Volume 34, Issue 8
Atmospheric Science
Free Access

Sensitivity of U.S. surface ozone to future emissions and climate changes

Zhining Tao

Zhining Tao

Illinois State Water Survey, University of Illinois at Urbana-Champaign, Champaign, Illinois, USA

Search for more papers by this author
Allen Williams

Allen Williams

Illinois State Water Survey, University of Illinois at Urbana-Champaign, Champaign, Illinois, USA

Search for more papers by this author
Ho-Chun Huang

Ho-Chun Huang

Illinois State Water Survey, University of Illinois at Urbana-Champaign, Champaign, Illinois, USA

Now at Scientific Applications International Corporation, Camp Springs, Maryland, USA.

Search for more papers by this author
Michael Caughey

Michael Caughey

Illinois State Water Survey, University of Illinois at Urbana-Champaign, Champaign, Illinois, USA

Search for more papers by this author
Xin-Zhong Liang

Xin-Zhong Liang

Illinois State Water Survey, University of Illinois at Urbana-Champaign, Champaign, Illinois, USA

Search for more papers by this author
First published: 27 April 2007
Citations: 44

Abstract

[1] The relative contributions of projected future emissions and climate changes to U.S. surface ozone concentrations are investigated focusing on California, the Midwest, the Northeast, and Texas. By 2050 regional average ozone concentrations increase by 2–15% under the IPCC SRES A1Fi (“dirty”) scenario, and decrease by 4–12% under the B1 (relatively “clean”) scenario. However, the magnitudes of ozone changes differ significantly between major metropolitan and rural areas. These ozone changes are dominated by the emissions changes in 61% area of the contiguous U.S. under the B1 scenario, but are largely determined by the projected climate changes in 46% area under the A1Fi scenario. In the ozone responses to climate changes, the biogenic emissions changes contribute strongly over the Northeast, moderately in the Midwest, and negligibly in other regions.

1. Introduction

[2] Numerous general circulation modeling (GCM) studies have projected important global climate changes for future decades in response to emissions changes resulting from human activities [Intergovernmental Panel on Climate Change, 2001]. These emissions and induced climate changes likely will impact surface ozone (O3) concentrations. Changes in emissions of O3 precursors, e.g., oxides of nitrogen (NOx) and volatile organic compounds (VOC), directly alter regional O3 concentrations through complex photochemical reactions. Meanwhile, climate changes affect regional O3 in two ways: (1) changing wind circulation and planetary boundary layer structure directly alter surface O3 concentrations through dilution and advection, and (2) increasing solar radiation and temperature enhance biogenic emissions and photochemical reaction rates, especially conversion between peroxyacetyl nitrate and NOx [e.g., Hogrefe et al., 2004; Murazaki and Hess, 2006]. Understanding not only the overall impacts of emissions and climate changes on, but also their individual contributions to, regional O3 distributions is critical for designing and implementing effective control strategies for the U.S. to meet the National Ambient Air Quality Standards in the future.

[3] Limited modeling studies have evaluated the impacts of emissions and climate changes on tropospheric background O3 [e.g., Fiore et al., 2002; Prather et al., 2003], and on regional O3 over the Europe [Szopa et al., 2006] and over the U.S. [Hogrefe et al., 2004; Murazaki and Hess, 2006; Steiner et al., 2006; J.-T. Lin et al., submitted manuscript, 2007]. For the U.S., the existing studies have generally used simulations under a single emissions scenario and presented results that either are not appropriate for direct comparison, or are inconclusive. The uncertainties in projecting future emissions and climate remain high, and thus, simulations with multiple emissions scenarios are necessary to better estimate the possible range of future surface O3 changes. In this study, we apply a newly developed regional climate-air quality model to investigate the individual and combined impacts of future emissions and climate changes on U.S. surface O3 concentrations under two emissions scenarios from the Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios (SRES) [Intergovernmental Panel on Climate Change (IPCC), 2000]. We also distinguish the effects of climate changes between meteorological conditions (e.g., temperature, radiation, and wind) and their resulting biogenic emissions. As such, this study will provide the first insight into the range of relative contributions of both anthropogenic and biogenic emissions, and of climate changes to future U.S. surface O3 projection.

2. Model Simulations and Observations

[4] The surface O3 concentrations are modeled using an improved version of the SJVAQS-AUSPEX Regional Modeling Adaptation Project Air Quality Model (SAQM) [Huang et al., 2007]. The meteorological fields are downscaled by the MM5-based regional climate model (CMM5) [Liang et al., 2004a, 2004b], as driven by a fully coupled GCM, the National Center for Atmospheric Research/Department of Energy Parallel Climate Model [Washington et al., 2000]. Liang et al. [2006] has shown that the CMM5, downscaled at 30 km grid resolution, significantly reduce the driving PCM bias in simulating the present climate. Future climates for 2050 are simulated under two IPCC SRES emissions scenarios, A1Fi – a “dirty” outlook, and B1 – a relatively “clean” outlook [IPCC, 2000]. The invariant clean background lateral boundary conditions are assumed for both present and future simulations [Huang et al., 2007].

[5] The present anthropogenic emissions inventories for the SAQM simulation are obtained from the 1999 National Emissions Inventory (http://www.epa.gov/air/data) for the U.S, the Big Bend Regional Aerosol and Visibility Observational Study Emissions Inventory for the northern 10 states of Mexico [Kuhns et al., 2003], and the National Pollutant Release Inventory for Canada (http://www.ec.gc.ca/pdb/npri). These inventories are integrated and processed through the Sparse Matrix Operator Kernel Emissions modeling system (SMOKE) [Houyoux et al., 2000]. The biogenic emissions are estimated using the Biogenic Emissions Inventory System embedded in the SMOKE. Future anthropogenic emissions are estimated by multiplying the present emissions with the growth factors for 2050 according to the A1Fi and B1 scenarios. A growth factor based on the Organization for Economic Cooperation and Development as at 1990 is used for the U.S. and Canada, and another based on Africa, Latin America and Middle East is applied to Mexico. Future biogenic emissions are estimated from the current land-use data and the future meteorological fields. The computational domains are illustrated in Figure 1, where the full outer domain and four inner nested subdomains are at 90 km and 30 km resolution, respectively, for the SAQM simulations. The following analysis focuses on summer (June, July, and August) results for the four inner subdomains, centered on California, the Midwest U.S., the Northeast U.S., and Texas, respectively.

Details are in the caption following the image
Summer average daily maximum 8-hour surface ozone concentration changes (ppb) between 2050 and 1998: (a) A1Fi scenario and (b) B1 scenario. Relative contributions (%) of the projected emissions (EMS) and climate (MET) changes to total surface O3 concentration trends between 2050 and 1998: (c) A1Fi scenario and (d) B1 scenario. Color code 1 (green) denotes the dominance of the EMS effect (contribution > 70%), 2 (yellow) the MET effect (contribution > 70%), and 3 (red) both effects to be important. Regions with solid lines represent the four 30-km-resolution subdomains, and the remaining areas use 90-km-resolution results.

[6] Four sets of simulations were carried out to assess the relative contributions of changes in emissions and climate to regional O3 changes (Table 1). The contributions are denoted as EMS for emissions changes, MET for climate changes, and BIO for the biogenic emissions changes induced by climate change.

Table 1. Experimental Designa
Case Anthropogenic Emissions Biogenic Emissions Climate
baseline 1998 1998 1998
1 2050 1998 1998
2 2050 2050 1998
3 2050 2050 2050
  • a EMS = case 1 − baseline; MET = case 3 − case 1; BIO = case 2 − case 1. 1998 represents the present and 2050, the future.

3. Results

[7] The modeled 1998 summer surface O3 concentrations are compared to the observations that are archived in the U.S. Environmental Protection Agency (Air Quality System, http://www.epa.gov/ttn/airs/airsaqs) database. Nearly 4000 stations within the four fine subdomains are selected while any null values in the AQS database are eliminated. If there are multiple stations in a grid cell, the average concentration is calculated to represent the observation for comparison with the modeled grid cell value. These grid cell values are then averaged in each subdomain to provide the regional statistics of normalized bias and gross error [Tesche et al., 1990] (Table 2). The model underestimates O3 in California, the Midwest, and Texas subdomains, while overestimates in the Northeast subdomain. The normalized gross errors range from 26% to 35% in the four selected regions. A more detailed model evaluation is given by Huang et al. [2007].

Table 2. Comparison of Modeled Surface O3 Concentrations With Observations in Four Subdomainsa
Subdomain Observation Number Average O Average P NB, % NGE, %
California 109,937 63.2 54.0 −9.7 26.2
Midwest 179,247 57.2 50.9 −8.5 29.0
Northeast 164,916 60.9 60.2 3.0 32.8
Texas 100,062 58.0 44.3 −20.6 34.8
  • a NB = equation image; NGE = equation image; P and O are the modeled and observed 1-hour average surface O3 concentration, respectively; N is the number of observations. The cutoff value is 40 ppb for observations. Average O and P are domain average values.

[8] The CMM5, as driven by the PCM, simulates moderate climate changes by 2050 under both the A1Fi and B1 scenarios. Surface temperature changes are less than 2 K in most areas, which are comparable to the observed interannual temperature variability between 1996 and 2000. Emissions of NOx and VOC, two important O3 precursors, vary according to the specified scenario and the resultant climate changes (Table 3). Changes in anthropogenic emissions are determined by the across-the-board growth factors, whereas changes in biogenic emissions depend largely on local landuse and meteorology. Because biogenic contributions to VOC emissions are much larger than those to NOx (> 50% vs. < 10%), under both scenarios the geographic distribution of percentage changes in VOC is less uniform than that of NOx. The different magnitude changes between NOx and VOC have important implications to regional O3 production as discussed below.

Table 3. 2050 Summer Average Changes of Surface Temperature and Emissionsa
Scenario Temperature, °C NOx, %b Isoprene, % TVOC, %b
CA MW NE TX CA MW NE TX CA MW NE TX CA MW NE TX
A1Fi 0.9 0.6 0.6 0.2 37 29 32 28 12 5 10 −3 10 4 6 0
B1 −1.0 1.2 1.2 1.1 −48 −50 −56 −49 −8 15 15 13 −16 −6 −9 1
  • a Changes are relative to 1998. CA, California; MW, Midwest; NE, Northeast; TX, Texas.
  • b NOx and TVOC are the sum of both anthropogenic and biogenic emissions.

[9] Regional O3 levels respond to changes in both emissions and climate. Under the A1Fi scenario (Figure 1a), summer average daily maximum 8-hour O3 trends upward by 3 to 20 ppb in vast rural areas, but changes less than 2 ppb in the Ohio River Valley and major metropolitan areas, such as New York City, Chicago, Detroit, and Houston. Decreases of 3 to 9 ppb are calculated for the Washington, D. C. – Baltimore, Los Angeles, and San Francisco areas. These regional differences may largely be explained by the contrast between the NOx and VOC emissions changes. NOx emissions increase by approximately 30% in all subdomains, while VOC emissions may either increase or decrease. VOC emissions increase by 10% in the California subdomain, by 6% in the Northeast, and 4% in the Midwest, but decrease by 0.2% in Texas. When emissions changes reduce the VOC/NOx ratio, O3 formation is enhanced in the NOx-sensitive areas, but decreased in the VOC-sensitive areas. Broad rural regions in the U.S. have been shown to be NOx-sensitive, so increased NOx will boost O3 production [Roselle, 1994; Pierce et al., 1998; Tao et al., 2003]. The major cities and the Ohio River Valley that emit large amount of NOx are VOC-sensitive, where additional NOx titrate O3, either reducing its concentrations or causing negligible O3 change. The result is reversed under the B1 scenario, where the percentage NOx reductions (ca. 50% averaged over four subdomains) are much larger than of the percentage VOC reductions (<10%) in all four subdomains. This leads to O3 decreases of 5 to 15 ppb in rural areas, and increases of more than 5 ppb in major cities (Figure 1b). Note that the overall pattern of daily mean surface O3 changes, when averaged over the summer, resemble the daily maximum 8-hour O3 pattern under both the A1Fi and B1 scenarios.

[10] By averaging over the subdomains, the daily maximum 8-hour surface O3 concentrations increase by 11%, 9%, 4%, and 16% for California, the Midwest, the Northeast, and Texas, respectively, under the A1Fi scenario. The respective increases of daily mean O3 concentrations are 10%, 4%, 2%, and 16%. Under the B1 scenario, the daily maximum 8-hour surface O3 concentrations decrease by 12%, 8%, 14%, and 6% in California, the Midwest, the Northeast, and Texas, respectively. The corresponding daily mean O3 concentrations reduce by 9%, 4%, 12%, and 1%.

[11] Two additional sensitivity experiments (Table 1) are conducted to determine the relative contributions of emissions versus climate changes to surface O3 responses. Denoting EMS and MET for the summer average daily mean surface O3 responses to the projected emissions and climate changes, respectively, the relative contributions at each grid cell are calculated as:
urn:x-wiley:00948276:media:grl22977:grl22977-math-0003
where X is either EMS or MET. Given that EMS and MET can be either positive or negative, and the magnitudes of O3 changes vary significantly from grid cell to grid cell, the normalized percentage contributions provide better visual presentation and easier interpretation of the results.

[12] Figures 1c and 1d illustrate the spatial distribution of the relative EMS and MET contributions to the surface O3 changes under the A1Fi and B1 scenarios. Under the A1Fi scenario (Figure 1c), the MET effect dominates (>70%) the surface O3 changes in large rural areas, the MET and EMS contributions are comparable in metropolitan areas, and the EMS becomes a bigger factor in the Lower Michigan and upper Ohio regions. On the other hand, the EMS is the dominant factor over vast areas in California and the Northeast, and the MET and EMS are both significant for major cities and large portions of the Midwest and Texas under the B1 scenario (Figure 1d).

[13] Table 4 summarizes the subdomain average statistics of the relative EMS and MET contributions to summer average daily mean O3 changes. Clearly under the A1Fi scenario, the MET effect dominates over the EMS effect in California, the Midwest, and Texas, but relatively unimportant in the Northeast. In contrast, under the B1 scenario, EMS becomes the dominant factor in all four regions although the MET contribution is also important in the Midwest and Texas.

Table 4. Relative Contributions of the Projected Emissions and Climate Changes to Summer Average Daily Mean Surface O3 Concentration Changesa
Subdomain A1Fi − EMS, % A1Fi − MET, % B1 − EMS, % B1 − MET, %
California 29 71 −94 6
Midwest −4 96 −64 36
Northeast 87 13 −92 8
Texas 14 86 −53 47
  • a EMS, projected emissions; MET, climate. Negative sign indicates reduction in surface O3.

[14] A portion of the MET effect is attributed to the O3 responses to the biogenic emissions changes (BIO) that result from the temperature and solar radiation changes [Guenther et al., 1995]. Figure 2 compares the subdomain average BIO contributions to the total MET effect for the two scenarios. The BIO portion is around 10% in Texas under both the A1Fi and B1 scenarios. In California, the BIO effect under the B1 scenario is negative, suggesting that the MET effect would otherwise be larger if the biogenic emissions changes induced by the climate changes were excluded. The BIO effect becomes significant in the Midwest (45%) and more so in the Northeast (>50%) under the A1Fi scenario. This is consistent with the result of Hogrefe et al. [2004], where the BIO effect under the A2 scenario accounts for approximately 50% of the total changes in summer average maximum O3 in the Midwest and Northeast.

Details are in the caption following the image
Relative contribution of biogenic emissions changes (BIO) to the total effect of the projected climate changes (MET) on future surface O3 concentrations (2050 minus 1998) as averaged over the subdomains of California, the Midwest, the Northeast, and Texas.

4. Conclusion and Discussion

[15] The U.S. regional surface O3 responds differently to the emissions and climate changes projected for 2050 under the different IPCC emissions scenarios. Under the A1Fi scenario (“dirty” outlook), the summer average daily mean and maximum 8-hour O3 concentrations show upward trends in vast rural areas in the U.S. For metropolitan areas, however, they both show small changes or even decrease. Under the B1 scenario (relatively “clean” outlook), both O3 quantities decline in broad rural areas, reflecting the effectiveness of cutting NOx emissions to mitigate the O3 problem. However, in major cities and metropolitan areas, large reductions in NOx emissions tend to increase surface O3 concentrations due to the reduced titration effect. Our sensitivity study shows that the projected emissions changes are largely responsible for future U.S. O3 changes in more than 61% area of the contiguous U.S. under the B1 scenario, while the projected climate changes are the dominant factor in 46% area under the A1Fi scenario. The biogenic emissions changes induced by the projected climate changes are dominant in the Northeast under both scenarios, contribute largely (moderately) in the Midwest under the A1Fi (B1) scenario, and are a minor factor in California and Texas.

[16] Due to the nonlinear nature of surface O3 production, the sensitivity of O3 concentrations to projected emissions and climate changes is largely dependent on local characteristics of NOx and VOC emissions, and meteorology. There might be an “optimal” point where surface O3 concentrations are sensitive to both emissions and climate changes, which would cause the largest surface O3 response. This question can only be resolved when more emissions scenarios and climate projections are available, and thus the ranges of surface O3 changes reported in this study should be applied with caution.

[17] Although it is not one of our four focal regions, it is worth noting that, according to our 90-km results and averaged over the region, by 2050 the southeast U.S. is projected to have reduced surface O3 concentrations by approximately 11% under the B1 scenario, and to experience about 14% O3 increases under the A1Fi scenario. Large reductions in NOx emissions under the B1 scenario are responsible for decreased O3 concentrations in this biogenic VOC abundant region [Tao et al., 2003]. On the other hand, climate changes play a larger role in O3 increases under the A1Fi scenario (Figure 1c). In the future, this region merits more studies.

[18] The above results contain some caveats. In particular, the analysis is based on simulations of a single year for the present (1998) and future (2050), and thus is subject to sampling influence. H.-C. Huang et al. (manuscript in preparation, 2007) analyzed the regional O3 changes, simulated by the same modeling system and the same forcing data of emissions and climate between 5 years of the present (1996–2000) and future (2048–2050). When both the climate and emissions are allowed to change (case 3 versus baseline), the spatial distributions and magnitudes in changes of average daily mean and daily maximum 8-hour surface O3 concentrations from our 1-year simulations resemble those from the 5-years' average given by H.-C. Huang et al. (manuscript in preparation, 2007). Table 5 compares the average daily mean O3 concentration changes based on the 5-year versus 1-year samples. Under both the A1Fi and B1 scenarios, the projected O3 are very close in all four subdomains, indicating that the influence of small sampling is not anticipated to change our main conclusions.

Table 5. Comparison of Future Summer Average Daily Mean Surface O3 Concentration Changes
Subdomain A1Fi − 5 Year, % A1Fi − 1 Year, % B1 − 5 Year, % B1 − 1 Year, %
California 8 10 −9 −9
Midwest 4 4 −5 −4
Northeast 2 2 −11 −12
Texas 10 16 −3 −1

[19] The effect of long range transport on regional O3 problem is not included in this study. An ongoing investigation (H.-C. Huang et al., manuscript in preparation, 2007) uses a global chemistry transport model and SAQM to examine such impact under the A1Fi and B1 scenarios. The preliminary result shows that the long range transport causes a 4 – 7% O3 increase by 2050. This compares to the 2 – 15% O3 increases under the A1Fi scenario, and 4 – 12% decreases under the B1 scenario reported here. Thus, it is of interest and merits further investigation to fully understand this issue.

[20] Nevertheless, we emphasize the necessity to conduct more detailed and systematic investigations, preferably using an ensemble of emissions scenarios and climate models (GCM/RCM) that are representative of the likely range of future emissions and climate changes to obtain a more credible projection of future U.S. air quality changes and uncertainty.

Acknowledgments

[21] We thank two anonymous reviewers for their constructive comments on the manuscript. We acknowledge the Forecast Systems Laboratory/National Oceanic & Atmospheric Administration and the National Center for Supercomputing Applications/University of Illinois at Urbana-Champaign for the supercomputing support. The research was partially supported by the United States Environmental Protection Agency Science to Achieve Results (STAR) Awards RD-83096301-0 and R831449. The views expressed are those of the authors and do not necessarily reflect those of the sponsoring agencies or the Illinois State Water Survey.