- Email: [email protected]
Ozone in the southeastern United States: An observation-based model using measurements from the SEARCH network
Atmospheric Environment 88 (2014) 192e200
Contents lists available at ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
Ozone in the southeastern United States: An observation-based model using measurements from the SEARCH network C.L. Blanchard a, *, G.M. Hidy b, S. Tanenbaum a a b
Envair, 526 Cornell Avenue, Albany, CA 94706, USA Envair/Aerochem, 6 Evergreen Dr, Placitas, NM 87043, USA
h i g h l i g h t s Observed ozone response to weather and ozone precursors is modeled. Ozone responses to changing concentrations of ozone precursors are projected. Empirically-determined results may be compared to photochemical model predictions. Model is useful for relating ozone trends to precursor trends and weather.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 28 August 2013 Received in revised form 28 January 2014 Accepted 6 February 2014 Available online 7 February 2014
A generalized additive model (GAM) is used to examine the influence of meteorological factors, nitrogen oxides (NOx ¼ NO þ NO2), and non-methane hydrocarbons (NMOC) on daily peak 8-h ozone (O3) concentrations. Application to 2002e2011 monitoring data from the Southeastern Aerosol Research and Characterization (SEARCH) program showed sensitivity of peak 8-h O3 to morning concentrations of nitric oxide (NO) and nitrogen dioxide (NO2) and to afternoon concentrations of NO2 reaction products (NOz). Peak O3 decreased with increasing NO and increased with increasing NO2 concentrations, consistent with reactions involving O3, NO, and NO2. Ozone production efficiency (OPE), estimated from the modeled relation between peak 8-h O3 and afternoon NOz, was w40e100 percent higher at rural compared to urban sites. OPE was nonlinear at all sites, decreasing with increasing NOz concentration. The mean ratio of NOz/NOy showed a two-fold increase from urban to rural sites, associated with chemical aging in stagnant air masses from one day (urban sites) to two or more days (non-urban sites). Peak 8-h O3 concentrations in Atlanta were sensitive to concentrations of both non-biogenic NMOC and NOz. Non-urban Yorkville, Georgia, peak 8-h O3 concentrations were sensitive to NOz but not to nonbiogenic NMOC concentrations. The results are consistent with expected NMOC and NOx sensitivity in urban and non-urban locales. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Ozone SEARCH Meteorology NOx VOC Generalized additive model
1. Introduction Ozone (O3) formation is a nonlinear function of sunlight acting on ambient volatile organic compounds (VOC) and oxides of nitrogen (NOx, NO þ NO2). Tropospheric O3 mixing ratios (concentrations) are affected by solar intensity, the rate of O3 formation, the rate of dispersion of O3 and its precursors, meteorological factors, and transport of urban plumes. Since 1970, emission control programs have reduced emissions and ambient concentrations of NOx and VOC, resulting in declining
* Corresponding author. E-mail address: [email protected] (C.L. Blanchard). http://dx.doi.org/10.1016/j.atmosenv.2014.02.006 1352-2310/Ó 2014 Elsevier Ltd. All rights reserved.
O3 concentrations in all major metropolitan areas of the U.S. However, slow rates of O3 reduction have led to concern about the reliability of predictive tools and the fundamental understanding of O3 formation (e.g., National Research Council, 1991; Hidy et al., 2011). Photochemical Air Quality Simulation Models (PAQSMs) have been used for over forty years to estimate the influence of precursor emissions on O3. Complementary, observation-based modeling developed by the 1990s (e.g, NARSTO, 2000; Hidy, 2000) and is accepted by the U.S. Environmental Protection Agency (EPA) and state agencies for use in “weight-of-evidence” plans to achieve the National Ambient Air Quality Standard (NAAQS) and state standards for O3 (EPA, 2007).
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Observational and PAQSM studies suggest that NOx is the limiting O3 precursor throughout much of the non-urban eastern United States (Liu et al., 1987; Sillman et al., 1990; Trainer et al., 1993; Chameides and Cowling, 1995). In the southeastern U.S., biogenic emissions of isoprene and anthropogenic NOx emissions affect O3 (e.g., Chameides et al., 1988). Evidence indicates responsiveness of post-1990s O3 formation to changes in NOx emissions from electric generating units (EGUs) in the eastern U.S. (Gego et al., 2007, 2008; Gilliland et al., 2008; Godowitch et al., 2008). Beginning in 1995, EPA’s acid rain program (CAAA, Title IV) implemented a nationwide reduction of SO2 and NOx emissions from EGUs (EPA, 2011a). The 2004 NOx state implementation plan (SIP) Call required further reductions in EGU NOx emissions in twenty-one eastern states (EPA, 2012). The Clean Air Interstate Transport Rule required 70 percent reductions of EGU emissions of SO2 and NOx in 28 eastern states (EPA, 2011b). The monitoring record from the Southeastern Aerosol Research and Characterization (SEARCH) network offers a rich data set for investigating the response of O3 to emission changes. Between 1999 and 2010, the annual O3-season (MarcheOctober) 95th percentiles of the peak daily 8-h O3 concentrations declined at SEARCH sites by 1.1 0.4 to 2.4 0.6 ppbv per year (Blanchard et al., 2013a). O3 declines were not attributable to a specific precursor using simple trend analyses (Blanchard et al., 2013a). Previous work suggests that the highest peak 8-h O3 concentrations in Atlanta, Georgia are sensitive to both ambient NMOC and NOy (NOx plus other oxidized nitrogen species, including, e.g., HNO3), but meteorological factors have a larger effect on day-to-day O3 variations (Blanchard et al., 2010a). In this paper, we use a generalized additive statistical model (GAM) to relate daily peak 8-h O3 concentrations to weather, to ambient concentrations of O3 precursors (NO and NO2), and to NOx reaction products (NOz) at SEARCH sites from 2002 to 2011. The approach is applicable to any well-instrumented monitoring location. 2. Methods 2.1. Ambient air quality measurements Hourly measurements of gases (CO, SO2, NO, NO2, NOy, HNO3, O3, and NH3) were obtained from SEARCH public archives (Atmospheric Research and Analysis, 2013). Network operations and measurement methods are documented in Hansen et al. (2003, 2006), Edgerton et al. (2005, 2006; 2007; 2009), and Saylor et al. (2010). Eight sites are located in the southeastern U.S.: Pensacola, Florida (PNS) and Gulfport, Mississippi (GFP), urban coastal sites (w5 km and 1.5 km from the shoreline, respectively); Pensacola e outlying (aircraft) landing field (OLF) and Oak Grove, Mississippi (OAK), non-urban coastal sites near the Gulf (w20 km and 80 km inland, respectively); Atlanta, GeorgiaeJefferson Street (JST) and North Birmingham, Alabama (BHM), urban inland sites; and Yorkville, Georgia (YRK) and Centreville, Alabama (CTR), non-urban inland sites. All sites measure meteorological parameters and gases at 10 m height (Hansen et al., 2003; Edgerton et al., 2007; Saylor et al., 2010). Measurements of NMOC, a VOC subset, were made with 24-h canister samples collected every day at JST through 2008 (Blanchard et al., 2010b). Additional 24-h and hourly NMOC measurements were made at Photochemical Assessment Monitoring Stations (PAMS) located at YRK and other sites in the Atlanta area (Blanchard et al., 2010b). Spatial and temporal variations of SEARCH air pollutant concentrations from 1999 to 2010 are described in Blanchard et al. (2013a, 2013b). We used data from 2002 through 2011, because measurements of NO2 began in 2002. Daily averages were created from hourly
193
data. In addition, daily peak 1-h and peak 8-h NO, NO2, and O3 concentrations were determined for each site and day. Morning (sample start hours of 7 through 10 a.m., i.e., sampling interval of 7e11 a.m.) and afternoon (sample start hours of 12 noon through 3 p.m., i.e., sampling interval of 12e4 p.m.) concentrations of NO, NO2, NOy, and NOz (NOz ¼ NOy NO NO2) were also computed. Table 1 lists mean NO, NO2, NOy, and NOz concentrations.
2.2. Statistical model EPA developed a GAM that predicts the highest daily peak 8h O3 within a metropolitan area as a function of surface and upper-air meteorological variables (Camalier et al., 2007). The EPA model (EPA, 2009) runs under the open-source R software system (The R Project, 2009). Blanchard et al. (2010a) describe an adaptation of the EPA statistical model. Further adaptation is implemented here. The GAM is expressed as:
lðO3 Þi ¼ m þ f1 ðx1 Þi þ . þ fm ðxm Þi þ g1 ðy1 Þi þ . þ gn ðyn Þi þ h1 ðz1 Þ þ . þ hp zp þ ei
(1)
Following Camalier et al. (2007), l(O3)i is the logarithm of the peak 8-h O3 on day “i,” but either a different O3 metric or a function other than the logarithm could be used. We apply the model site by site. For convenience, the terms f1(x1)i through fm(xm)i represent the effects of meteorological variables on peak 8-h O3, and g1(y1)i through gn(yn)i represent the effects of ambient concentrations of O3 precursors on O3. For consistency with l(O3)i, the variables y1 through yn are logarithms of air quality measurements. The terms h1(z1) through hp(zp) represent the effects of temporal variables, including “day of week” and “year”. The last term, ei, is the difference between observed and predicted O3 (error). Each term expresses the effect of one parameter on daily peak 8-h O3 as a deviation from the long-term mean, “m” (logarithm of peak 8-h O3 averaged overall days). “Day of week” and “year” are categorical variables and are used to represent weekly cycles and trend, respectively. Trend need not be linear or monotonic. The GAM uses natural splines (Hastie and Tibshirani, 1990) to model nonlinear dependence of O3 on other predictor variables. Meteorological and air quality data were used to predict daily peak 8-h O3 at each SEARCH site during March through October of 2002 through 2011. Camalier et al. (2007) found that the most consistently significant predictors of peak 8-h O3 concentrations in the 39 eastern U.S. metropolitan areas studied were: (1) daily maximum surface temperature (T), (2) mid-day (10 a.m.e4 p.m.) relative humidity (RH), (3) morning (7 a.m.e10 a.m.) average wind speed, (4) afternoon (1 p.m.e4 p.m.) average wind speed, (5) morning (w1200 UTC) difference between 925 mb T and surface T, (6) deviation of morning (w1200 UTC) 850 mb T from 10-year monthly average, (7) air mass transport direction and distance (determined from back trajectories), (8) occurrence of rain (as number of hours), (9) julian day, (10) day of week, and (11) year. We replaced the 925 mb e surface T difference with the 850 mb T e daily minimum surface temperature because the 925 mb height in coastal areas is not always above the marine layer and may not represent temperatures aloft. We eliminated precipitation amount (or hours) and 850 mb T because they were not statistically significant at any site. We added sea-level pressure gradients from Birmingham to Mobile, AL, and from Birmingham to Atlanta. We refit the GAM using the preceding meteorological parameters, but added the daily maximum 1-h NO and the morning NO2 concentrations. The maximum NO concentrations tended to occur in the morning (hours 5 through 9 on 46e73 percent of the days), coinciding with morning traffic, while maximum NO2
194
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Table 1 Mean 2002e2011 NO, NO2, NOy, and NOz concentrations (ppbv). Site
NO Max 1 h
BHM CTR GFP JST OAK OLF PNS YRK
36.81 0.95 5.51 48.36 0.46 4.30 11.25 1.22
NO2 7e11 a.m. 1.13 0.03 0.25 1.5 0.01 0.16 0.51 0.04
12.43 2.06 5.35 17.39 1.37 3.72 5.64 3.50
0.16 0.03 0.08 0.15 0.02 0.05 0.09 0.06
NO 12e4 p.m. 1.29 0.11 0.46 2.01 0.07 0.22 0.55 0.31
concentrations occurred during both morning (hours 5 through 9 on 14e34 percent of the days) and evening (hours 17 through 23 on 28e58 percent of the days). Since evening NO2 maxima follow the time of the peak 8-h O3, the model used morning (7e10 a.m.) NO2 concentrations. We then added afternoon (12 noon to 4 p.m.) concentrations of NOz (NOz ¼ NOy NO NO2). Conceptually, afternoon NOz concentration indicates the amount of photochemical processing (Trainer et al., 1993). Ambient NOz concentrations depend on cumulative mass loading of NO and NO2 and on the reactivity of species in air masses (affecting the fraction of NOx that is converted to reaction products). Afternoon NOz concentrations were weakly correlated or uncorrelated with 1-h NO maxima and morning NO2 concentrations (r2 ¼ 0.0e0.2). Measurements of 24-h average isoprene and non-biogenic NMOC concentrations were incorporated into additional versions of the GAM for JST and YRK. Non-biogenic NMOC was computed as the sum of measured species concentrations less the concentrations of isoprene, a-pinene, b-pinene, d-limonene, and D3-carene (only isoprene was measured at YRK). The computation assumes that the measured C2eC12 alkanes, alkenes, and aromatics are predominantly anthropogenic with the exception of the listed biogenic species. The frequency of 24-h NMOC sampling was daily at JST but once every six days at YRK. Morning (sample starting hours 7e10 a.m.) non-biogenic NMOC and isoprene concentrations were computed and substituted for the 24-h values. Hourly NMOC measurements from the PAMS site at South DeKalb were used for JST. The PAMS hourly measurements were made daily during June through August, except at YRK the sampling frequency was every third day before 2004. During the Southern Oxidant Study (SOS) in the 1990s, gas data were acquired at CTR, OAK, and YRK (Atmospheric Research and Analysis, 2013). A longer time period (1995e2011) was examined using the earlier data. Since NO2 measurements were not made in the earlier years, NOy NO (denoted NOw) was computed; 1h maximum NOw was substituted for 1-h maximum NO2 and noon e 4 p.m. NOw was substituted for noon e 4 p.m. NOz. Comparisons were made of the various versions of the GAM. The statistical fit of each model was evaluated using (1) the R2 between predicted and observed peak 8-h O3, (2) the root mean square error (RMSE), and (3) forecasting metrics (Table S1). We utilized SEARCH meteorological measurements for the daily maximum and minimum T, mid-day RH, and morning and afternoon wind speed and direction. We obtained 1995e2007 transport distance and direction, derived from air mass back trajectories computed using the HYSPLIT model (Draxler et al., 2009), from the EPA’s “Omnibus Meteorological Data Set” (OMD) (EPA, 2008). Trajectories were initiated each day at 12-noon local time at 300 m above ground level, and transport distance and direction were determined from trajectory locations 24 h prior to initiation (EPA, 2008). We used the HYSPLIT model to reproduce OMD transport distance and direction for a comparison time period for starting
NO2 12e4 p.m.
0.03 0.00 0.02 0.06 0.00 0.01 0.02 0.01
4.57 0.58 1.68 6.09 0.44 1.02 2.05 1.25
0.09 0.02 0.04 0.09 0.01 0.03 0.05 0.04
NOy 12e4 p.m. 8.07 2.13 3.53 11.4 1.61 2.57 3.91 3.77
0.12 0.03 0.06 0.15 0.02 0.04 0.08 0.07
NOz 12e4 p.m. 2.21 1.44 1.39 3.29 1.09 1.34 1.32 2.20
0.04 0.02 0.02 0.05 0.02 0.02 0.03 0.04
locations at airports in Atlanta, Birmingham, and Mobile and to determine additional back trajectories for each day of 2007 through 2011. 3. Results and discussion 3.1. Sensitivity of daily peak 8-h O3 to meteorological factors Sensitivities are determined as the deviation of peak 8-h O3 from its mean as a function of each single predictor variable while holding constant the contributions from all other predictor variables (Camalier et al., 2007). Variations in mid-day RH and daily maximum temperature (Tmax) cause peak 8-h O3 concentrations to vary by w30 percent from mean peak 8-h O3, the largest effects attributable to any variables (Fig. 1). Peak 8-h O3 decreases with increasing mid-day RH at all sites, leveling off at RH > w80 percent at 5 of the 8 sites (Fig. 1). At the SEARCH sites, mid-day RH correlates inversely with mean daily solar radiation (r2 w 0.3e0.6). Substitution of solar radiation for mid-day RH in the GAM indicates that O3 increases with increasing solar radiation (Fig. S1). When both RH and solar radiation were included, only RH was statistically significant. RH measurements are widely available, and the R2 values were slightly higher (by 0.02e0.05) with mid-day RH than with mean daily solar radiation. Peak 8-h O3 increases with increasing Tmax by up to w1 percent per degree C, leveling off or reversing for Txmax > 25e30 C at three sites (Fig. 1). Sensitivity to Tmax remained when solar radiation, rather than RH, was included in the model (Fig. S1). Transport distance and direction cause variations of up to w10 percent from mean peak 8-h O3 (Fig. 2). Back trajectories originating from the south (w150e200 ) exhibit peak 8-h O3 that is w5e10 percent lower than average. For the coastal sites, trajectories originating from w300 exhibit peak 8-h O3 that is w5e7 percent higher than average, and trajectories originating from w0 to 60 exhibit peak 8-h O3 that is w3e5 percent higher than average (Fig. 2). For the inland sites, trajectories originating from w30 to 90 exhibit peak 8-h O3 that is w5e7 percent higher than average. The result accords with higher O3 in association with weak easterly flow around large, high-pressure systems (Wolff and Lioy, 1980). The sensitivities to transport distance show declining O3 as trajectory distance increases from zero to w600 km, consistent with association of higher O3 concentrations with air mass stagnation rather than transport (Blanchard et al., 2013b). The coastal sites showed peak O3 that was w2e5 percent higher than average when local morning winds were southerly (w160e 200 ), as did OLF for afternoon winds (Fig. S2). GFP and PNS exhibited peak O3 that was w5e7 percent higher than average when afternoon winds were southeasterly (w120 ). These results suggest the influence of sea-breeze and land-breeze circulation at coastal sites. More varied patterns are evident for inland sites. Higher BHM peak O3 concentrations correspond to morning upvalley flow (from the metropolitan core) and afternoon down-
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Fig. 1. Predicted effects of mid-day RH and daily maximum temperature on peak 8-h O3. The predictions are relative to the site mean peak 8-h O3 concentrations.
Fig. 2. Predicted effects of transport distance and direction on peak 8-h O3. The predictions are relative to the site mean peak 8-h O3 concentrations.
195
196
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
valley flow. JST exhibits higher peak O3 concentrations that correspond to morning flow from the metropolitan core (w150 ) and reverse afternoon flow (w270e320 ). YRK, located w70 km ENE of the Atlanta metropolitan core, exhibits w3e5 percent higher than average peak O3 concentrations corresponding to morning flow from the metropolitan core (w100e150 ) and w7 percent higher peak O3 concentrations when afternoon winds are from w90 .
RO2 þ NO0RO þ NO2
3.2. Sensitivity of peak O3 to NO, NO2, and NOz
HO2 þ NO0HO þ NO2
RH þ HO 0R þ H2 O
R þ O2 0RO2
NO þ O3 0NO2 þ O2
(R5)
The sensitivities of peak 8-h O3 to NO, NO2, and NOz (Fig. 3) are interpreted with reference to known O3 formation and loss reactions, R1 through R10 (Seinfeld, 1986):
(R4)
(R6)
(R7)
HO þ NO2 0HNO3
(R8)
CH3 CðOÞO2 þ NO0CH3 O2 þ NO2 þ CO2
(R9)
(R1)
NO2 þ hn0NO þ O
(R2)
O þ O2 0O3
(R3)
CH3 CðOÞO2 þ NO2 5CH3 CðOÞO2 NO2
(R10)
Reaction of O3 with NO removes O3 (R1), so higher concentrations of NO result in lower O3 at all sites; peak 8-h O3 deviates from its long-term mean by 10e30 percent due to variations in daily NO maxima (Fig. 3). Photolysis of NO2 yields O3 (R2 and R3), so
Fig. 3. Predicted effects of maximum 1-h NO, morning NO2, and afternoon NOz concentrations on peak 8-h O3, expressed as percentage deviations relative to site mean peak 8-h O3.
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200 Table 2 Regression of daily peak 8-h O3 against afternoon (noone4 p.m.) NOz concentration, listed in order of increasing slope. Site
Type
N days
r2
Intercept (ppbv)
JST PNS YRK BHM OLF GFP CTR OAK
Urban Urban Rural Urban Suburban Urban Rural Rural
1792 1047 1814 1250 1657 1271 1759 1521
0.299 0.210 0.439 0.318 0.358 0.327 0.439 0.329
38.57 37.50 37.48 34.09 35.26 34.80 32.27 32.96
0.58 0.62 0.45 0.68 0.48 0.63 0.50 0.52
Slope (ppbv per ppbv) 4.02 6.14 6.28 6.36 9.04 9.84 11.19 11.32
0.15 0.37 0.17 0.26 0.30 0.40 0.30 0.42
higher morning NO2 concentrations result in higher peak 8-h O3 (Fig. 3). Over the range of observed NO2 concentrations, peak 8-h O3 deviates from its long-term mean by 20e30 percent due to NO2 variations. Accumulation of O3 occurs when oxidation of VOCs (R4) generates radical species () that convert NO to NO2 without consuming O3 (R5eR7). Radical reactions also produce NOx reaction products comprising NOz, such as nitric acid (HNO3) (R8) and peroxyacetlynitrate (PAN) (R9 and R10). Thus, the afternoon NOz concentration is an indicator of the amount of photochemical activity and of O3 production efficiency (OPE) (Trainer et al., 1993), defined as the mean number of O3 molecules formed per NOx molecule consumed. Peak 8-h O3 ranged from w20e40 percent lower than average to w30e60 percent higher than average as NOz concentrations ranged from their lowest to highest values (Fig. 3). While reactions R8 and R10 each convert NO2 to a reaction product, R8 terminates radical production; in contrast, R10 is reversible because PAN decomposes as temperature increases. PAN therefore is a reservoir of NO2 for O3 production (Singh and Hanst, 1981; Singh, 1987; Pollack et al., 2013). The observed sensitivity of O3 to NOz potentially varies depending on the relative proportions of PAN and HNO3 (Pollack et al., 2013). Application of the GAM to 1995e2011 data from CTR, OAK, and YRK yielded sensitivities comparable to those for the shorter 2002e
197
2011 time period. This result supports earlier conclusions that regional O3 concentrations are NOx-sensitive (e.g. Chameides et al., 1988; Chameides and Cowling, 1995). The sensitivity coefficients for NO and NO2 did not vary if the model included or excluded afternoon NOz concentrations (Figs. S3 and S4). This result indicates that cumulative photochemical activity is a contributing factor separate from same-day mass loadings of NO and NO2. The sensitivity coefficients for NOz reflect both the amount of NOx that has been converted to NOz by afternoon and the OPE. Mean afternoon NOz is 27e39 percent of mean afternoon NOy at urban sites, 52 percent at OLF (suburban), and 58e68 percent at non-urban sites (Table 1). Therefore, the average reservoir of unreacted afternoon NOx ranged from 32 to 73 percent among sites. OPE has been estimated as the slope of the regression line of O3 versus NOz. Table 2 shows that OPE estimated in this way ranges from 4:1 at JST to 11:1 at CTR and OAK, with uncertainties ranging from 0.2 to 0.4 (one standard error). These values are consistent with previous studies (e.g., Trainer et al., 1993; Demerjian et al., 2013). In contrast, the GAM sensitivity curves are nonlinear, varying among sites and within different NOz concentrations ranges at each site. The JST, YRK, and CTR results in Fig. 3 are quasi-linear within the range of 5e10 ppbv NOz with O3/NOz slopes of 1.4, 3.1, and 1.8, respectively; within the range of 1e5 ppbv NOz, the JST, YRK, and CTR O3/NOz slopes are 3.5, 5.0, and 7.1, respectively. For afternoon NOz concentrations less than 1 ppbv, the YRK and CTR slopes increase to 9.2 and 12.7, respectively, while the JST slope reverses from positive to negative. The GAM was used to project O3 responses to future reductions of ambient NO, NO2, and NOz concentrations by replacing the 2009 through 2011 NO, NO2, and NOz concentrations with one-half, onethird, and 10 percent of their actual values. When ambient NO, NO2, and NOz concentrations were 10 percent of the observed 2009e 2011 values, the projected 50th and 95th percentiles of peak 8-h O3 concentrations ranged from 21 to 38 ppbv and from 32 to 50 ppbv, respectively (Fig. 4; see also S5). The projected 50th percentiles were 26e28 ppbv at three non-urban sites (CTR, OAK, OLF) and 35e 38 ppbv at three urban (BHM, JST, GFP) and one urban-influenced (YRK) site; the projected values for PNS (21 ppbv at 50th
Fig. 4. Distributions of predicted daily peak 8-h O3 changes with concentrations of NO, NO2, and NOz each reduced to 10 percent of the observed 2009e2011 concentrations.
198
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Fig. 5. GAM-predicted mean annual peak 8-h O3 change due to (top) NOz versus percent deviation of mean annual afternoon NOz from overall site mean NOz and (bottom) nonbiogenic NMOC versus mean annual non-biogenic NMOC concentration. The predicted O3 changes are expressed as percentage deviations relative to site mean peak 8-h O3.
percentile) were based on 2009 data only and may therefore be less reliable. Mean afternoon NOy concentrations would be w0.2 ppbv at CTR, OAK, and OLF, and w0.4e1 ppbv at BHM, JST, and YRK, when reduced to 10 percent of their 2009e2011 means (Table 1). Assuming that the GAM projections reliably represent the consequences of large (50e90 percent) precursor reductions, the 90 percent decreases of NO, NO2, and NOz suggest that peak 8-h O3 concentrations are unlikely to fall below limits within the range of w25e40 ppbv. 3.3. Sensitivity of peak O3 to NMOC Additional applications of the GAM yielded a significant relationship between non-biogenic NMOC and peak O3 at JST but not at YRK (Fig. S6). The results agree with reported sensitivity of O3 to both NMOC and NOy in Atlanta but NOx-sensitive O3 at non-urban YRK (Blanchard et al., 2010a). Projections based on replacing observed ambient NO, NO2, NOz, and non-biogenic NMOC concentrations by one-half, one-third, and 10 percent of their actual values yielded median peak 8-h O3 concentrations of 25 ppbv at YRK and 30 ppbv at JST (Fig. S7). For JST, the GAM results indicated that maximum O3 concentrations were weakly sensitive (decreasing) to isoprene concentrations (either as 24-h or morning
averages). For YRK, the sensitivity of maximum O3 concentrations to isoprene concentrations was weak or negligible (Fig. S6). These results are inconsistent with the expectation that isoprene should increase O3 maxima. The inconsistency may indicate an uncertainty in measuring isoprene reliably because of its known high reactivity. Alternatively, isoprene-O3 chemistry may be complicated by air mass movement and mixing, by the potential lack of NMOC response in NOx sensitive regimes, or by the responses of vegetation to oxidant exposure, temperature and soil moisture (e.g., Loreto et al., 2001; Sharkey et al., 2008). The apparent anomalies were studied further by examining average diurnal concentrations of isoprene and non-biogenic NMOC that were differentiated by O3, NOx, NOz concentrations and by meteorological properties. Evaluation of these results did not provide a simple explanation. 3.4. Temporal variations Both observed and predicted daily peak 8-h O3 concentrations varied among the days of the week (Table S2). Large (w10e70 percent) reductions of NO, NO2, and NOz concentrations occurred on Sundays compared with midweek (Table S3), but did not account for day-of-week variations in peak O3. The day of the week was a statistically significant predictor in the GAM, which suggests
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
that O3 formation depends on the ambient concentrations of species in addition to NO, NO2, and NOz. The GAM predicted interannual variations in mean annual peak 8-h O3 that matched the observed year-to-year variations at all sites (Fig. S8). Variations in weather were the largest contributor to interannual variations (Fig. S9). Trends in the unadjusted mean annual peak 8-h O3 concentrations were not statistically significant (at p ¼ 0.05) at 7 of the 8 sites (average rate of decrease was 0.37 ppbv per year) (Table S4). When adjusted for weather, trends were statistically significant at 5 of the 8 sites and the average decrease among all sites was 0.90 ppbv per year. Camalier et al. (2007) interpreted trends in the meteorologically-adjusted peak O3 as evidence for effects of emission reductions. When adjusted for meteorological effects and for changing ambient concentrations of NO, NO2, and NOz, the remaining O3 trends were statistically significant at only CTR and the multi-site average decrease was 0.25 ppbv per year. This result indicates that the changing concentrations of NO, NO2, and NOz explained the majority of the observed downward O3 trend; the residual trend is possibly due to factors not included in the model. When normalized as a percent-to-percent comparison, a consistent relationship between O3 and NOz concentrations emerges (Fig. 5). The sites exhibit an average 0.2 percent change in mean annual peak 8-h O3 (0.13e0.18 at urban sites, 0.21 at OLF, and 0.22e0.27 at rural sites) for each percent NOz change. The mean annual afternoon concentrations of NO, NO2, and NOz declined at all sites and mean afternoon NOz as a percentage of mean afternoon NOy showed no trend (Fig. S10). Therefore, the principal determinant of the average afternoon NOz concentration was the amount of NOx emitted and present in the atmosphere. However, the modeled NOz effect on peak O3 results from changes in both NOx mass loading and OPE, and the latter is linked to NMOC concentrations through rates of radical production and removal. At the urban sites, afternoon NOz is typically w30 percent of NOy; in contrast, NOz at the rural sites is w65 percent (Table 1). If urban conditions represent O3 photochemistry of about one day, the urban-rural differences suggest a regional aging or carryover of two days or more. For JST, the interannual variations of non-biogenic NMOC concentrations yielded variations of up to w3 percent from the mean peak 8-h O3, compared with the NOz effect of w10 percent (Fig. 5). 4. Conclusions The monitoring record from the eight sites of the SEARCH network offers a rich data set for investigating the observed response of O3 to emission changes in the southeastern U.S. Application of a generalized additive model (GAM) to a ten-year record of O3, NO, NO2, NOz, and meteorological measurements indicates the importance of temperature, humidity, and NOx to O3 maxima in the southeastern U.S. The GAM estimates of O3 sensitivity to NO, NO2, and NOz are qualitatively consistent with expectations from photochemistry, and are potentially important for evaluation of PAQSM predictions. The relationships of O3 to surface wind speed and direction, and to 24-h back trajectory distance and direction, suggest that regional and local O3 components are dominated by stagnation on a w400 km scale, and not by transport over greater distances such as has been observed in the northeastern U.S. When NO, NO2, and NOz concentrations in the model were reduced to 10 percent of their observed 2009e2011 values, the projected 50th percentiles (26e38 ppbv) and 95th percentiles (36e50 ppbv) of peak 8-h O3 provide evidence for an irreducible O3 component. In Atlanta, O3 was less sensitive to NMOC than to NOx; at a rural location outside Atlanta, no evidence was found for sensitivity of O3 to NMOC.
199
Acknowledgments This study was sponsored by the Southern Company. The authors thank Eric Edgerton and John Jansen for reviewing preliminary drafts of this manuscript. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.atmosenv.2014.02.006. References Atmospheric Research and Analysis, 2013. http://www.atmospheric-research.com/ studies/SEARCH/index.html (accessed July, 2013). Blanchard, C.L., Hidy, G.M., Tanenbaum, S., 2010a. NMOC, ozone, and organic aerosol in the southeastern states, 1999e2007: 2. Ozone trends and sensitivity to NMOC emissions in Atlanta, Georgia. Atmospheric Environment 44, 4840e 4849. http://dx.doi.org/10.1016/j.atmosenv.2010.07.030. Blanchard, C.L., Hidy, G.M., Tanenbaum, S., Rasmussen, R., Watkins, R., Edgerton, E., 2010b. NMOC, ozone, and organic aerosol in the southeastern states, 1999e 2007: 1. Spatial and temporal variations of NMOC concentrations and composition in Atlanta, Georgia. Atmospheric Environment 44, 4827e4839. http:// dx.doi.org/10.1016/j.atmosenv.2010.08.036. Blanchard, C.L., Hidy, G.M., Tanenbaum, S., Edgerton, E.S., Hartsell, B.E., 2013a. The Southeastern Aerosol Research and Characterization (SEARCH) Study: temporal trends in gas and PM concentrations and composition, 1999-2010. Journal of the Air & Waste Management Association 63, 247e259. http://dx.doi.org/10.1080/ 10962247.2012.748523. Blanchard, C.L., Hidy, G.M., Tanenbaum, S., Edgerton, E.S., Hartsell, B.E., 2013b. The Southeastern Aerosol Research and Characterization (SEARCH) study: spatial variations and chemical climatology, 1999e2010. Journal of the Air & Waste Management Association 63, 260e275. http://dx.doi.org/10.1080/ 10962247.2012.749816. Camalier, L., Cox, W., Dolwick, P., 2007. The effects of meteorology and their use in assessing ozone trends. Atmospheric Environment 41, 7127e7137. Chameides, W., Lindsay, R., Richardson, J., Kiang, C., 1988. The role of biogenic hydrocarbons in urban photochemical smog: Atlanta as a case study. Science 241, 1473. Chameides, W.L., Cowling, E.B., 1995. The State of the Southern Oxidants Study: Policy-relevant Findings in Ozone Pollution Research, 1988e1994. Southern Oxidants Study, College of Forest Resources, North Carolina State University, Raleigh, NC. Demerjian, K., Schwab, J.J., Roychowdhury, U., Sun, Y., Zhang, Q., 2013. Research Measurements of Chemical Constituents and Related Processes Affecting New York State’s Regional and Urban O3 and PM Air Quality. Prepared for New York State Energy Research and Development Authority, Albany, New York. Contract number 10602. Draxler, R., Stunder, B., Rolph, G., Stein, A., Taylor, A., 2009. Hysplit4 User’s Guide. National Oceanic and Atmospheric Administration. http://www.arl.noaa.gov/ documents/reports/hysplit_user_guide.pdf (accessed April 20, 2009). Edgerton, E.S., Hartsell, B.E., Saylor, R.D., Jansen, J.J., Hansen, D.A., Hidy, G.M., 2005. The Southeastern Aerosol Research and Characterization Study: part 2 e filterbased measurements of PM2.5 and PMcoarse mass and composition. Journal of the Air & Waste Management Association 55, 1527e1542. Edgerton, E.S., Hartsell, B.E., Saylor, R.D., Jansen, J.J., Hansen, D.A., Hidy, G.M., 2006. The Southeastern Aerosol Research and Characterization Study, part 3: continuous measurements of fine particulate matter mass and composition. Journal of the Air & Waste Management Association 56, 1325e1341. Edgerton, E.S., Saylor, R.D., Hartsell, B.E., Jansen, J.J., Hansen, D.A., 2007. Ammonia and ammonium measurements from the southeastern United States. Atmospheric Environment 41, 3339e3351. Edgerton, E.S., Casuccio, G.S., Saylor, R.D., Lersch, T.L., Hartsell, B.E., Jansen, J.J., Hansen, D.A., 2009. Measurements of OC and EC in coarse particulate matter in the southeastern United States. Journal of the Air & Waste Management Association 59, 78e90. Gego, E., Porter, P.S., Gilliland, A., Rao, S.T., 2007. Observation-based assessment of the impact of nitrogen oxides emissions reductions on ozone air quality over the eastern United States. Journal of Applied Meteorology and Climatology 46, 994e1008. http://dx.doi.org/10.1175/JAM2523.1. Gego, E., Gilliland, A., Godowitch, J., Rao, S.T., Porter, P.S., Hogrefe, C., 2008. Modeling analyses of the effects of changes in nitrogen oxides emissions from the electric power sector on ozone levels in the eastern United States. Journal of the Air & Waste Management Association 58, 580e588. Gilliland, A., Hogrefe, C., Pinder, R.W., Godowitch, J., Foley, K.L., Rao, S.T., 2008. Dynamic evaluation of regional air quality models: assessing changes in O3 stemming from changes in emissions and meteorology. Atmospheric Environment 42, 5110e5123. http://dx.doi.org/10.1016/j.atmosenv.2008.02.018. Godowitch, J., Gilliland, A., Draxler, R.R., Rao, S.T., 2008. Modeling assessment of point source NOx emission reductions on ozone levels in the eastern United States. Atmospheric Environment 42, 87e100.
200
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Hansen, D.A., Edgerton, E.S., Hartsell, B.E., Jansen, J.J., Hidy, G.M., Kandasamy, K., Blanchard, C.L., 2003. The Southeastern Aerosol Research and Characterization Study (SEARCH): 1. Overview. Journal of the Air & Waste Management Association 53, 1460e1471. Hansen, D.A., Edgerton, E., Hartsell, B., Jansen, J., Burge, H., Koutrakis, P., Rogers, C., Suh, H., Chow, J., Zielinska, B., McMurry, P., Mulholland, J., Russell, A., Rasmussen, R., 2006. Air quality measurements for the aerosol research and inhalation epidemiology study. Journal of the Air & Waste Management Association 56, 1445e1458. Hastie, T.J., Tibshirani, R.J., 1990. Generalized Additive Models. Chapman and Hall/ CRC, Boca Raton, p. 335. Hidy, G.M., 2000. Ozone process insights from field experiments e part 1: overview. Atmospheric Environment 34, 2001e2022. Hidy, G.M., Brook, J.R., Demerjian, K.L., Molina, T. Luisa, Pennell, W.T., Scheffe, R.D. (Eds.), 2011. Technical Challenges of Multipollutant Air Quality Management. Springer, New York. Liu, S.C., Trainer, M., Fehsenfeld, F.C., Parrish, D.D., Williams, E.J., Fahey, D.W., Hubler, G., Murphy, P.C., 1987. Ozone production in the rural troposphere and the implications for regional and global ozone distributions. Journal of Geophysical Research 92 (D4), 4191e4207. Loreto, F., Mannozzi, M., Maris, C., Nascetti, P., Ferranti, F., Pasqualini, S., 2001. Ozone quenching properties of isoprene and its antioxidant role in leaves. Plant Physiology 126, 993e1000. National Research Council, 1991. Rethinking the Ozone Problem in Urban and Regional Air Pollution. National Academy Press, Washington, D.C, p. 489. NARSTO, 2000. An Assessment of Tropospheric Ozone Pollution. Report 1000040. NARSTO, Pasco, WA (Also available from: EPRI, Palo Alto, CA). Pollack, I., Ryerson, T., Trainer, M., Neuman, J., Roberts, J., Parrish, D., 2013. Trends in ozone its precursors and related secondary oxidation products in Los Angeles, California: a synthesis of measurements from 1960e2010. Journal of Geophysical Research 118, 5891e5911. Saylor, R., Edgerton, E.S., Hartsell, B.E., Baumann, K., Hansen, D.A., 2010. Continuous gaseous and total ammonia measurements from the southeastern aerosol research and characterization (SEARCH) study. Atmospheric Environment 44, 4994e5004. Seinfeld, J.H., 1986. Atmospheric Chemistry and Physics of Air Pollution. John Wiley and Sons, New York. Sharkey, T., Wiberley, A., Donohue, A., 2008. Isoprene emissions from plants: why and how? Annals of Botany 10, 5e18.
Sillman, S., Logan, J.A., Wofsy, S.C., 1990. The sensitivity of ozone to nitrogen oxides and hydrocarbons in regional ozone episodes. Journal of Geophysical Research 95 (D2), 1837e1851. Singh, H.B., Hanst, P.L., 1981. Peroxyacetyl nitrate (PAN) in the unpolluted atmosphere: an important reservoir for nitrogen oxides. Geophysical Research Letters 8, 941e944. Singh, H., 1987. Reactive nitrogen in the troposphere: chemistry and transport of NOx and PAN. ES&T 21, 320e327. The R Project, 2009. The R Project for Statistical Computing. http://www.R-project. org (accessed January 25, 2013). Trainer, M., Parish, D.D., Buhr, M.P., Norton, R.B., Fehsenfeld, F.C., Anlauf, K.G., Bottenheim, J.W., Tang, Y.Z., Wiebe, H.A., Roberts, J.M., Tanner, R.L., Newman, L., Bowersox, V.C., Meagher, J.F., Olszyna, K.J., Rodgers, M.O., Wang, T., Berresheim, H., Demerjian, K.L., Roychowdhury, U.K., 1993. Correlation of ozone with NOy in photochemically aged air. Journal of Geophysical Research 98 (D2), 2917e2925. U.S. Environmental Protection Agency, 2007. Guidance on the Use of Models and Other Analyses for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze. EPA-454/B-07-002. U.S. Environmental Protection Agency, 2008. Details of the Omnibus Meteorological Database (OMD or “MetDat”). http://www.epa.gov/scram001/meteorology/ omnibus_meteorological_data_set.pdf (accessed December 2008). U.S. Environmental Protection Agency, 2009. R Code to Model Daily Maximum 8hour Ozone as a Function of Daily Meteorological Variables Plus Terms to Account for Calendar Day, Dow-of-week and a Term “Year” to Account for Trends over Time. http://www.epa.gov/air/airtrends/weather/metadj_code.txt (accessed January 25, 2013). U.S. Environmental Protection Agency, 2011a. Clean Air Interstate Rule (CAIR). http://www.epa.gov/airmarkets/progsregs/cair/index.html (accessed April 16, 2011). U.S. Environmental Protection Agency, 2011b. SO2 Reductions and Allowance Trading under the Acid Rain Program. http://www.epa.gov/airmarkets/ progsregs/arp/s02.html (accessed April 16, 2011). U.S. Environmental Protection Agency, 2012. The NOx SIP Call. http://www.epa.gov/ airmarkets/progsregs/nox/noxsipcall.html (accessed March 14, 2012). Wolff, G.T., Lioy, P.J., 1980. Development of an ozone river associated with synoptic scale episodes in the eastern United States. Environmental Science & Technology 14, 1257e1260.
Contents lists available at ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
Ozone in the southeastern United States: An observation-based model using measurements from the SEARCH network C.L. Blanchard a, *, G.M. Hidy b, S. Tanenbaum a a b
Envair, 526 Cornell Avenue, Albany, CA 94706, USA Envair/Aerochem, 6 Evergreen Dr, Placitas, NM 87043, USA
h i g h l i g h t s Observed ozone response to weather and ozone precursors is modeled. Ozone responses to changing concentrations of ozone precursors are projected. Empirically-determined results may be compared to photochemical model predictions. Model is useful for relating ozone trends to precursor trends and weather.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 28 August 2013 Received in revised form 28 January 2014 Accepted 6 February 2014 Available online 7 February 2014
A generalized additive model (GAM) is used to examine the influence of meteorological factors, nitrogen oxides (NOx ¼ NO þ NO2), and non-methane hydrocarbons (NMOC) on daily peak 8-h ozone (O3) concentrations. Application to 2002e2011 monitoring data from the Southeastern Aerosol Research and Characterization (SEARCH) program showed sensitivity of peak 8-h O3 to morning concentrations of nitric oxide (NO) and nitrogen dioxide (NO2) and to afternoon concentrations of NO2 reaction products (NOz). Peak O3 decreased with increasing NO and increased with increasing NO2 concentrations, consistent with reactions involving O3, NO, and NO2. Ozone production efficiency (OPE), estimated from the modeled relation between peak 8-h O3 and afternoon NOz, was w40e100 percent higher at rural compared to urban sites. OPE was nonlinear at all sites, decreasing with increasing NOz concentration. The mean ratio of NOz/NOy showed a two-fold increase from urban to rural sites, associated with chemical aging in stagnant air masses from one day (urban sites) to two or more days (non-urban sites). Peak 8-h O3 concentrations in Atlanta were sensitive to concentrations of both non-biogenic NMOC and NOz. Non-urban Yorkville, Georgia, peak 8-h O3 concentrations were sensitive to NOz but not to nonbiogenic NMOC concentrations. The results are consistent with expected NMOC and NOx sensitivity in urban and non-urban locales. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Ozone SEARCH Meteorology NOx VOC Generalized additive model
1. Introduction Ozone (O3) formation is a nonlinear function of sunlight acting on ambient volatile organic compounds (VOC) and oxides of nitrogen (NOx, NO þ NO2). Tropospheric O3 mixing ratios (concentrations) are affected by solar intensity, the rate of O3 formation, the rate of dispersion of O3 and its precursors, meteorological factors, and transport of urban plumes. Since 1970, emission control programs have reduced emissions and ambient concentrations of NOx and VOC, resulting in declining
* Corresponding author. E-mail address: [email protected] (C.L. Blanchard). http://dx.doi.org/10.1016/j.atmosenv.2014.02.006 1352-2310/Ó 2014 Elsevier Ltd. All rights reserved.
O3 concentrations in all major metropolitan areas of the U.S. However, slow rates of O3 reduction have led to concern about the reliability of predictive tools and the fundamental understanding of O3 formation (e.g., National Research Council, 1991; Hidy et al., 2011). Photochemical Air Quality Simulation Models (PAQSMs) have been used for over forty years to estimate the influence of precursor emissions on O3. Complementary, observation-based modeling developed by the 1990s (e.g, NARSTO, 2000; Hidy, 2000) and is accepted by the U.S. Environmental Protection Agency (EPA) and state agencies for use in “weight-of-evidence” plans to achieve the National Ambient Air Quality Standard (NAAQS) and state standards for O3 (EPA, 2007).
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Observational and PAQSM studies suggest that NOx is the limiting O3 precursor throughout much of the non-urban eastern United States (Liu et al., 1987; Sillman et al., 1990; Trainer et al., 1993; Chameides and Cowling, 1995). In the southeastern U.S., biogenic emissions of isoprene and anthropogenic NOx emissions affect O3 (e.g., Chameides et al., 1988). Evidence indicates responsiveness of post-1990s O3 formation to changes in NOx emissions from electric generating units (EGUs) in the eastern U.S. (Gego et al., 2007, 2008; Gilliland et al., 2008; Godowitch et al., 2008). Beginning in 1995, EPA’s acid rain program (CAAA, Title IV) implemented a nationwide reduction of SO2 and NOx emissions from EGUs (EPA, 2011a). The 2004 NOx state implementation plan (SIP) Call required further reductions in EGU NOx emissions in twenty-one eastern states (EPA, 2012). The Clean Air Interstate Transport Rule required 70 percent reductions of EGU emissions of SO2 and NOx in 28 eastern states (EPA, 2011b). The monitoring record from the Southeastern Aerosol Research and Characterization (SEARCH) network offers a rich data set for investigating the response of O3 to emission changes. Between 1999 and 2010, the annual O3-season (MarcheOctober) 95th percentiles of the peak daily 8-h O3 concentrations declined at SEARCH sites by 1.1 0.4 to 2.4 0.6 ppbv per year (Blanchard et al., 2013a). O3 declines were not attributable to a specific precursor using simple trend analyses (Blanchard et al., 2013a). Previous work suggests that the highest peak 8-h O3 concentrations in Atlanta, Georgia are sensitive to both ambient NMOC and NOy (NOx plus other oxidized nitrogen species, including, e.g., HNO3), but meteorological factors have a larger effect on day-to-day O3 variations (Blanchard et al., 2010a). In this paper, we use a generalized additive statistical model (GAM) to relate daily peak 8-h O3 concentrations to weather, to ambient concentrations of O3 precursors (NO and NO2), and to NOx reaction products (NOz) at SEARCH sites from 2002 to 2011. The approach is applicable to any well-instrumented monitoring location. 2. Methods 2.1. Ambient air quality measurements Hourly measurements of gases (CO, SO2, NO, NO2, NOy, HNO3, O3, and NH3) were obtained from SEARCH public archives (Atmospheric Research and Analysis, 2013). Network operations and measurement methods are documented in Hansen et al. (2003, 2006), Edgerton et al. (2005, 2006; 2007; 2009), and Saylor et al. (2010). Eight sites are located in the southeastern U.S.: Pensacola, Florida (PNS) and Gulfport, Mississippi (GFP), urban coastal sites (w5 km and 1.5 km from the shoreline, respectively); Pensacola e outlying (aircraft) landing field (OLF) and Oak Grove, Mississippi (OAK), non-urban coastal sites near the Gulf (w20 km and 80 km inland, respectively); Atlanta, GeorgiaeJefferson Street (JST) and North Birmingham, Alabama (BHM), urban inland sites; and Yorkville, Georgia (YRK) and Centreville, Alabama (CTR), non-urban inland sites. All sites measure meteorological parameters and gases at 10 m height (Hansen et al., 2003; Edgerton et al., 2007; Saylor et al., 2010). Measurements of NMOC, a VOC subset, were made with 24-h canister samples collected every day at JST through 2008 (Blanchard et al., 2010b). Additional 24-h and hourly NMOC measurements were made at Photochemical Assessment Monitoring Stations (PAMS) located at YRK and other sites in the Atlanta area (Blanchard et al., 2010b). Spatial and temporal variations of SEARCH air pollutant concentrations from 1999 to 2010 are described in Blanchard et al. (2013a, 2013b). We used data from 2002 through 2011, because measurements of NO2 began in 2002. Daily averages were created from hourly
193
data. In addition, daily peak 1-h and peak 8-h NO, NO2, and O3 concentrations were determined for each site and day. Morning (sample start hours of 7 through 10 a.m., i.e., sampling interval of 7e11 a.m.) and afternoon (sample start hours of 12 noon through 3 p.m., i.e., sampling interval of 12e4 p.m.) concentrations of NO, NO2, NOy, and NOz (NOz ¼ NOy NO NO2) were also computed. Table 1 lists mean NO, NO2, NOy, and NOz concentrations.
2.2. Statistical model EPA developed a GAM that predicts the highest daily peak 8h O3 within a metropolitan area as a function of surface and upper-air meteorological variables (Camalier et al., 2007). The EPA model (EPA, 2009) runs under the open-source R software system (The R Project, 2009). Blanchard et al. (2010a) describe an adaptation of the EPA statistical model. Further adaptation is implemented here. The GAM is expressed as:
lðO3 Þi ¼ m þ f1 ðx1 Þi þ . þ fm ðxm Þi þ g1 ðy1 Þi þ . þ gn ðyn Þi þ h1 ðz1 Þ þ . þ hp zp þ ei
(1)
Following Camalier et al. (2007), l(O3)i is the logarithm of the peak 8-h O3 on day “i,” but either a different O3 metric or a function other than the logarithm could be used. We apply the model site by site. For convenience, the terms f1(x1)i through fm(xm)i represent the effects of meteorological variables on peak 8-h O3, and g1(y1)i through gn(yn)i represent the effects of ambient concentrations of O3 precursors on O3. For consistency with l(O3)i, the variables y1 through yn are logarithms of air quality measurements. The terms h1(z1) through hp(zp) represent the effects of temporal variables, including “day of week” and “year”. The last term, ei, is the difference between observed and predicted O3 (error). Each term expresses the effect of one parameter on daily peak 8-h O3 as a deviation from the long-term mean, “m” (logarithm of peak 8-h O3 averaged overall days). “Day of week” and “year” are categorical variables and are used to represent weekly cycles and trend, respectively. Trend need not be linear or monotonic. The GAM uses natural splines (Hastie and Tibshirani, 1990) to model nonlinear dependence of O3 on other predictor variables. Meteorological and air quality data were used to predict daily peak 8-h O3 at each SEARCH site during March through October of 2002 through 2011. Camalier et al. (2007) found that the most consistently significant predictors of peak 8-h O3 concentrations in the 39 eastern U.S. metropolitan areas studied were: (1) daily maximum surface temperature (T), (2) mid-day (10 a.m.e4 p.m.) relative humidity (RH), (3) morning (7 a.m.e10 a.m.) average wind speed, (4) afternoon (1 p.m.e4 p.m.) average wind speed, (5) morning (w1200 UTC) difference between 925 mb T and surface T, (6) deviation of morning (w1200 UTC) 850 mb T from 10-year monthly average, (7) air mass transport direction and distance (determined from back trajectories), (8) occurrence of rain (as number of hours), (9) julian day, (10) day of week, and (11) year. We replaced the 925 mb e surface T difference with the 850 mb T e daily minimum surface temperature because the 925 mb height in coastal areas is not always above the marine layer and may not represent temperatures aloft. We eliminated precipitation amount (or hours) and 850 mb T because they were not statistically significant at any site. We added sea-level pressure gradients from Birmingham to Mobile, AL, and from Birmingham to Atlanta. We refit the GAM using the preceding meteorological parameters, but added the daily maximum 1-h NO and the morning NO2 concentrations. The maximum NO concentrations tended to occur in the morning (hours 5 through 9 on 46e73 percent of the days), coinciding with morning traffic, while maximum NO2
194
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Table 1 Mean 2002e2011 NO, NO2, NOy, and NOz concentrations (ppbv). Site
NO Max 1 h
BHM CTR GFP JST OAK OLF PNS YRK
36.81 0.95 5.51 48.36 0.46 4.30 11.25 1.22
NO2 7e11 a.m. 1.13 0.03 0.25 1.5 0.01 0.16 0.51 0.04
12.43 2.06 5.35 17.39 1.37 3.72 5.64 3.50
0.16 0.03 0.08 0.15 0.02 0.05 0.09 0.06
NO 12e4 p.m. 1.29 0.11 0.46 2.01 0.07 0.22 0.55 0.31
concentrations occurred during both morning (hours 5 through 9 on 14e34 percent of the days) and evening (hours 17 through 23 on 28e58 percent of the days). Since evening NO2 maxima follow the time of the peak 8-h O3, the model used morning (7e10 a.m.) NO2 concentrations. We then added afternoon (12 noon to 4 p.m.) concentrations of NOz (NOz ¼ NOy NO NO2). Conceptually, afternoon NOz concentration indicates the amount of photochemical processing (Trainer et al., 1993). Ambient NOz concentrations depend on cumulative mass loading of NO and NO2 and on the reactivity of species in air masses (affecting the fraction of NOx that is converted to reaction products). Afternoon NOz concentrations were weakly correlated or uncorrelated with 1-h NO maxima and morning NO2 concentrations (r2 ¼ 0.0e0.2). Measurements of 24-h average isoprene and non-biogenic NMOC concentrations were incorporated into additional versions of the GAM for JST and YRK. Non-biogenic NMOC was computed as the sum of measured species concentrations less the concentrations of isoprene, a-pinene, b-pinene, d-limonene, and D3-carene (only isoprene was measured at YRK). The computation assumes that the measured C2eC12 alkanes, alkenes, and aromatics are predominantly anthropogenic with the exception of the listed biogenic species. The frequency of 24-h NMOC sampling was daily at JST but once every six days at YRK. Morning (sample starting hours 7e10 a.m.) non-biogenic NMOC and isoprene concentrations were computed and substituted for the 24-h values. Hourly NMOC measurements from the PAMS site at South DeKalb were used for JST. The PAMS hourly measurements were made daily during June through August, except at YRK the sampling frequency was every third day before 2004. During the Southern Oxidant Study (SOS) in the 1990s, gas data were acquired at CTR, OAK, and YRK (Atmospheric Research and Analysis, 2013). A longer time period (1995e2011) was examined using the earlier data. Since NO2 measurements were not made in the earlier years, NOy NO (denoted NOw) was computed; 1h maximum NOw was substituted for 1-h maximum NO2 and noon e 4 p.m. NOw was substituted for noon e 4 p.m. NOz. Comparisons were made of the various versions of the GAM. The statistical fit of each model was evaluated using (1) the R2 between predicted and observed peak 8-h O3, (2) the root mean square error (RMSE), and (3) forecasting metrics (Table S1). We utilized SEARCH meteorological measurements for the daily maximum and minimum T, mid-day RH, and morning and afternoon wind speed and direction. We obtained 1995e2007 transport distance and direction, derived from air mass back trajectories computed using the HYSPLIT model (Draxler et al., 2009), from the EPA’s “Omnibus Meteorological Data Set” (OMD) (EPA, 2008). Trajectories were initiated each day at 12-noon local time at 300 m above ground level, and transport distance and direction were determined from trajectory locations 24 h prior to initiation (EPA, 2008). We used the HYSPLIT model to reproduce OMD transport distance and direction for a comparison time period for starting
NO2 12e4 p.m.
0.03 0.00 0.02 0.06 0.00 0.01 0.02 0.01
4.57 0.58 1.68 6.09 0.44 1.02 2.05 1.25
0.09 0.02 0.04 0.09 0.01 0.03 0.05 0.04
NOy 12e4 p.m. 8.07 2.13 3.53 11.4 1.61 2.57 3.91 3.77
0.12 0.03 0.06 0.15 0.02 0.04 0.08 0.07
NOz 12e4 p.m. 2.21 1.44 1.39 3.29 1.09 1.34 1.32 2.20
0.04 0.02 0.02 0.05 0.02 0.02 0.03 0.04
locations at airports in Atlanta, Birmingham, and Mobile and to determine additional back trajectories for each day of 2007 through 2011. 3. Results and discussion 3.1. Sensitivity of daily peak 8-h O3 to meteorological factors Sensitivities are determined as the deviation of peak 8-h O3 from its mean as a function of each single predictor variable while holding constant the contributions from all other predictor variables (Camalier et al., 2007). Variations in mid-day RH and daily maximum temperature (Tmax) cause peak 8-h O3 concentrations to vary by w30 percent from mean peak 8-h O3, the largest effects attributable to any variables (Fig. 1). Peak 8-h O3 decreases with increasing mid-day RH at all sites, leveling off at RH > w80 percent at 5 of the 8 sites (Fig. 1). At the SEARCH sites, mid-day RH correlates inversely with mean daily solar radiation (r2 w 0.3e0.6). Substitution of solar radiation for mid-day RH in the GAM indicates that O3 increases with increasing solar radiation (Fig. S1). When both RH and solar radiation were included, only RH was statistically significant. RH measurements are widely available, and the R2 values were slightly higher (by 0.02e0.05) with mid-day RH than with mean daily solar radiation. Peak 8-h O3 increases with increasing Tmax by up to w1 percent per degree C, leveling off or reversing for Txmax > 25e30 C at three sites (Fig. 1). Sensitivity to Tmax remained when solar radiation, rather than RH, was included in the model (Fig. S1). Transport distance and direction cause variations of up to w10 percent from mean peak 8-h O3 (Fig. 2). Back trajectories originating from the south (w150e200 ) exhibit peak 8-h O3 that is w5e10 percent lower than average. For the coastal sites, trajectories originating from w300 exhibit peak 8-h O3 that is w5e7 percent higher than average, and trajectories originating from w0 to 60 exhibit peak 8-h O3 that is w3e5 percent higher than average (Fig. 2). For the inland sites, trajectories originating from w30 to 90 exhibit peak 8-h O3 that is w5e7 percent higher than average. The result accords with higher O3 in association with weak easterly flow around large, high-pressure systems (Wolff and Lioy, 1980). The sensitivities to transport distance show declining O3 as trajectory distance increases from zero to w600 km, consistent with association of higher O3 concentrations with air mass stagnation rather than transport (Blanchard et al., 2013b). The coastal sites showed peak O3 that was w2e5 percent higher than average when local morning winds were southerly (w160e 200 ), as did OLF for afternoon winds (Fig. S2). GFP and PNS exhibited peak O3 that was w5e7 percent higher than average when afternoon winds were southeasterly (w120 ). These results suggest the influence of sea-breeze and land-breeze circulation at coastal sites. More varied patterns are evident for inland sites. Higher BHM peak O3 concentrations correspond to morning upvalley flow (from the metropolitan core) and afternoon down-
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Fig. 1. Predicted effects of mid-day RH and daily maximum temperature on peak 8-h O3. The predictions are relative to the site mean peak 8-h O3 concentrations.
Fig. 2. Predicted effects of transport distance and direction on peak 8-h O3. The predictions are relative to the site mean peak 8-h O3 concentrations.
195
196
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
valley flow. JST exhibits higher peak O3 concentrations that correspond to morning flow from the metropolitan core (w150 ) and reverse afternoon flow (w270e320 ). YRK, located w70 km ENE of the Atlanta metropolitan core, exhibits w3e5 percent higher than average peak O3 concentrations corresponding to morning flow from the metropolitan core (w100e150 ) and w7 percent higher peak O3 concentrations when afternoon winds are from w90 .
RO2 þ NO0RO þ NO2
3.2. Sensitivity of peak O3 to NO, NO2, and NOz
HO2 þ NO0HO þ NO2
RH þ HO 0R þ H2 O
R þ O2 0RO2
NO þ O3 0NO2 þ O2
(R5)
The sensitivities of peak 8-h O3 to NO, NO2, and NOz (Fig. 3) are interpreted with reference to known O3 formation and loss reactions, R1 through R10 (Seinfeld, 1986):
(R4)
(R6)
(R7)
HO þ NO2 0HNO3
(R8)
CH3 CðOÞO2 þ NO0CH3 O2 þ NO2 þ CO2
(R9)
(R1)
NO2 þ hn0NO þ O
(R2)
O þ O2 0O3
(R3)
CH3 CðOÞO2 þ NO2 5CH3 CðOÞO2 NO2
(R10)
Reaction of O3 with NO removes O3 (R1), so higher concentrations of NO result in lower O3 at all sites; peak 8-h O3 deviates from its long-term mean by 10e30 percent due to variations in daily NO maxima (Fig. 3). Photolysis of NO2 yields O3 (R2 and R3), so
Fig. 3. Predicted effects of maximum 1-h NO, morning NO2, and afternoon NOz concentrations on peak 8-h O3, expressed as percentage deviations relative to site mean peak 8-h O3.
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200 Table 2 Regression of daily peak 8-h O3 against afternoon (noone4 p.m.) NOz concentration, listed in order of increasing slope. Site
Type
N days
r2
Intercept (ppbv)
JST PNS YRK BHM OLF GFP CTR OAK
Urban Urban Rural Urban Suburban Urban Rural Rural
1792 1047 1814 1250 1657 1271 1759 1521
0.299 0.210 0.439 0.318 0.358 0.327 0.439 0.329
38.57 37.50 37.48 34.09 35.26 34.80 32.27 32.96
0.58 0.62 0.45 0.68 0.48 0.63 0.50 0.52
Slope (ppbv per ppbv) 4.02 6.14 6.28 6.36 9.04 9.84 11.19 11.32
0.15 0.37 0.17 0.26 0.30 0.40 0.30 0.42
higher morning NO2 concentrations result in higher peak 8-h O3 (Fig. 3). Over the range of observed NO2 concentrations, peak 8-h O3 deviates from its long-term mean by 20e30 percent due to NO2 variations. Accumulation of O3 occurs when oxidation of VOCs (R4) generates radical species () that convert NO to NO2 without consuming O3 (R5eR7). Radical reactions also produce NOx reaction products comprising NOz, such as nitric acid (HNO3) (R8) and peroxyacetlynitrate (PAN) (R9 and R10). Thus, the afternoon NOz concentration is an indicator of the amount of photochemical activity and of O3 production efficiency (OPE) (Trainer et al., 1993), defined as the mean number of O3 molecules formed per NOx molecule consumed. Peak 8-h O3 ranged from w20e40 percent lower than average to w30e60 percent higher than average as NOz concentrations ranged from their lowest to highest values (Fig. 3). While reactions R8 and R10 each convert NO2 to a reaction product, R8 terminates radical production; in contrast, R10 is reversible because PAN decomposes as temperature increases. PAN therefore is a reservoir of NO2 for O3 production (Singh and Hanst, 1981; Singh, 1987; Pollack et al., 2013). The observed sensitivity of O3 to NOz potentially varies depending on the relative proportions of PAN and HNO3 (Pollack et al., 2013). Application of the GAM to 1995e2011 data from CTR, OAK, and YRK yielded sensitivities comparable to those for the shorter 2002e
197
2011 time period. This result supports earlier conclusions that regional O3 concentrations are NOx-sensitive (e.g. Chameides et al., 1988; Chameides and Cowling, 1995). The sensitivity coefficients for NO and NO2 did not vary if the model included or excluded afternoon NOz concentrations (Figs. S3 and S4). This result indicates that cumulative photochemical activity is a contributing factor separate from same-day mass loadings of NO and NO2. The sensitivity coefficients for NOz reflect both the amount of NOx that has been converted to NOz by afternoon and the OPE. Mean afternoon NOz is 27e39 percent of mean afternoon NOy at urban sites, 52 percent at OLF (suburban), and 58e68 percent at non-urban sites (Table 1). Therefore, the average reservoir of unreacted afternoon NOx ranged from 32 to 73 percent among sites. OPE has been estimated as the slope of the regression line of O3 versus NOz. Table 2 shows that OPE estimated in this way ranges from 4:1 at JST to 11:1 at CTR and OAK, with uncertainties ranging from 0.2 to 0.4 (one standard error). These values are consistent with previous studies (e.g., Trainer et al., 1993; Demerjian et al., 2013). In contrast, the GAM sensitivity curves are nonlinear, varying among sites and within different NOz concentrations ranges at each site. The JST, YRK, and CTR results in Fig. 3 are quasi-linear within the range of 5e10 ppbv NOz with O3/NOz slopes of 1.4, 3.1, and 1.8, respectively; within the range of 1e5 ppbv NOz, the JST, YRK, and CTR O3/NOz slopes are 3.5, 5.0, and 7.1, respectively. For afternoon NOz concentrations less than 1 ppbv, the YRK and CTR slopes increase to 9.2 and 12.7, respectively, while the JST slope reverses from positive to negative. The GAM was used to project O3 responses to future reductions of ambient NO, NO2, and NOz concentrations by replacing the 2009 through 2011 NO, NO2, and NOz concentrations with one-half, onethird, and 10 percent of their actual values. When ambient NO, NO2, and NOz concentrations were 10 percent of the observed 2009e 2011 values, the projected 50th and 95th percentiles of peak 8-h O3 concentrations ranged from 21 to 38 ppbv and from 32 to 50 ppbv, respectively (Fig. 4; see also S5). The projected 50th percentiles were 26e28 ppbv at three non-urban sites (CTR, OAK, OLF) and 35e 38 ppbv at three urban (BHM, JST, GFP) and one urban-influenced (YRK) site; the projected values for PNS (21 ppbv at 50th
Fig. 4. Distributions of predicted daily peak 8-h O3 changes with concentrations of NO, NO2, and NOz each reduced to 10 percent of the observed 2009e2011 concentrations.
198
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Fig. 5. GAM-predicted mean annual peak 8-h O3 change due to (top) NOz versus percent deviation of mean annual afternoon NOz from overall site mean NOz and (bottom) nonbiogenic NMOC versus mean annual non-biogenic NMOC concentration. The predicted O3 changes are expressed as percentage deviations relative to site mean peak 8-h O3.
percentile) were based on 2009 data only and may therefore be less reliable. Mean afternoon NOy concentrations would be w0.2 ppbv at CTR, OAK, and OLF, and w0.4e1 ppbv at BHM, JST, and YRK, when reduced to 10 percent of their 2009e2011 means (Table 1). Assuming that the GAM projections reliably represent the consequences of large (50e90 percent) precursor reductions, the 90 percent decreases of NO, NO2, and NOz suggest that peak 8-h O3 concentrations are unlikely to fall below limits within the range of w25e40 ppbv. 3.3. Sensitivity of peak O3 to NMOC Additional applications of the GAM yielded a significant relationship between non-biogenic NMOC and peak O3 at JST but not at YRK (Fig. S6). The results agree with reported sensitivity of O3 to both NMOC and NOy in Atlanta but NOx-sensitive O3 at non-urban YRK (Blanchard et al., 2010a). Projections based on replacing observed ambient NO, NO2, NOz, and non-biogenic NMOC concentrations by one-half, one-third, and 10 percent of their actual values yielded median peak 8-h O3 concentrations of 25 ppbv at YRK and 30 ppbv at JST (Fig. S7). For JST, the GAM results indicated that maximum O3 concentrations were weakly sensitive (decreasing) to isoprene concentrations (either as 24-h or morning
averages). For YRK, the sensitivity of maximum O3 concentrations to isoprene concentrations was weak or negligible (Fig. S6). These results are inconsistent with the expectation that isoprene should increase O3 maxima. The inconsistency may indicate an uncertainty in measuring isoprene reliably because of its known high reactivity. Alternatively, isoprene-O3 chemistry may be complicated by air mass movement and mixing, by the potential lack of NMOC response in NOx sensitive regimes, or by the responses of vegetation to oxidant exposure, temperature and soil moisture (e.g., Loreto et al., 2001; Sharkey et al., 2008). The apparent anomalies were studied further by examining average diurnal concentrations of isoprene and non-biogenic NMOC that were differentiated by O3, NOx, NOz concentrations and by meteorological properties. Evaluation of these results did not provide a simple explanation. 3.4. Temporal variations Both observed and predicted daily peak 8-h O3 concentrations varied among the days of the week (Table S2). Large (w10e70 percent) reductions of NO, NO2, and NOz concentrations occurred on Sundays compared with midweek (Table S3), but did not account for day-of-week variations in peak O3. The day of the week was a statistically significant predictor in the GAM, which suggests
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
that O3 formation depends on the ambient concentrations of species in addition to NO, NO2, and NOz. The GAM predicted interannual variations in mean annual peak 8-h O3 that matched the observed year-to-year variations at all sites (Fig. S8). Variations in weather were the largest contributor to interannual variations (Fig. S9). Trends in the unadjusted mean annual peak 8-h O3 concentrations were not statistically significant (at p ¼ 0.05) at 7 of the 8 sites (average rate of decrease was 0.37 ppbv per year) (Table S4). When adjusted for weather, trends were statistically significant at 5 of the 8 sites and the average decrease among all sites was 0.90 ppbv per year. Camalier et al. (2007) interpreted trends in the meteorologically-adjusted peak O3 as evidence for effects of emission reductions. When adjusted for meteorological effects and for changing ambient concentrations of NO, NO2, and NOz, the remaining O3 trends were statistically significant at only CTR and the multi-site average decrease was 0.25 ppbv per year. This result indicates that the changing concentrations of NO, NO2, and NOz explained the majority of the observed downward O3 trend; the residual trend is possibly due to factors not included in the model. When normalized as a percent-to-percent comparison, a consistent relationship between O3 and NOz concentrations emerges (Fig. 5). The sites exhibit an average 0.2 percent change in mean annual peak 8-h O3 (0.13e0.18 at urban sites, 0.21 at OLF, and 0.22e0.27 at rural sites) for each percent NOz change. The mean annual afternoon concentrations of NO, NO2, and NOz declined at all sites and mean afternoon NOz as a percentage of mean afternoon NOy showed no trend (Fig. S10). Therefore, the principal determinant of the average afternoon NOz concentration was the amount of NOx emitted and present in the atmosphere. However, the modeled NOz effect on peak O3 results from changes in both NOx mass loading and OPE, and the latter is linked to NMOC concentrations through rates of radical production and removal. At the urban sites, afternoon NOz is typically w30 percent of NOy; in contrast, NOz at the rural sites is w65 percent (Table 1). If urban conditions represent O3 photochemistry of about one day, the urban-rural differences suggest a regional aging or carryover of two days or more. For JST, the interannual variations of non-biogenic NMOC concentrations yielded variations of up to w3 percent from the mean peak 8-h O3, compared with the NOz effect of w10 percent (Fig. 5). 4. Conclusions The monitoring record from the eight sites of the SEARCH network offers a rich data set for investigating the observed response of O3 to emission changes in the southeastern U.S. Application of a generalized additive model (GAM) to a ten-year record of O3, NO, NO2, NOz, and meteorological measurements indicates the importance of temperature, humidity, and NOx to O3 maxima in the southeastern U.S. The GAM estimates of O3 sensitivity to NO, NO2, and NOz are qualitatively consistent with expectations from photochemistry, and are potentially important for evaluation of PAQSM predictions. The relationships of O3 to surface wind speed and direction, and to 24-h back trajectory distance and direction, suggest that regional and local O3 components are dominated by stagnation on a w400 km scale, and not by transport over greater distances such as has been observed in the northeastern U.S. When NO, NO2, and NOz concentrations in the model were reduced to 10 percent of their observed 2009e2011 values, the projected 50th percentiles (26e38 ppbv) and 95th percentiles (36e50 ppbv) of peak 8-h O3 provide evidence for an irreducible O3 component. In Atlanta, O3 was less sensitive to NMOC than to NOx; at a rural location outside Atlanta, no evidence was found for sensitivity of O3 to NMOC.
199
Acknowledgments This study was sponsored by the Southern Company. The authors thank Eric Edgerton and John Jansen for reviewing preliminary drafts of this manuscript. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.atmosenv.2014.02.006. References Atmospheric Research and Analysis, 2013. http://www.atmospheric-research.com/ studies/SEARCH/index.html (accessed July, 2013). Blanchard, C.L., Hidy, G.M., Tanenbaum, S., 2010a. NMOC, ozone, and organic aerosol in the southeastern states, 1999e2007: 2. Ozone trends and sensitivity to NMOC emissions in Atlanta, Georgia. Atmospheric Environment 44, 4840e 4849. http://dx.doi.org/10.1016/j.atmosenv.2010.07.030. Blanchard, C.L., Hidy, G.M., Tanenbaum, S., Rasmussen, R., Watkins, R., Edgerton, E., 2010b. NMOC, ozone, and organic aerosol in the southeastern states, 1999e 2007: 1. Spatial and temporal variations of NMOC concentrations and composition in Atlanta, Georgia. Atmospheric Environment 44, 4827e4839. http:// dx.doi.org/10.1016/j.atmosenv.2010.08.036. Blanchard, C.L., Hidy, G.M., Tanenbaum, S., Edgerton, E.S., Hartsell, B.E., 2013a. The Southeastern Aerosol Research and Characterization (SEARCH) Study: temporal trends in gas and PM concentrations and composition, 1999-2010. Journal of the Air & Waste Management Association 63, 247e259. http://dx.doi.org/10.1080/ 10962247.2012.748523. Blanchard, C.L., Hidy, G.M., Tanenbaum, S., Edgerton, E.S., Hartsell, B.E., 2013b. The Southeastern Aerosol Research and Characterization (SEARCH) study: spatial variations and chemical climatology, 1999e2010. Journal of the Air & Waste Management Association 63, 260e275. http://dx.doi.org/10.1080/ 10962247.2012.749816. Camalier, L., Cox, W., Dolwick, P., 2007. The effects of meteorology and their use in assessing ozone trends. Atmospheric Environment 41, 7127e7137. Chameides, W., Lindsay, R., Richardson, J., Kiang, C., 1988. The role of biogenic hydrocarbons in urban photochemical smog: Atlanta as a case study. Science 241, 1473. Chameides, W.L., Cowling, E.B., 1995. The State of the Southern Oxidants Study: Policy-relevant Findings in Ozone Pollution Research, 1988e1994. Southern Oxidants Study, College of Forest Resources, North Carolina State University, Raleigh, NC. Demerjian, K., Schwab, J.J., Roychowdhury, U., Sun, Y., Zhang, Q., 2013. Research Measurements of Chemical Constituents and Related Processes Affecting New York State’s Regional and Urban O3 and PM Air Quality. Prepared for New York State Energy Research and Development Authority, Albany, New York. Contract number 10602. Draxler, R., Stunder, B., Rolph, G., Stein, A., Taylor, A., 2009. Hysplit4 User’s Guide. National Oceanic and Atmospheric Administration. http://www.arl.noaa.gov/ documents/reports/hysplit_user_guide.pdf (accessed April 20, 2009). Edgerton, E.S., Hartsell, B.E., Saylor, R.D., Jansen, J.J., Hansen, D.A., Hidy, G.M., 2005. The Southeastern Aerosol Research and Characterization Study: part 2 e filterbased measurements of PM2.5 and PMcoarse mass and composition. Journal of the Air & Waste Management Association 55, 1527e1542. Edgerton, E.S., Hartsell, B.E., Saylor, R.D., Jansen, J.J., Hansen, D.A., Hidy, G.M., 2006. The Southeastern Aerosol Research and Characterization Study, part 3: continuous measurements of fine particulate matter mass and composition. Journal of the Air & Waste Management Association 56, 1325e1341. Edgerton, E.S., Saylor, R.D., Hartsell, B.E., Jansen, J.J., Hansen, D.A., 2007. Ammonia and ammonium measurements from the southeastern United States. Atmospheric Environment 41, 3339e3351. Edgerton, E.S., Casuccio, G.S., Saylor, R.D., Lersch, T.L., Hartsell, B.E., Jansen, J.J., Hansen, D.A., 2009. Measurements of OC and EC in coarse particulate matter in the southeastern United States. Journal of the Air & Waste Management Association 59, 78e90. Gego, E., Porter, P.S., Gilliland, A., Rao, S.T., 2007. Observation-based assessment of the impact of nitrogen oxides emissions reductions on ozone air quality over the eastern United States. Journal of Applied Meteorology and Climatology 46, 994e1008. http://dx.doi.org/10.1175/JAM2523.1. Gego, E., Gilliland, A., Godowitch, J., Rao, S.T., Porter, P.S., Hogrefe, C., 2008. Modeling analyses of the effects of changes in nitrogen oxides emissions from the electric power sector on ozone levels in the eastern United States. Journal of the Air & Waste Management Association 58, 580e588. Gilliland, A., Hogrefe, C., Pinder, R.W., Godowitch, J., Foley, K.L., Rao, S.T., 2008. Dynamic evaluation of regional air quality models: assessing changes in O3 stemming from changes in emissions and meteorology. Atmospheric Environment 42, 5110e5123. http://dx.doi.org/10.1016/j.atmosenv.2008.02.018. Godowitch, J., Gilliland, A., Draxler, R.R., Rao, S.T., 2008. Modeling assessment of point source NOx emission reductions on ozone levels in the eastern United States. Atmospheric Environment 42, 87e100.
200
C.L. Blanchard et al. / Atmospheric Environment 88 (2014) 192e200
Hansen, D.A., Edgerton, E.S., Hartsell, B.E., Jansen, J.J., Hidy, G.M., Kandasamy, K., Blanchard, C.L., 2003. The Southeastern Aerosol Research and Characterization Study (SEARCH): 1. Overview. Journal of the Air & Waste Management Association 53, 1460e1471. Hansen, D.A., Edgerton, E., Hartsell, B., Jansen, J., Burge, H., Koutrakis, P., Rogers, C., Suh, H., Chow, J., Zielinska, B., McMurry, P., Mulholland, J., Russell, A., Rasmussen, R., 2006. Air quality measurements for the aerosol research and inhalation epidemiology study. Journal of the Air & Waste Management Association 56, 1445e1458. Hastie, T.J., Tibshirani, R.J., 1990. Generalized Additive Models. Chapman and Hall/ CRC, Boca Raton, p. 335. Hidy, G.M., 2000. Ozone process insights from field experiments e part 1: overview. Atmospheric Environment 34, 2001e2022. Hidy, G.M., Brook, J.R., Demerjian, K.L., Molina, T. Luisa, Pennell, W.T., Scheffe, R.D. (Eds.), 2011. Technical Challenges of Multipollutant Air Quality Management. Springer, New York. Liu, S.C., Trainer, M., Fehsenfeld, F.C., Parrish, D.D., Williams, E.J., Fahey, D.W., Hubler, G., Murphy, P.C., 1987. Ozone production in the rural troposphere and the implications for regional and global ozone distributions. Journal of Geophysical Research 92 (D4), 4191e4207. Loreto, F., Mannozzi, M., Maris, C., Nascetti, P., Ferranti, F., Pasqualini, S., 2001. Ozone quenching properties of isoprene and its antioxidant role in leaves. Plant Physiology 126, 993e1000. National Research Council, 1991. Rethinking the Ozone Problem in Urban and Regional Air Pollution. National Academy Press, Washington, D.C, p. 489. NARSTO, 2000. An Assessment of Tropospheric Ozone Pollution. Report 1000040. NARSTO, Pasco, WA (Also available from: EPRI, Palo Alto, CA). Pollack, I., Ryerson, T., Trainer, M., Neuman, J., Roberts, J., Parrish, D., 2013. Trends in ozone its precursors and related secondary oxidation products in Los Angeles, California: a synthesis of measurements from 1960e2010. Journal of Geophysical Research 118, 5891e5911. Saylor, R., Edgerton, E.S., Hartsell, B.E., Baumann, K., Hansen, D.A., 2010. Continuous gaseous and total ammonia measurements from the southeastern aerosol research and characterization (SEARCH) study. Atmospheric Environment 44, 4994e5004. Seinfeld, J.H., 1986. Atmospheric Chemistry and Physics of Air Pollution. John Wiley and Sons, New York. Sharkey, T., Wiberley, A., Donohue, A., 2008. Isoprene emissions from plants: why and how? Annals of Botany 10, 5e18.
Sillman, S., Logan, J.A., Wofsy, S.C., 1990. The sensitivity of ozone to nitrogen oxides and hydrocarbons in regional ozone episodes. Journal of Geophysical Research 95 (D2), 1837e1851. Singh, H.B., Hanst, P.L., 1981. Peroxyacetyl nitrate (PAN) in the unpolluted atmosphere: an important reservoir for nitrogen oxides. Geophysical Research Letters 8, 941e944. Singh, H., 1987. Reactive nitrogen in the troposphere: chemistry and transport of NOx and PAN. ES&T 21, 320e327. The R Project, 2009. The R Project for Statistical Computing. http://www.R-project. org (accessed January 25, 2013). Trainer, M., Parish, D.D., Buhr, M.P., Norton, R.B., Fehsenfeld, F.C., Anlauf, K.G., Bottenheim, J.W., Tang, Y.Z., Wiebe, H.A., Roberts, J.M., Tanner, R.L., Newman, L., Bowersox, V.C., Meagher, J.F., Olszyna, K.J., Rodgers, M.O., Wang, T., Berresheim, H., Demerjian, K.L., Roychowdhury, U.K., 1993. Correlation of ozone with NOy in photochemically aged air. Journal of Geophysical Research 98 (D2), 2917e2925. U.S. Environmental Protection Agency, 2007. Guidance on the Use of Models and Other Analyses for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze. EPA-454/B-07-002. U.S. Environmental Protection Agency, 2008. Details of the Omnibus Meteorological Database (OMD or “MetDat”). http://www.epa.gov/scram001/meteorology/ omnibus_meteorological_data_set.pdf (accessed December 2008). U.S. Environmental Protection Agency, 2009. R Code to Model Daily Maximum 8hour Ozone as a Function of Daily Meteorological Variables Plus Terms to Account for Calendar Day, Dow-of-week and a Term “Year” to Account for Trends over Time. http://www.epa.gov/air/airtrends/weather/metadj_code.txt (accessed January 25, 2013). U.S. Environmental Protection Agency, 2011a. Clean Air Interstate Rule (CAIR). http://www.epa.gov/airmarkets/progsregs/cair/index.html (accessed April 16, 2011). U.S. Environmental Protection Agency, 2011b. SO2 Reductions and Allowance Trading under the Acid Rain Program. http://www.epa.gov/airmarkets/ progsregs/arp/s02.html (accessed April 16, 2011). U.S. Environmental Protection Agency, 2012. The NOx SIP Call. http://www.epa.gov/ airmarkets/progsregs/nox/noxsipcall.html (accessed March 14, 2012). Wolff, G.T., Lioy, P.J., 1980. Development of an ozone river associated with synoptic scale episodes in the eastern United States. Environmental Science & Technology 14, 1257e1260.