Final Report: Extreme Air Quality Events Using a Hierarchy of Models: Present and Future

EPA Grant Number: R835205
Title: Extreme Air Quality Events Using a Hierarchy of Models: Present and Future
Investigators: Hess, Peter , Berner, Judith , Grigoriu, Mircea Dan , Mahowald, Natalie M.
Institution: Cornell University , National Center for Atmospheric Research
EPA Project Officer: Chung, Serena
Project Period: June 1, 2012 through May 31, 2015 (Extended to May 31, 2016)
Project Amount: $746,825
RFA: Extreme Event Impacts on Air Quality and Water Quality with a Changing Global Climate (2011) RFA Text |  Recipients Lists
Research Category: Air Quality and Air Toxics , Global Climate Change , Water and Watersheds , Climate Change , Air , Water

Objective:

This grant funded interdisciplinary research to address the following broad questions: Under current conditions what is the probability of an extreme pollution event? During the next century, how are the probability, frequency, duration and severity of high pollution episodes likely to change under future emission and climate scenarios? What are the geographic, meteorological, climatological and chemical conditions that could contribute to extreme pollution episodes in the United States? What parts of the country are particularly sensitive to extreme pollution events now and in the future? How do extreme pollution events relate to heat waves? What are the feedbacks between heat waves and severe pollution events?

In order to address the questions raised above we proposed the following tasks: (1) Apply extreme value theory to measurements. (2) Evaluate the impact of adding stochastic parameterizations to the Community Earth System Model (CESM) on extreme values. (3) Compare extreme values in model predictions to observations. (4) Evaluate future extremes for a range of climate and emission scenarios. (5) Develop surrogate models for extreme pollution events.

Summary/Accomplishments (Outputs/Outcomes):

Tasks 1–4 have been completed as discussed in detail below. We have applied extreme value theory to the measurements and to simulations of the present and future climate. In doing so we focused on the response of ozone and on ozone extremes to meteorology: in particular to temperature, temperature extremes, persistent high temperatures, persistent stagnation events, the frequency of cyclone passages and overall meteorological flow. The hypothesis that heat waves will become more commonplace and persistent in the future has guided much of our research. We analyzed the response of ozone through the use of measurements and in chemistry-climate simulations of the present and future climate. We introduced the concept of an upper bound to characterize ozone distributions across the continental United States in the many locations where the distributions are not heavy tailed. In addition, we introduced a new statistic to measure the joint extrema of temperature and ozone. Excessive variability in the CESM is the dominant reason why temperature extremes are overestimated, for both warm and cold extremes. The bias in temperature variability is a major impediment for accurate air quality predictions over North America. In addition to simulating excessive variability over the Northern Hemispheric continents, the CESM also simulates excessive variability in the tropical Pacific and Indian Ocean. We investigated the possibility that stochastic parameterizations would correct for the excessive variability.

Details of Technical Aspects of the Project: Major Findings

Pakawat Phalitnonkiat, a graduate student in applied mathematics, under the primary guidance of Gennady Samorodnitsky, Mircea Grigoriu and Peter Hess, applied extreme value theory to Clean Air Status and Trends Network (CASTNET) ozone measurements. Results were published in Atmospheric Environment (Phalitnonkiat, et al., 2016) and presented at the American Geophysical Union (AGU) meeting in San Francisco (Phalitnokiat, et al., 2014). Using a novel approach, a versatile model, the Generalized Pareto Distribution, was fit to the extrema of the ozone concentration using a combination of the Maximum Likelihood Estimation (MLE) and a Hill estimator approach. We were careful to separate the tails from the middle of the distribution and about appropriately transforming the data prior to extreme value analysis. We contend that determining the tail parameter of ozone distributions is trickier than it may appear.

This is particularly true with limited datasets. We compare our approach to previous approaches using synthetic data. We find that under a variety of conditions the procedure using MLE is likely to underestimate the tails. For heavy-tailed distributions we recommend a procedure based on Hill’s methodology. We introduce the concept of an upper bound to characterize ozone distributions across the continental United States when the distribution is light tailed. A finite upper bound gives the maximum value that will be reached. In addition to the return time to describe extremes of ozone, the upper bound can also be used as a defining characteristic of ozone distributions with light tails. The upper bound answers the question, how bad can it get? Using CASTNET data we evaluate decadal-level return periods and, where appropriate, the ultimate upper limits of the ozone concentration over the continental United States. Based on our analysis procedure, we find the upper bound of ozone ranges between 60 and 120 ppb when it exists (it depends on the particular station). We find however, that at certain stations an upper bound does not exist and the distribution is heavy tailed. At these locations one can expect more extreme behavior, although of course physically there will always be a limit as to how high the ozone concentrations can be. We find that the NOX State Implementation Plan (SIP) is likely associated with increases in the estimated shape parameter of the ozone distribution in many locations, implying it rendered the ozone distribution more heavy tailed while at the same time decreased the mean ozone.

Brownsteiner, et al. (2015), (partially funded on this grant) first compared ozone extremes in CASTNET data. Brownsteiner, et al. (2015), compared various configurations of CESM against CASTNET ozone and temperature data to evaluate their ability to simulate future climate chemistry by evaluating their ability to simulate present-day chemistry. The version of the CESM using analyzed meteorological inputs best simulates temperature over the United States but does not outperform a configuration using simulated meteorology in many of the other performance metrics examined. A configuration with simulated meteorology and 56 vertical levels is markedly better in capturing observed ozone-temperature relationships and 100 ppb return intervals than a configuration that is identical except that it contains 26 vertical levels. We note that in the calculation of the 100 ppb return intervals the mean ozone bias is removed. The paper recommends caution in the use of GCCMs in simulating surface chemistry, as differences in a variety of model parameters have a significant impact on the resulting chemical and climate variables. Isoprene emissions depend strongly on surface temperature, and the resulting ozone chemistry is dependent on isoprene emissions but also on cloud cover, photolysis, the number of vertical levels and the choice of meteorology. These dependencies must be accounted for in the interpretation of GCM results.

In a second study, Phalitnonkiat, et al. (2017), conducted an analysis using extreme value theory to examine the joint extremes of temperature and pollution in the present and future climate (circa 2100) using the CESM with comparisons to CASTNET data. This follows on the work of Brownsteiner, et al. (2015), who first compared ozone extremes in CASTNET data and various configurations of the CESM (see above). Phalitnonkiat, et al. (2017), is still in preparation (with partial funding from a National Science Foundation [NSF] grant) but should be ready to submit shortly. While the mean behavior of ozone with temperature has been investigated in several studies, we are interested in understanding the impact of temperature on the tail of the ozone distribution. (i.e., How do ozone extremes change with temperature extremes?) The research presents an overall geographical record of the effect of heat waves on ozone. We measure how ozone and temperature extremes change in the future compared to mean changes using the quantity ψ.

Here GCM2100 is a simulation of the future climate in the CESM, GCM2000 is a simulation of the present day climate, the mean is taken over 25 years in these simulations, and X and Y are either temperature or ozone. Taking X and Y as temperature, we find that over the western United States future changes in temperature extremes (temperatures above 90th percentile) decrease with respect to changes in the mean temperature, but they increase over much of the eastern half of the United States by as much as 30 percent. Future changes in ozone extremes with respect to the mean decrease over much of the United States. Changes in the sensitivity of future ozone extremes to future temperature extremes varies strongly by region (where X is ozone and Y is temperature), but with increases over much of the eastern half of the United States by over 30 percent.

We introduce a new statistic, φ, to quantify the extent to which temperature extremes and ozone extremes occur at the same time. While there is a well-known increase of ozone with temperature, analysis of the model simulations and CASTNET data show the joint extremes of ozone and temperature as quantitifed by φ occur simultaneously at most 20–30 percent of the time. We find future changes in φ are, on average, small. Thus while high ozone days and high temperature days are obviously related, the extremes in both are only modestly related.

Sun, et al. (2017a), investigates the impact of meteorological persistence on ozone extremes using CASTNET data. This work was also presented at the San Francisco AGU (Sun, et al., 2015). CASTNET ozone and temperature data and large-scale meteorological analysis are used to quantify the extent to which meteorological events and their persistence impact ozone with an emphasis on the high end of the ozone distribution (greater than the 90th percentile). Ozone increases with each successive stagnation day in all regions of the United States, with the highest increase in the Northeastern United States (0.4 standard deviation or ∼4.7 ppb per successive stagnation day). Ozone increases with days since cyclone passage only in the northeastern and Mid-Atlantic regions of the United States but on average not enough to reach the 90th percentile concentration. Persistent high temperature does not result in further ozone increases in any region. On the interannual timescale there is little evidence that summers with large numbers of the above events increase ozone preferentially on the high end of the ozone distribution. Examining the interannual variability in the current climate, we do not find evidence to support an amplication of ozone extremes in a future climate. In particular, we do not find a tail enhancement in relation to changes in synoptic frequency.

In the first study of its kind, Sun, et al. (2016), and Sun, et al. (2017b), examine the impact of waviness of the 500 hPa flow on ozone concentration (quantified through the metric of "wave activity") in CASTNET data and in present day and future simulations (circa 2100) using the CESM. The final paper detailing this work (Sun, et al., 2017b) has been partially funding from the NSF and is currently in preparation. Wave activity can be used as a metric for measuring the general "waviness" of the meteorological flow. Wave activity is a dynamically meaningful quantity related to midlatitude circulation and its changes, with a proven relationship to meteorological extremes. Changes in wave activity should encapsulate changes in stagnant conditions, as well as changes in blocking and the passage of synoptic fronts, but in a succinct manner. Thus wave activity provides a metric to relate changes in the flow to pollution events. Interannual changes in wave activity are well correlated with interannual changes in measured and simulated ozone in the current climate over the United States. As the circulation changes in a future climate, we find that wave activity also changes. Using correlations between wave activity and ozone in the current climate we estimate that future changes in wave activity act to increase future ozone by 1–3 ppb over large regions of the United States, particularly the eastern United States.

The CESM exhibits too much variability over the Northern Hemispheric continents but also in the tropical Pacific and Indian Ocean. This excessive variability is the dominant reason why temperature extremes are overestimated, for both warm and cold extremes. The bias in temperature variability is a major impediment for accurate air quality predictions over North America. To overcome this bias we tested two stochastic parameterizations as a means to add the effect of small-scale noise on general circulation model simulations. While neither of the stochastic parameterization schemes tested were able to remedy this shortcoming over North America, one of the stochastic parameterizations schemes evaluated, namely the stochastically perturbed tendency scheme (SPPT), did remedy the excessive temperature variability in the tropics. In this region, SPPT reduces the variability of sea surface temperatures in the El Nino 3.4 region by 60 percent, leading to much better agreement between the simulated and observed spectra. These findings were summarized in an article submitted to the Journal of Climate and are currently under review (Christensen, et al., 2017).

In the final period of the grant, the mechanisms leading to the improved temperature variability in the tropical Pacific were investigated to see how a stochastic scheme would need to be designed to improve temperature variability and extremes over North America. It was found that SPPT increased convection over the maritime continent and decreased convection over the subsidence region in the Eastern Pacific. This leads to a stronger Walker circulation and less intermittence and thus variability in the tropical belt. The decreased atmospheric variability results in a weaker ocean forcing and makes the simulated El Nino–Southern Oscillation variability more similar to the observed one. These results were presented in an invited talk at the European Centre for Medium-Range Weather Forecasts (ECMWF)/World Weather Research Programme workshop on model error, ECMWF, Reading, UK, April 11–15, 2016.

The findings suggest that the excessive variability over North America in model simulations might be the result of a too-strong coupling between the atmospheric and land model components. If this coupling is weakened, for example, but not necessarily by stochastic methods, the bias in temperature variability might be reduced, which is guaranteed to lead to a better estimation of temperature extremes. This is a novel hypothesis, which to our knowledge has not been considered as a possible mechanism for an improved simulation of near-surface temperatures in climate model simulations.

Conclusions:

Tropospheric ozone has severe impacts on health and is consequently regulated by the U.S. Environmental Protection Agency. High-pollution episodes have been attributed to high temperatures, stagnant conditions and a reduction in the frequency of cyclone passages. Climate change may increase the persistence of these types of conditions, acting to exacerbate future pollution episodes. This concern has been heightened by the severe pollution events associated with the European heat wave of 2003 and the Russian heat wave of 2010, events which resulted in many premature fatalities due to both high temperatures and air pollution. A number of studies suggest that climate change will particularly increase ozone on the high end of its distribution, making the bad pollution days even worse. We find the following notable results vis-à-vis ozone extremes: (1) In many locations over the United States, the ozone distribution is not characterized by extreme concentrations but can be mathematically characterized as bounded where the upper bound of the ozone distribution (the highest concentration of ozone that can exist at a particular location) can be obtained from an analysis of the distribution itself. (2) The change in temperature and ozone extremes circa 2100 compared to changes in the mean show considerable regional variation across the United States, where in some regions the extremes become more extreme but in other regions are less extreme. (3) While there is a well-known increase of ozone with temperature, analysis of the model simulations and CASTNET data show the joint extremes of ozone and temperature occur together at most 20–30 percent of the time. (4) An examination of the interannual variability of ozone in the present climate suggests that ozone extreme concentrations will not be enhanced in a future climate. (5) Successive stagnation days cause ozone to build up in all regions of the country, but successive high-temperature days do not. (6) Changes in circulation in a future climate act to increase ozone in many regions of the country. (7) Excessive coupling between the atmospheric and land component of the CESM may result in the excessive temperature variability resulting in an overestimation of temperature extremes, a major impediment for accurate future air quality predictions over North America.


Journal Articles on this Report : 10 Displayed | Download in RIS Format

Other project views: All 22 publications 10 publications in selected types All 10 journal articles
Type Citation Project Document Sources
Journal Article Berner J, Fossell KR, Ha S-Y, Hacker JP, Snyder C. Increasing the skill of probabilistic forecasts: understanding performance improvements from model-error representations. Monthly Weather Review 2015;143(4):1295-1320. R835205 (2013)
R835205 (Final)
  • Full-text: AMS-Full Text PDF
    Exit
  • Abstract: AMS-Abstract
    Exit
  • Journal Article Brown-Steiner B, Hess PG, Lin MY. On the capabilities and limitations of GCCM simulations of summertime regional air quality:a diagnostic analysis of ozone and temperature simulations in the US using CESM CAM-Chem. Atmospheric Environment 2015;101:134-148. R835205 (2013)
    R835205 (2014)
    R835205 (Final)
    R834283 (Final)
  • Full-text: ScienceDirect-Full Text HTML
    Exit
  • Abstract: ScienceDirect-Abstract
    Exit
  • Other: ScienceDirect-Full Text PDF
    Exit
  • Journal Article Christensen HM, Berner J, Coleman DRB, Palmer TN. Stochastic parameterization and El Niño–Southern Oscillation. Journal of Climate 2017;30(1):17-38. R835205 (Final)
  • Full-text: Journal of Climate-Full Text HTML
    Exit
  • Abstract: Journal of Climate-Abstract
    Exit
  • Other: Journal of Climate-Full Text PDF
    Exit
  • Journal Article Clark SK, Ward DS, Mahowald NM. The sensitivity of global climate to the episodicity of fire aerosol emissions. Atmospheric Chemistry and Physics Discussions 2013;13:23691-23717. R835205 (2013)
    R835205 (Final)
  • Full-text: EGU Journals-Full Text PDF
    Exit
  • Abstract: EGU Journals-Abstract HTML
    Exit
  • Journal Article Franzke CLE, O'Kane TJ, Berner J, Williams PD, Lucarini V. Stochastic climate theory and modeling. WIREs Climate Change 2015;6(1):63-78. R835205 (2013)
    R835205 (Final)
  • Full-text: Wiley-Full Text PDF
    Exit
  • Abstract: Wiley-Abstract & Full Text HTML
    Exit
  • Journal Article Phalitnonkiat P, Sun W, Grigoriu MD, Hess P, Samorodnitsky G. Extreme ozone events:tail behavior of the surface ozone distribution over the U.S. Atmospheric Environment 2016;128:134-146. R835205 (2014)
    R835205 (Final)
  • Full-text: ScienceDirect-Full Text HTML
    Exit
  • Abstract: ScienceDirect-Abstract
    Exit
  • Other: ScienceDirect-Full Text PDF
    Exit
  • Journal Article Riddick S, Ward D, Hess P, Mahowald N, Massad R, Holland E. Estimate of changes in agricultural terrestrial nitrogen pathways and ammonia emissions from 1850 to present in the Community Earth System Model. Biogeosciences 2016;13(11):3397-3426. R835205 (2013)
    R835205 (Final)
  • Full-text: Biogeosciences-Full Text PDF
    Exit
  • Abstract: Biogeosciences-Abstract
    Exit
  • Journal Article Romine GS, Schwartz CS, Berner J, Fossell KR, Snyder C, Anderson JL, Weisman ML. Representing forecast error in a convection-permitting ensemble system. Monthly Weather Review 2014;142(12):4519-4541. R835205 (2013)
    R835205 (Final)
  • Full-text: AMS Journals-Full Text PDF
    Exit
  • Abstract: AMS Journals-Abstract
    Exit
  • Journal Article Sun W, Hess P, Liu C. The impact of meteorological persistence on the distribution and extremes of ozone. Geophysical Research Letters 2017;44(3):1545-1553. R835205 (Final)
  • Abstract: Wiley-Abstract
    Exit
  • Journal Article Tagle F, Berner J, Grigoriu MD, Mahowald NM, Samorodnitsky G. Temperature extremes in the Community Atmosphere Model with stochastic parameterizations. Journal of Climate 2016;29(1):241-258. R835205 (Final)
  • Abstract: Journal of Climate-Abstract
    Exit
  • Other: UCAR-Abstract
    Exit
  • Progress and Final Reports:

    Original Abstract
    2012 Progress Report
    2013 Progress Report
    2014 Progress Report