A preliminary synthesis of modeled climate change impacts on U.S. regional ozone concentrations

and by increasing (but modifying in an quality model simulation of a California 0 3 daily peak 0 3 the of the 0 3 -forming NO 0 3 , via a decrease in PAN production. Recent EPA STAR-funded results yielded similar insights for the EPA global change-air quality assessment. in a high-resolution simulation of a 5-day 0 3 episode over California, found that temperature pertur-bations consistent with plausible 2050s climate change led to increases in afternoon 0 3 concentrations of 1-5 ppb across the state. et al. using a different modeling system, found similar effects of temperature modification when simulating 0 3 concentrations during a weeklong period over the eastern United States.

Dzyxwvut srqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA iscussion of the potential sensitivity of air quality to climate change has increased in recent years. In 2001, the NRC (acronyms defined in Table 1) posed the question "to what extent will the United States be in control of its own air quality in the coming decades?" noting that"... changing climatic conditions could significantly affect the air quality in some regions of the United States ..." and called for the expansion of air quality studies to include investigation of how U.S. air quality is affected by long-term climatic changes (NRC 2001). A subsequent NRC report emphasized that the U.S. air quality management system must be "flexible and vigilant" to ensure the effectiveness of pollution mitigation strategies in the face of climate change (NRC 2004). The recent IPCC Fourth Assessment Report warned of the possibility of significant air quality degradation in some regions as a result of climate-related changes in the dispersion rate of pollutants, the chemical environment for 03 and aerosol generation, and the strength of emissions from the biosphere, fires, and dust (Solomon et al. 2007).
The mission of the EPA is to protect human health and the environment. To achieve this mission, the where the emissions of VOCs and NOx are also large and that 03concentraitons increase even more when meteorological conditions most strongly favor net photochemical production-persistent high pressure, stagnant air, lack of convection, clear skies, and warm temperatures (e.g., U.S. EPA 1989;NRC 1991;Cox and Chu 1993;Bloomfield et al. 1996;Morris et al. 1995;Sillman and Samson 1995;U.S. EPA 1999;Thompson et al. 2001;Camalier et al. 2007; among many others).
Consequently, the 03 NAAQS are most often exceeded during summertime hot spells in places with large natural or anthropogenic precursor emissions (e.g., cities and suburban areas). Table 2 highlights a number of key meteorology-related factors.
Because climate change may alter weather patterns and hence potentially increase the frequency, duration, and intensity of 03 episodes in some regions, it has the potential to create additional challenges for air quality managers. However, the causal chain linking (i) long-term global climate change, (ii) short-term meteorological variability that most directly drives peak 03 episodes, and (iii) 03 changes that ultimately result from the interaction of these meteorological changes with the pollutants present in the environment (which may themselves be sensitive to meteorology) is not straightforward.
Changes in the 03 distribution of a given region as a result of climate change will reflect a balance among competing or reinforcing changes in multiple factors.
The meteorological variables that affect O do not, in general, vary independently of each other, nor must they vary in concert with corresponding effects on 03 concentrations. The 1991 NRC report noted that the relationship between temperature and 03 "cannot readily be extrapolated to a warmer climate because higher temperatures are often correlated empirically with sunlight and meteorology" (NRC 1991). How the relationship between 03 and its meteorological drivers is perceived depends on the timescale considered (see the sidebar on p. 5 for additional information about the temperature-03 relationship). TABLEzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 2. Meteorological variables with the potential to affect regional air quality (adapted from U.S. EPA 1989).
The average maximum or minimum temperature and/or changes in their spatial distribution and duration, leading to a change in reaction rate coefficients and the solubility of gases in cloud water solution; The frequency and pattern of cloud cover, leading to a change in reaction rates and rates of conversion of S02to acid deposition; The frequency and intensity of stagnation episodes or a change in the mixing layer, leading to more or less mixing of polluted air with background air; Background boundary layer concentrations of water vapor, hydrocarbons, NOx, and 03, leading to more or less dilution of polluted air in the boundary layer and altering the chemical transformation rates; The vegetative and soil emission of hydrocarbons and NOxthat are sensitive to temperature and light levels, leading to changes in their concentrations; Deposition rates to vegetative surfaces whose absorption of pollutants is a function of moisture, temperature, light intensity, and other factors, leading to changes in concentrations; and Circulations and precipitation patterns, leading to a change in the abundance of pollutants deposited locally versus those exported off the continent, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONM (NRC 2001), a major component of the assessment approach is the development and application of global to regional climate and air quality modeling systems. Having multiple groups-with differences in emphasis and using a range (albeit still limited) of models, chemical and physical parameterizations, and greenhouse gas scenarios-address the same problem enhances the richness of the EPA assessment effort; the collective results may reveal choices to which the results are particularly sensitive, thereby building insight into the workings of the coupled system. Table 3 provides a summary of the global and regional modeling experiments available to date from this first phase of the assessment, highlighting the different combinations of modeling tools and other aspects of simulation design. Collectively, these simulations (described in more detail in the papers listed in  will likely depend at least as much on how many of these meteorological episodes that promote 03 formation occur, and how long they last, as on how hot it is during each one. In other words, how often in a given summer that cool, cloudy, rainy, and windy conditions give way to spells of hot, clear, dry, and stagnant conditions will play a large role in determining whether it was a "high 03" or "low 03" summer. At this time scale, temperature and 03 will also be positively correlated; however, here the "temperature-03" relationship exists at least partly because temperature itself is highly correlated with these other meteorological conditions-such as more sunlight and less ventilation--that also favor increased 03 concentrations.  Table 3 refer only to greenhouse gas concentrations and not to precursor pollutants. As emphasized earlier, all of the results shown here are from simulations that held anthropogenic emissions of precursor pollutants constant at present-day levels but allowed climate-sensitive natural emissions of biogenic VOCs to vary in response to the simulated climate changes. Figure 1 shows summertime mean MDA8 03 concentration differences between future and present-day climates. This air quality metric is selected because of its direct relevance to U.S. air quality standards.

LONG-TERM CLIMATE
Several cited. This is significant because it is the high-03 episodes that most concern air quality managers in the United States.
There are also significant differences, however,  Certain regions show greater agreement across experiments than others, at least in a very general sense.
For example, Fig. 1 shows that a loosely bounded area, This m ay have som e consequences for direct com parison, which will be discussed further later in the paper. zyxwvutsrqponmlkjihgfedcb  One way to summarize what Figs. 3-5, in conjunction with Fig. 1, are showing us is that 03 largely responds to the meteorological drivers in a qualitatively consistent manner across the different climate change experiments, but the regional patterns of relative changes in these drivers are highly variable across these sets of simulations. In other words, there are important differences in the simulated future regional climate changes that seem to drive the differences in  Table 3).   Table 3 shows that all the regional model experiments whose results are shown in Fig. 1 have chemical mechanisms that do recycle isoprene nitrate. Figure 6 shows the averaging subregions used in also shows the averages for the two global modeling experiments discussed below.) The regional modeling findings presented here are generally consistent with the relatively few regional  Table 3 supports the most general conclusions from the regional modeling studies; that is, large regions of the country show future 03 concentration increases of a few to several parts per billion, and there are significant differences in the spatial patterns of these changes between the simulations. In a global context, the results from these simulations are generally consistent with other GCTM climate change experiments (e.g., see Murazaki and Hess 2006;Stevenson et al. 2006;Zeng et al. 2008  In Fig. 9, which shows the same quantities for the CMU experiment, a different regional pattern of change emerges. In the CMU experiment, the major increase in future 03 concentration is instead centered on the Gulf Coast and eastern seaboard, with minimal 03 changes in the upper Midwest and northern plains states. The differences between these two sets of results can seemingly mostly be explained by two factors:  Table 3).
pressure anomalies in the present-day and future CMU simulations and found only relatively small changes in these regions. These results suggest that storm-track activity does not decrease as much in this CMU model simulation [see also Leibensperger et al. (2008) for further discussion]. In any case, it seems plausible that differences in simulated future largescale circulation patterns explain the differences in future 03 changes simulated in the two experiments for the northern part of the country.
The even larger differences in simulated future 03 changes in the southern half of the country likely arise because of differences in how isoprene chemistry is described in the Harvard-1 and CMU modeling systems, leading to differences in how 03 responds to the climate-induced changes in biogenic VOC emissions.
The spatial patterns of future-minus-present changes in isoprene emissions shown in Figs. 8d and 9d are qualitatively similar, with the largest increases centered on the Southeast and Gulf Coast regions for both groups.
Examining the CMU results in Fig. 9, it appears that increases in temperature and decreases in cloud cover (and hence increases in insolation) have combined to lead to increases in both isoprene emissions and 03 concentrations in this region. An additional CMU simulation with future meteorology but scaled-back isoprene emissions has confirmed that the enhanced 03 chemical production resulting from these enhanced emissions are largely responsible for the simulated future 03 increases (Racherla and Adams 2008).
This is in contrast to the Harvard-1 results in One factor to which this striking difference between the two sets of results might be traced is the modeled isoprene nitrate chemistry, as mentioned earlier.
Although increased emissions of biogenic VOCs are often associated with increases in 03 concentrations, these increased emissions can also lead to decreases in 03 concentrations via different pathways. For example, high concentrations of isoprene can reduce O amounts through direct ozonolysis, and they can also suppress 03 production in NOx-limited regimes (e.g., rural areas) zyxwvutsrqponm FIGzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA . 9. As AMERICAN zyxwvutsrqponmlkjihgfedcbaYXWVUTSRPONMLKJIHGFEDCBA in Fig. 8 but f or t he CM U global modeling experiment (see Table 3).
increasing isoprene emissions seem to result in little change, or even decreases in 03 amounts, perhaps because the model chemistry represents these isoprene nitrates as a "terminal" sink for NO . In the absence of   Lastly, although this paper discusses the problem of climate change effects alone on air quality, it is of course unrealistic to assume that emissions will stay the same into the future in the face of future economic and technological development and future regulatory regimes. As described earlier in the paper, understanding the interactions and combined effects of both climate and emissions changes is the focus of the second phase of the EPA assessment effort, and a number of the modeling groups mentioned here have made some initial efforts in this direction (e.g., see Hogrefe et al. 2004b;Nolte et al. 2008;Racherla and Adams 2008;Steiner et al. 2006;Tagaris et al. 2007;Tao et al. 2007;Wu et al. 2008a,b;Zhang et al. 2008). An initial model intercomparison study of the first-order relative effects of climate and emissions changes on U.S. regional 03 concentrations has been conducted and is being prepared for a separate publication.
For the scientific research community, assessments such as the one being carried out by the EPA help convey the key knowledge gaps that limit our understanding of the problem and/or create barriers to the use and interpretation of scientific information by decision makers. In this case, coupled global climate-regional air quality science is still in a relatively youthful state. Because air quality-from a health, environmental, and regulatory perspective-is largely determined by episodes that occur during specific, sporadic weather events, the ability of available modeling tools to simulate these events and capture the variability and future changes in these episodes is important. The focus of the climate modeling community has been shifting in recent years from long-term mean values of variables such as temperature and precipitation to increased consideration of changes in variability, extremes, and the frequency of specific weather patterns. Some of this effort should be directed into more detailed considerations of the climate metrics and statistics most relevant for air quality and more evaluations of climate models for these metrics and statistics. New research carried out under the auspices of this assessment, as summarized in Leung and Gustafson (2005) and Gustafson and Leung (2007), represent advances in this direction and provide useful insights. Additionally, this assessment has helped improve the understanding of the sensitivity of simulated meteorology, and hence air quality endpoints, to model physical parameterizations (e.g., Liang et al. 2004aLiang et al. ,b 2006Lynn et al. 2007;Kunkel et al. 2007;Tao et al. 2008