, nonparametric neural network to describe rare events. Inspired by recent extreme cold winter weather events in North America caused by atmospheric blocking, we examine several probabilistic generative models for the entire multivariate probability distribution of daily boreal winter surface air temperature. We propose metrics to measure spatial asymmetries, such as long-range anticorrelated patterns that commonly appear in temperature fields during blocking events. Compared to vine copulas, the statistical standard for multivariate copula modeling, deep learning methods show improved ability to reproduce complicated asymmetries in the spatial distribution of ERA5 temperature reanalysis, including the spatial extent of in-sample extreme events.

Krock, Mitchell L.↗

Tail Dependence as a Measure of Teleconnected Warm and Cold Extremes of North American Wintertime Temperatures

Current models for spatial extremes are concerned with the joint upper (or lower) tail of the distribution at two or more locations. Such models cannot account for teleconnection patterns of 2-m surface air temperature ( T 2m ) in North America, where very low temperatures in the contiguous United States may coincide with very high temperatures in Alaska in the wintertime. This dependence between warm and cold extremes motivates the need for a model with opposite-tail dependence in spatial extremes. This work develops a statistical modeling framework that has flexible behavior in all four pairings of high and low extremes at pairs of locations. In particular, we use a mixture of rotations of common Archimedean copulas to capture various combinations of four-corner tail dependence. We study teleconnected T 2m extremes using ERA5 of daily average 2-m temperature during the boreal winter. Further, the estimated mixture model quantifies the strength of opposite-tail dependence between warm temperatures in Alaska and cold temperatures in the midlatitudes of North America, as well as the reverse pattern. These dependence patterns are shown to correspond to blocked and zonal patterns of midtropospheric flow. This analysis extends the classical notion of correlation-based teleconnections to considering dependence in higher quantiles.

54 ENVIRONMENTAL SCIENCES↗

Scalable computations for nonstationary Gaussian processes

Nonstationary Gaussian process models can capture complex spatially varying dependence structures in spatial datasets. However, the large number of observations in modern datasets makes fitting such models computationally intractable with conventional dense linear algebra. In addition, derivative-free or even first-order optimization methods can be very slow to converge when estimating many spatially varying parameters. In this paper, we present a computational framework which couples an algebraic block diagonal plus low-rank covariance matrix approximation with stochastic trace estimation to facilitate the efficient use of second-order solvers for maximum likelihood estimation of Gaussian process models with many parameters. We demonstrate the effectiveness of these methods by simultaneously fitting 192 parameters in the popular nonstationary model of Paciorek and Schervish using 107,600 sea surface temperature anomaly measurements.

97 MATHEMATICS AND COMPUTING↗

Neural networks for parameter estimation in intractable models

The goal is to use deep learning models to estimate parameters in statistical models when standard likelihood estimation methods are computationally infeasible. For instance, inference for max-stable processes is exceptionally challenging even with small datasets, but simulation is straightforward. Data from model simulations are used to train deep neural networks and learn statistical parameters from max-stable models. The proposed neural network-based method provides a competitive alternative to current approaches, as demonstrated by considerable accuracy and computational time improvements. Finally, it serves as a proof of concept for deep learning in statistical parameter estimation and can be extended to other estimation problems.

97 MATHEMATICS AND COMPUTING↗

Fitting Matérn smoothness parameters using automatic differentiation

The Mat$\acute{e}$rn covariance function is ubiquitous in the application of Gaussian processes to spatial statistics and beyond. Perhaps the most important reason for this is that the smoothness parameter $\nu$ gives complete control over the mean-square differentiability of the process, which has significant implications for the behavior of estimated quantities such as interpolants and forecasts. Unfortunately, derivatives of the Mat$\acute{e}$rn covariance function with respect to $\nu$ require derivatives of the modified second-kind Bessel function $K$ $\nu$ with respect to $\nu$. While closed form expressions of these derivatives do exist, they are prohibitively difficult and expensive to compute. For this reason, many software packages require fixing $\nu$ as opposed to estimating it, and all existing software packages that attempt to offer the functionality of estimating $\nu$ use finite difference estimates for $\partial$ $\nu$ $K$ $\nu$ . In this work, we introduce a new implementation of $K$$\nu$ that has been designed to provide derivatives via automatic differentiation (AD), and whose resulting derivatives are significantly faster and more accurate than those computed using finite differences. Here, we provide comprehensive testing for both speed and accuracy and show that our AD solution can be used to build accurate Hessian matrices for second-order maximum likelihood estimation in settings where Hessians built with finite difference approximations completely fail.

97 MATHEMATICS AND COMPUTING↗

Flexible nonstationary spatiotemporal modeling of high-frequency monitoring data

Many physical datasets are generated by collections of instruments that make measurements at regular time intervals. For such regular monitoring data, we extend the framework of half-spectral covariance functions to the case of nonstationarity in space and time and demonstrate that this method provides a natural and tractable way to incorporate complex behaviors into a covariance model. Further, we use this method with fully time-domain computations to obtain bona fide maximum likelihood estimators—as opposed to using Whittle-type likelihood approximations, for example—that can still be computed conveniently. Additionally, we apply this method to very high-frequency Doppler LIDAR vertical wind velocity measurements, demonstrating that the model can expressively capture the extreme nonstationarity of dynamics above and below the atmospheric boundary layer and, more importantly, the interaction of the process dynamics across it.

54 ENVIRONMENTAL SCIENCES↗

The Evolving Distribution of Relative Humidity Conditional Upon Daily Maximum Temperature in a Warming Climate

Abstract The impacts of heat waves in a warming climate depend not only on changing temperatures but also on changing humidity. Using 35 simulations from the Community Earth System Model Large Ensemble (CESM LENS), we investigate the long‐term evolution of the joint distribution of summer relative humidity (RH) and daily maximum temperature ( Tmax ) near four U.S. cities (New York City, Chicago, Phoenix, and New Orleans) under the high‐emissions Representative Concentration Pathway (RCP) 8.5. We estimate the conditional quantiles of RH given Tmax with quantile regression models, using functions of temperature for each city in July for three time periods (1990–2005, 2026–2035, and 2071–2080). Quality‐of‐fit diagnostics indicate that these models accurately estimate conditional quantiles for each city. As expected, each quantile of Tmax increases from 1990–2005 to 2071–2080, while mean RH decreases modestly. Conditional upon a fixed quantile of Tmax , the median and high quantiles of RH decrease, while those of the Heat Index (HI) and dew point both increase. This result suggests that, despite a modest decrease in median relative humidity, heat stress measured by metrics considering both humidity and temperature in a warming climate will increase faster than that measured by temperatures alone would indicate. For a fixed Tmax , the high quantiles of RH (and thus of HI and dew point) increase from 1990–2005 to 2071–2080 in all four cities. This result suggests that the heat stress of a day at a given Tmax will increase in a warming climate due to the increase of RH.

Yuan, Jiacan↗

Future climate emulations using quantile regressions on large ensembles

The study of climate change and its impacts depends on generating projections of future temperature and other climate variables. For detailed studies, these projections usually require some combination of numerical simulation and observations, given that simulations of even the current climate do not perfectly reproduce local conditions. We present a methodology for generating future climate projections that takes advantage of the emergence of climate model ensembles, whose large amounts of data allow for detailed modeling of the probability distribution of temperature or other climate variables. The procedure gives us estimated changes in model distributions that are then applied to observations to yield projections that preserve the spatiotemporal dependence in the observations. We use quantile regression to estimate a discrete set of quantiles of daily temperature as a function of seasonality and long-term change, with smooth spline functions of season, long-term trends, and their interactions used as basis functions for the quantile regression. A particular innovation is that more extreme quantiles are modeled as exceedances above less extreme quantiles in a nested fashion, so that the complexity of the model for exceedances decreases the further out into the tail of the distribution one goes. We apply this method to two large ensembles of model runs using the same forcing scenario, both based on versions of the Community Earth System Model (CESM), run at different resolutions. The approach generates observation-based future simulations with no processing or modeling of the observed climate needed other than a simple linear rescaling. The resulting quantile maps illuminate substantial differences between the climate model ensembles, including differences in warming in the Pacific Northwest that are particularly large in the lower quantiles during winter. We show how the availability of two ensembles allows the efficacy of the method to be tested with a “perfect model” approach, in which we estimate transformations using the lower-resolution ensemble and then apply the estimated transformations to single runs from the high-resolution ensemble. Finally, we describe and implement a simple method for adjusting a transformation estimated from a large ensemble of one climate model using only a single run of a second, but hopefully more realistic, climate model.

54 ENVIRONMENTAL SCIENCES↗

Estimating Changes in Temperature Distributions in a Large Ensemble of Climate Simulations Using Quantile Regression

Understanding future changes in extreme temperature events in a transient climate is inherently challenging. A single model simulation is generally insufficient to characterize the statistical properties of the evolving climate, but ensembles of repeated simulations with different initial conditions greatly expand the amount of data available. In this paper, we present a new approach for using ensembles to characterize changes in temperature distributions based on quantile regression that more flexibly characterizes seasonal changes. Specifically, our approach uses a continuous representation of seasonality rather than breaking the dataset into seasonal blocks; that is, we assume that temperature distributions evolve smoothly both day to day over an annual cycle and year to year over longer secular trends. To demonstrate our method’s utility, we analyze an ensemble of 50 simulations of the Community Earth System Model (CESM) under a scenario of increasing radiative forcing to 2100, focusing on North America. As previous studies have found, we see that daily temperature bulk variability generally decreases in wintertime in the continental mid- and high latitudes (>40°). A more subtle result that our approach uncovers is that differences in two low quantiles of wintertime temperatures do not shrink as much as the rest of the temperature distribution, producing a more negative skew in the overall distribution. Although the examples above concern temperature only, the technique is sufficiently general that it can be used to generate precise estimates of distributional changes in a broad range of climate variables by exploiting the power of ensembles.

54 ENVIRONMENTAL SCIENCES↗

Changes in Spatiotemporal Precipitation Patterns in Changing Climate Conditions

Climate models robustly imply that some significant change in precipitation patterns will occur. Models consistently project that the intensity of individual precipitation events increases by approximately 6-7%/K, following the increase in atmospheric water content, but that total precipitation increases by a lesser amount (2-3%/K in the global average). Some other aspect of precipitation events must then change to compensate for this difference. We develop here a new methodology for identifying individual rainstorms and studying their physical characteristics - including starting location, intensity, spatial extent, duration, and trajectory - that allows identifying that compensating mechanism. We apply this technique to precipitation over the contiguous U.S. from both radar-based data products and high-resolution model runs simulating 100 years of business-as-usual warming. In model studies, we find that the dominant compensating mechanism is a reduction of storm size. In summer, rainstorms become more intense but smaller; in winter, rainstorm shrinkage still dominates, but storms also become less numerous and shorter duration. These results imply that flood impacts from climate change will be less severe than would be expected from changes in precipitation intensity alone. We show also that projected changes are smaller than model-observation biases, implying that the best means of incorporating them into impact assessments is via "data-driven simulations" that apply model-projected changes to observational data. We therefore develop a simulation algorithm that statistically describes model changes in precipitation characteristics and adjusts data accordingly, and show that, especially for summertime precipitation, it outperforms simulation approaches that do not include spatial information.

54 ENVIRONMENTAL SCIENCES↗

Evaluation of dynamically downscaled extreme temperature using a spatially-aggregated generalized extreme value (GEV) model

The Weather Research and Forecast (WRF) model downscaling skill in extreme maximum daily temperature is evaluated by using the generalized extreme value (GEV) distribution. While the GEV distribution has been used extensively in climatology and meteorology for estimating probabilities of extreme events, accurately estimating GEV parameters based on data from a single pixel can be difficult, even with fairly long data records. This work proposes a simple method assuming that the shape parameter, the most difficult of the three parameters to estimate, does not vary over a relatively large region. This approach is applied to evaluate 31-year WRF-downscaled extreme maximum temperature through comparison with North American Regional Reanalysis (NARR) data. Uncertainty in GEV parameter estimates and the statistical significance in the differences of estimates between WRF and NARR are accounted for by conducting bootstrap resampling. Despite certain biases over parts of the United States, overall, WRF shows good agreement with NARR in the spatial pattern and magnitudes of GEV parameter estimates. Both WRF and NARR show a significant increase in extreme maximum temperature over the southern Great Plains and southeastern United States in January and over the western United States in July. The GEV model shows clear benefits from the regionally constant shape parameter assumption, for example, leading to estimates of the location and scale parameters of the model that show coherent spatial patterns.

Wang, Jiali↗