Potential regional climate change and implications to U.S. air quality

Authors


Abstract

[1] Regional climate change scenarios were generated by dynamical downscaling to assess the potential effects of climate change on U.S. air quality. Comparing the climate simulation for 2045–2055 based on the IPCC A1B scenario with the control simulation of 1995–2005, large atmospheric changes that could affect air quality were found in several regions. Analyses were based on changes in surface air temperature and downward solar radiation, precipitation frequency, stagnation events, and ventilation. Changes in the Midwest and Texas during summer are of opposite sign, suggesting negative impacts on air quality in Texas and small positive or no impact in the Midwest. During fall, large warming, increased solar radiation, reduced rainfall frequency, increased stagnation occurrence, and reduced ventilation in the western U.S. all suggest negative impacts on regional air quality. These changes are related to an anticyclonic pattern in the 500 hPa height change that is also found in other GCM projections.

1. Introduction

[2] In the last decade, climate projections have been used to assess the societal impacts of greenhouse warming in sectors such as water resources, agriculture, and ecosystems. More recently, several studies have begun to investigate how regional air quality may be affected by emission and climate changes in the future. Prather et al. [2003] summarized the projections of tropospheric ozone concentrations based on 10 global models. They found large increases in ozone concentrations over certain tropical hot spots where high sunlight and larger shifts in emissions prevail. Their study, however, only accounts for changes in ozone due to anthropogenic emission changes. Obviously, changes in climate as a result of greenhouse warming, and changes in biogenic emissions in response to climate change can alter ozone concentrations in a complex manner. In addition, global climate and chemistry models [e.g., Mickley et al., 2004; Langner et al., 2005] do not have enough spatial resolution to resolve atmospheric, chemical, and surface processes for assessing regional air quality.

[3] Hogrefe et al. [2004] reported the first study that applied both global and regional climate and chemistry models to assess the impacts of climate change and anthropogenic emission on regional air quality over the eastern U.S. This study is part of an on-going effort to investigate the impacts of global climate change on U.S. air quality. We have applied dynamical downscaling of global climate simulations to generate regional scale climate change scenarios. Because interannual climate variations may obscure long term trends, we examine climate projections for two 10 year periods of the 1990s and 2050s. Regional model outputs will be used to drive an air quality model and estimate changes in biogenic emissions as a result of vegetation changes. Our ultimate goal is to assess climate change effects on air quality based on consistent scenarios of emission, climate, and vegetation through one-way coupling of global climate, regional climate, vegetation, and air quality models.

[4] This paper analyzes the atmospheric changes in the regional climate change simulations to determine where (what regions) and when (what seasons) we may expect significant changes in regional air quality. Although air quality changes are ultimately governed by complex interactions between emissions, climate, and chemistry, analysis of atmospheric changes is needed to guide the experimental design and analysis and interpretation of air quality simulations.

[5] Of particular concern in regional air quality issues is the surface ozone concentration that has known effects on human health, crop, and the natural ecosystem. Surface ozone (O3) is produced by complex reactions between hydrocarbons and NOx emitted from a variety of natural and anthropogenic sources. Besides variations in emission rate, type, and sources and the mix of precursor species, O3 concentration is affected by solar radiation, air temperature, and mixing/transport. While solar radiation moderates O3 production primarily by influencing NOx photolysis rate, air temperature affects ozone by changing biogenic emissions and conversion between NOx and PAN [e.g., Jacob et al., 1993]. Mixing/transport can dilute/advect ozone and precursor concentrations and affect O3 formation and concentrations. Other pollutants such as sulfate, nitrate, and particulate matter can be removed by wet deposition so changes in precipitation frequency can also have large impacts on air quality.

[6] Synoptic conditions are often used as a surrogate of temperature, solar radiation (or cloud cover), and mixing/transport to explain variability in ozone concentrations. For example, high ozone episodes in the East and Midwest are often related to stagnating surface anticyclones [e.g., Comrie and Yarnal, 1992]. In the Pacific Northwest, the development of a thermal trough and upper-level ridge are climatologically significant factors that control summertime O3 concentrations [e.g., McKendry, 1994]. However the relationships between mixing or precipitation and synoptic condition are not robust because factors such as mesoscale circulation and orographic forcing can have large effects. With a modeling approach, analysis can be performed based on atmospheric variables that directly affect ozone concentrations, as well as synoptic conditions simulated by the model. By examining a suite of atmospheric variables, we aim to identify particular regions and seasons within the U.S. where consistency emerges regarding potential changes in air quality.

2. Numerical Experiments

[7] To simulate current and future climate at the regional scale for air quality assessment, a regional climate model based on the Penn State/NCAR Mesoscale Model (MM5) [Grell et al., 1994] was used to downscale a global climate simulation generated by the Goddard Institute of Space Studies (GISS) model [Rind et al., 1999]. The GISS global climate simulation was described by Mickley et al. [2004], who applied the model at a horizontal resolution of 4° latitude by 5° longitude. The transient climate simulation covers the period 1950 to 2055. Observed greenhouse gas concentrations were used during 1950–2000, and the simulation followed the Intergovernmental Panel of Climate Change A1B scenario [Houghton et al., 2001] for greenhouse gases during 2000–2055.

[8] A two-way nested configuration was used in the regional climate model at 108 km and 36 km horizontal resolution, respectively, for the outer and inner domains, with 23 vertical levels. The inner domain covers the entire continental U.S., part of Canada and Mexico, and the surrounding oceans. Leung et al. [2003] described the physics parameterizations used. Two regional climate simulations were performed based on large scale conditions from the GISS simulation for 1995–2005 and 2045–2055. The model was initialized on June 1 of 1995 and 2045. Lateral boundary conditions were applied every 6 hours within 15 grid points in the buffer zone of the outer domain. Model outputs were archived every hour to provide meteorological conditions for driving air quality models.

3. Results

[9] To estimate the change in atmospheric conditions, we calculated the difference between the future (2045–2055) and control (1995–2005) climate for the inner domain. The downscaled control simulation captures the seasonal cycle and regional variations in surface temperature and precipitation quite well when compared with observations. Averaged seasonally, surface temperature bias is generally about 1–2°C, with larger warm biases of up to 4°C in parts of the Southeast during summer. Precipitation is realistically simulated over the western U.S., but larger dry biases of 50–80% are found over parts of the eastern and southeastern U.S. during summer. Biases in the control simulation are comparable to those reported by Leung et al. [2004]. More detailed evaluation of the downscaled control simulation will be reported in a separate paper.

[10] The analyses presented below focus on changes in atmospheric conditions related to air quality. As discussed above, these include air temperature, solar radiation, and rainfall that affect biogenic emission, precursor concentrations, chemical mechanism, and wet deposition. Mixing/transport depends on wind, atmospheric stability, and mixing depth. We use stagnation and ventilation as collective measures of mixing and transport that affect air quality. The frequency of stagnation events further provide information about the persistence of weather conditions conducive to high ozone episodes found over many parts of the U.S. The following analyses focus mainly on regions and seasons with atmospheric changes that collectively suggest positive or negative impacts on air quality.

[11] Figures 1a and 1b show the change in the 2-meter air temperature simulated by the regional model for the summer and fall seasons. There are large regional and seasonal differences in how surface temperature responds to greenhouse forcing. During winter and spring (not shown), warming is between 0 and 3°C in the U.S., but reaches 4°C at the higher latitudes over northeastern Canada. During summer, large warming between 2–4°C is found over the southwestern U.S. and Mexico. However, the warming is much smaller over the Midwestern U.S., with some regions even showing small cooling related to changes in cloud cover in the future climate. During fall, warming of up to 4°C occurs over much of the western U.S. The maximum daytime temperature has larger impacts on ozone concentrations. In the simulations, the maximum daytime temperature is usually warmed by 1–2°C more than the increase in the minimum nighttime temperature to reach a warming of 5°C in southwest Texas during summer and part of California, Nevada, and South Dakota during fall.

Figure 1.

Changes in 2-meter air temperature (°C), for (a) summer and (b) fall, and rainfall frequency (in days per season), for (c) summer and (d) fall, based on the difference between the mean future (2045–2055) and control (1995–2005) regional climate simulations.

[12] Changes in downward solar radiation at the surface follow quite closely the spatial distribution of temperature change (not shown). During winter and spring, the change is mostly within 8 W/m2 in the U.S. During summer, it reaches 30 W/m2 over Texas. In the Midwest, cloud cover changes reduce solar radiation by up to 30 W/m2, which is consistent with the temperature change. During fall, the change is positive everywhere, with a maximum over the western U.S. of 15 W/m2.

[13] Rainfall frequency can affect wet deposition. Based on the regional simulations, daily rainfall intensity increases in most regions across the continental U.S., but the change in daily rainfall frequency is more spatially variable. Similar to changes in temperature and solar radiation, changes in rainfall frequency (within 4 days per season) are small during winter and spring. During summer, the contrast between the Midwest and Texas, as found in the temperature and solar radiation changes, shows up clearly in the rainfall frequency (Figures 1c and 1d), with an increase of up to 5 days per season in the Midwest, and a decrease of up to 6–8 days per season in Texas. During fall, consistent with the larger warming and reduced solar radiation, rainfall frequency decreases by about 5 days per season over a large part of the western U.S.

[14] Stagnation events strongly correlate with poor air quality. Similar to Korshover and Angell [1982], we defined a time to be stagnant if the following three criteria are met consecutively for a 4 day period: (a) the 10 m wind speed is less than 4 m/s at 7:00 am LST, (b) the 500 mb wind speed is less than 13 m/s at 7:00 am LST, and (c) rainfall total is less than 0.001 cm during the 4 day period. The control simulation only slightly overestimated the number of stagnation days over the eastern U.S. compared with previous analysis based on observations [e.g., Korshover and Angell, 1982].

[15] Figures 2a and 2b show the change in the number of stagnation days during summer and fall. The changes during winter and spring are again much smaller and spatially incoherent. During summer, the number of stagnation days increases by less than 4 days per season over southern Texas. This change, though positive, is not large compared to the average 15 days per season in the control simulation. Larger changes in stagnation occurrence are noted over the intermountain zone of the West and in parts of the South Central U.S. However, we cannot find changes in solar radiation and rain frequency in these regions during summer that suggest similar positive or negative impacts on air quality as implied by the change in stagnation occurrence. During fall, there is an increase of up to 8 stagnation days per season in the western U.S. compared to the typical 5–10 stagnation days in the control simulation for the same region. Hence the percentage change in stagnation occurrence is very significant.

Figure 2.

Similar to Figure 1, but for the change in the number of stagnation days (in days per season) (a and b), and the change in the unvented hours (in hours per day) (c and d).

[16] Mixing and transport within the boundary layer are important mechanisms for diluting or removing air pollutants. The ventilation coefficient can be used to collectively represent these processes, and it is defined as the product of the mean wind speed within the boundary layer, and the boundary layer depth. A deeper boundary layer dilutes pollutants, and strong winds remove pollutants locally. We calculated the number of unvented hours from the control and future simulations by counting the total time when the ventilation coefficient is less than 6000 m2/s [e.g., Pielke et al., 1991]. An increase in the number of unvented hours is an indication of the negative impacts of climate change on air quality. In the control run, the mean number of unvented hours during summer and fall typically differs from the estimate based on the North American Regional Reanalysis (http://wwwt.emc.ncep.noaa.gov/mmb/rreanl/) by less than 3 hours, compared to the average of 10 to 15 hours per day.

[17] During summer, ventilation is reduced in the upper Midwest as a result of reduced boundary layer depth consistent with larger cloud cover, reduced solar radiation, and small warming/cooling (Figure 2c). On average, the diurnal maximum boundary layer depth decreases by 300–500 m over most areas during summer. In Texas, ventilation is enhanced as indicated by the decrease in the number of unvented hours of up to 2 hours per day. Although the boundary layer depth and wind speed increase by no more than 100 m and 0.5 m/s during daytime, the larger increase in wind speed of about 1 m/s during nighttime causes ventilation to last about 2 hours longer into the early morning hours. In parts of the South Central and Southeast U.S., increased ventilation and reduced stagnation may have positive impacts on air quality; however, increased solar radiation and rain frequency suggest otherwise.

[18] During fall, larger changes are located over the western U.S. (Figure 2d) where the diurnal maximum boundary layer depth increases by up to 150 m, compared to small changes of less than 80 m in all other regions, as a result of the large warming. However, consistent with the increased frequency of stagnation and higher pressure, wind speed is significantly reduced, which leads to an increase of 0.5–2 unvented hours per day. It is interesting to note that the spatial patterns of changes in stagnation and ventilation are similar, but not identical. Stagnation reflects the large scale circulation and its persistence, which affects ventilation, but the latter is not dictated by large scale circulation alone. Examining the changes in both is important for assessing potential climate change impacts on air quality.

4. Discussion

[19] Summarizing the above analyses, we found consistent changes in a suite of atmospheric variables related to regional air quality over three regions during summer and fall. Changes in other regions and seasons are either relatively small and/or inconsistent among the variables in terms of air quality impacts. During summer, Texas is marked by warming (1–3°C), increased downward solar radiation (up to 40 W/m2), less frequent rainfall (more than 8 days less per season), and slightly more frequent stagnation (up to 4 days more per season) that all suggest an increase in ozone concentrations. One exception, however, is the extension of ventilation time by about 2 hours during the early morning hours.

[20] The opposite conditions are found in the Midwest with very small warming or even cooling, reduced downward solar radiation (up to 30 W/m2), more frequent rainfall (up to 6 days more per season), less frequent stagnation (up to 8 days less per season), and reduced ventilation (up to 3 more unvented hours per day). Depending on the relative impacts of these parameters, ozone concentrations may remain similar or slightly decrease, based on the simulated atmospheric changes alone.

[21] During fall, all indicators consistently suggest increased ozone concentrations in the western U.S. These include warmer temperature (2–4°C), more downward solar radiation (up to 15 W/m2), less frequent rainfall (up to 8 days less per season), more frequent stagnation (up to 15 days more per season), and reduced ventilation (up to 2 more unvented hours per day). These conditions are consistent with the change in 500 hPa height that shows a distinct anticyclonic pattern centering over the western U.S. and the adjacent ocean. Meehl and Tebaldi [2004] studied the projected change in heat waves and found an increased frequency of heat waves over the western U.S. in the 21st century. This signal was related to the anticyclonic pattern in the 500 hPa height change located over the western U.S. in their model simulations. Meehl and Tebaldi found a comparable pattern in an ensemble of seven additional global climate projections. Consistent with these large scale changes projected by other GCMs, our simulations in the western U.S. suggest negative impacts on regional air quality.

[22] Comparing the regional simulations with the global simulations that were used to provide boundary conditions for downscaling, large differences were found in the climate change signals in the Midwest during summer. While the regional simulations show increased cloud cover and reduced stagnation, Mickley et al. [2004] found a decline in mid-latitude cyclone frequency across southern Canada that reduces ventilation in the Midwestern and eastern U.S. Their results show little change elsewhere and during other seasons.

[23] The large disparity between the GISS (∼400 km) and MM5 (36 km) horizontal resolution may explain some of the differences. Indeed, the MM5 storm tracks east of the Rockies were displaced compared to GISS because MM5 has a more realistic topographic representation of the Rocky Mountain. The regional model's ability to resolve the complex terrain and processes such as clouds and turbulence that are affected by the orography may also be responsible for differences between GISS and MM5 over the western U.S. Differences in physics parameterizations, such as convection, used in the models may further amplify differences in atmospheric conditions used in air quality assessment. This issue, together with model biases in the control simulation and uncertainty arising from downscaling different global climate projections, has important implications to the use of climate scenarios for assessing air quality and will be examined in more detail in follow on papers. Lastly, although atmospheric conditions exert important controls on air quality, the complex interactions between emissions, atmospheric changes, and chemistry must be considered for a more comprehensive assessment of the influence of climate change on regional air quality.

Acknowledgments

[24] The U.S. Environmental Protection Agency (EPA) through its Office of Research and Development funded this research under IAG DW-89-93963401. The views expressed in this paper are those of the authors and do not necessarily reflect the views or policies of the EPA. We thank Loretta Mickley for providing the global climate simulations used in this study. Pacific Northwest National Laboratory is operated for the U.S. Department of Energy by Battelle Memorial Institute under Contract DE-AC06-76RLO 1830.

Ancillary