Supplement of Chemical climatology of the southeastern United States , 1999 – 2013

Abstract. A series of experiments (the Southern Oxidant and Aerosol Study – SOAS) took place in central Alabama in June–July, 2013 as part of the broader Southern Atmosphere Study (SAS). These projects were aimed at studying oxidant photochemistry and formation and impacts of aerosols at a detailed process level in a location where high biogenic organic vapor emissions interact with anthropogenic emissions, and the atmospheric chemistry occurs in a subtropical climate in North America. The majority of the ground-based experiments were located at the Southeastern Aerosol Research and Characterization (SEARCH) Centreville (CTR) site near Brent, Alabama, where extensive, unique aerometric measurements of trace gases and particles and meteorology were made beginning in the early 1990s through 2013. The SEARCH network data permits a characterization of the temporal and spatial context of the SOAS findings. Our earlier analyses of emissions and air quality trends are extended through 2013 to provide a perspective for continued decline in ambient concentrations, and the implications of these changes to regional sulfur oxide, nitrogen–ozone, and carbon chemistry. The narrative supports the SAS program in terms of long-term average chemistry (chemical climatology) and short-term comparisons of early summer average spatial variability across the southeastern US at high temporal (hourly) resolution. The long-term measurements show that the SOAS experiments took place during the second wettest and coolest year in the 2000–2013 period, with lower than average solar radiation. The pollution levels at CTR and other SEARCH sites were the lowest since full measurements began in 1999. Changes in anthropogenic gas and particle emissions between 1999 and 2013 account for the decline in pollutant concentrations at the monitoring sites in the region. The data provide an opportunity to contrast SOAS results with temporally and spatially variable conditions in support of the development of tests for the robustness of SOAS findings.


Introduction
An important element of atmospheric science is the reconciliation of the results of short-term field campaigns with long-term measurements.Such comparisons are used to estimate the extent to which the results from special studies apply to other times, different locations, or under different aerometric conditions.Short-term campaigns for studying atmospheric processes are amenable to detailed observations of complex physicochemical processes.While it is impractical to obtain such detail over many years, long-term measurements can provide basic indicators of key atmospheric and chemical processes.The Southern Oxidant and Aerosol Study (SOAS) campaign was part of the Southern Atmosphere Study (SAS) held over the period of 1 June-15 July 2013 (EPA, 2014a;UCAR, 2014).SAS encompassed several major intensive field campaigns, of which SOAS was one.The campaigns involved a number of research groups in order to inform a series of coordinated research questions, which included the following: (1) the amounts, variations,

G. M. Hidy e al.: Chemical climatology of the southeastern United States
and controlling processes for biogenic fluxes of organic carbon and nitrogen, (2) aerosol composition, secondary aerosol formation processes, associated driving factors, and impacts, (3) aqueous-phase and cloud chemistry, (4) climate relevant properties of biogenic aerosol, (5) oxidation processes and oxidant concentrations, (6) cycling of reactive nitrogen compounds, and (7) mercury emissions, transformation, and fate.
A key recurring theme of all the SAS research was to investigate how interactions of biogenic and anthropogenic emissions affect nitrogen, mercury, oxidant, and secondary organic aerosols (SOA) at the surface and aloft (e.g., Carlton et al., 2013;Hunt, 2013;NOAA, 2013).
This paper provides a context for, and an assessment of, the representativeness of the SOAS experimental period using long-term surface data, excluding mercury (mercury data are available for some sites in Atmospheric Research and Analysis (ARA) 2014), collected at the observational site at Centreville (Brent), Alabama (Supplement Fig. S1), and at nearby sites of the Southeastern Aerosol Research and Characterization (SEARCH) network (ARA, 2014;EPRI, 2013).Our approach extends through 2013 earlier analyses of trends (Blanchard et al., 2013a, b, c) ending in 2010-2011.In addition to extending the earlier relationships for annual emissions with ambient concentrations of gases and particles, we add statistically significant linear relationships between ambient concentrations and emissions for 1999-2013.Also included are several new analyses and trends estimates for the SOAS early summer period, including aspects of the changes in concentrations of primary pollutants compared with secondary species.We first consider the climate of the southeastern United States in which the rural Centreville site (CTR) is located.This description is followed by a summary of pollutant emission trends in the region and a survey of the role of meteorology in mediating the southeastern air chemical climatology.The post-1999 trends in concentrations of key natural and pollutant species at CTR and the region are presented, with consideration of contrasts between rural and urban conditions, and between inland and coastal conditions.Finally, we discuss the long-term context and representativeness of the SOAS-SAS experiments from 1 June to 15 July 2013.The narrative is supported with additional tables and figures in the Supplement.

The infrastructure and climate of the Southeast
The southeastern US is characterized by a warm, moist, subtropical climate, with relatively isolated urban communities surrounded by agricultural areas and forests in hilly terrain.Anthropogenic activities in the region are known to emit volatile organic compounds (VOCs), CO, SO x , NO x , NH 3 , and particles, including organic carbon (OC) and black (elemental) carbon (EC) (see Appendix Table A1 glossary; e.g., SAMI, 2002;Ingram et al., 2013).The urban areas and the larger southeastern region have experienced a growth in population, transportation, and industry, resulting in increased  air pollution since the 1960s, followed by post-1990s reductions in pollution (e.g., Irving, 1991;Ingram et al., 2013).Climatologically, summers are warm and humid, with occasional thunderstorms associated with continental frontal systems passing through the United States.Winter conditions are cooler, with occasional freezing and isolated snow or ice storms inland, but generally milder conditions along the northern coast of the Gulf of Mexico.The region also is known for its extreme weather conditions where drought occurs, tornadoes develop in large convective systems in summer and early fall, and tropical storms from the Atlantic Ocean and the Gulf of Mexico penetrate inland on occasion (e.g., Peterson et al., 2014).Extended summer periods of light winds and air mass stagnation are prevalent, leading to accumulation of pollution (e.g., McNider et al., 1998).The region is exposed to smoke from wild and prescribed fires, as well as intrusions of transported dust and other pollution from North America and Africa (e.g., Hidy, 2009).
The SEARCH sites include continuous or semi-continuous measurements of indicators of pollutant emissions, reactants in the atmosphere, and products of chemistry  Hansen et al., 2003Hansen et al., , 2006;;Edgerton et al., 2005Edgerton et al., , 2006)).In addition to the long-term measurements, SEARCH has hosted a number of studies supporting methods development and urban epidemiology.
The data have been used for understanding ambient composition and chemistry, and a variety of analyses including characterization of trends (publications using SEARCH data for such purposes are found at http://www.atmosphericresearch. com/studies/SEARCH.index.html).SEARCH data also have been used in the evaluation of air quality models, including Zhang et al. (2004), Morris et al. (2005), Marmur et al. (2006Marmur et al. ( , 2009)), Pun and Seigneur (2008), Pun et al. (2008), and Tesche et al. (2008).The SEARCH sites were established in urban-rural (or urban-suburban) pairs to represent urban-rural contrasts, and to capture the potential influences of inland-continental vs. marine aerometric conditions.A number of other monitoring sites for O 3 , NO y , and particulate matter are located in the region, including those of IMPROVE, CSN, ASACA, and CASTNET (Table A1), and those data have been compared with SEARCH network data (e.g.Blanchard et al., 2013a).
CTR is believed to be a regionally representative location in an agricultural area surrounded by mixed deciduousevergreen forest including the Talladega National Forest a few kilometers to the south and west.Blanchard et al. (2013a) have discussed the representativeness of air chemistry at CTR. Support for this conclusion is also included later in this paper based on regional aerometric data and its relative uniformity in azimuthal direction of concentration variation.The region, represented by Alabama, Georgia, Mississippi, and northwestern Florida, is of particular interest in that relatively large quantities of natural biogenic emissions occur as VOCs and as detritus from vegetation.CTR is approximately 80 km SSW of the Birmingham, Alabama metropolitan area.The site is known to be exposed occasionally to pollution from Birmingham, Tuscaloosa (∼ 45 km WNW), and Montgomery (∼ 120 km SE), along with major vehicle traffic routes and certain large point sources in the region, including power plants, pulp and paper mills, and metal processing (e.g., Birmingham) indicated, for example, in Supplement Table S2.For regional gas and particle emissions, see Supplement Fig. S2 and the discussion in Blanchard et al., 2013c).The Southeast also experiences a substantial amount of biomass burning (Wade et al., 2000;Haines et al., 2001).Overall, the CTR aerometric conditions represent relatively low gas and aerosol concentrations typical of the rural Southeast from Louisiana eastward to Georgia, and from Tennessee and the Carolinas southward to the Gulf of Mexico (Blanchard et al., 2013a).In the Southeast, essentially all rural areas experience exposure to sustained diluted air pollution (e.g., Goldan et al., 1995Goldan et al., , 2000;;SAMI, 2002).This regime of low, annual-average regional concentrations of monitored pollutants tends to approach nominal background concentrations characteristic of the eastern North American continent, e.g., estimated for different gases and particles, for example, by Seinfeld and Pandis (1998) for NO x , Hidy and Blanchard (2005) for PM 2.5 , and Lefohn et al. (2014) for O 3 .The presence of anthropogenic pollution in the Southeast combined with large amounts of natural biogenic emissions yields the potential for a regional air chemistry that has elevated ozone levels, partially driven by isoprene from vegetation and anthropogenic NMOC (e.g., Chameides et al., 1988;Chameides and Cowling, 1995;Meagher et al., 1998;SOS, 2014).Various species are associated with the regional photochemistry, including sulfur oxides, reactive nitrogen, VOC and the "secondary" pollutants, inorganic SO 4 , (particle) pNO 3 , O 3 , and secondary organic compounds.
The Southern Oxidant Study (SOS), beginning in 1988, is of particular interest for elucidating ozone chemistry in the Southeast.This study involved several years of O 3 and precursor monitoring and major campaigns in Atlanta and later in Nashville, Tennessee (SOS, 2014).The SOS provided evidence of the regional and urban character of summer O 3 concentrations, affected by precursor emissions distributions, active photochemistry, and meteorological influences.The accumulation of O 3 in the region depends on an interplay of urban plumes, elevated regional concentrations, and point-source plumes.SOS confirmed the hypothesis of Chameides et al. (1988), showing the importance of isoprene and the (regional) NO x sensitivity of O 3 production.Observation-based O 3 modeling matured from the SOS experience, and the nature of O 3 production efficiency using the O 3 / NO z (or O 3 / NO 2 ) ratio was proposed (e.g., Solomon et al., 2000).The role of southeastern meteorology, including multiple-scale transport winds horizontally dispersing pollution and vertical mixing, was observed.Also of interest were the observations showing that winds repeatedly arising at night in the boundary layer potentially result in transport of O 3 .The SOS experience served as the basis for characterizing the long-term trends in chemical climatology of oxidants in the Southeast, and in part catalyzed interest in the aerosol component of southeastern pollution.As discussed below, the aerometric conditions for O 3 accumulation currently remain the same in the Southeast, but both urban and regional concentrations of O 3 and its precursors have declined since the mid-1990s.Recent information about O 3 and precursor chemistry is derived from the SEARCH studies (e.g., Blanchard et al., 2010a, b;2011).
Scientists recognized by the 1960s that secondary organic carbon (SOC) was present in atmospheric aerosols both naturally and from anthropogenic VOC oxidation (e.g., Went, 1960;O'Brien et al., 1975;Grosjean and Friedlander, 1980).Regional and urban SOC in the Southeast was expected given the presence of isoprene and terpenes as precursors, as well as anthropogenic > C 7 VOCs.Lim and Turpin (2002), Zheng et al. (2002Zheng et al. ( , 2006)), Saylor et al. (2006), Ding et al. (2008), and Blanchard et al. (2008, 2013a) reported estimates of SOC at SEARCH sites using different methods, all of which suggest that secondary organic carbon, SOC (vs.primary organic carbon, POC) is the smaller fraction of annual-average OC in the SEARCH region.
Theories linking oxidant-O 3 chemistry with reactions forming SOC are summarized in Seinfeld and Pandis (1998), based on hypothesized mechanisms mainly for > C 7 hydrocarbons (especially C = C containing compounds), and smog chamber experiments.In the 1990s, advanced laboratory experiments showed that lower carbon-number VOCs, including isoprene, could produce aerosols.Hallquist et al. (2009) reviewed the emerging homogeneous and heterogeneous chemical mechanisms and caveats relating to volatility or emitted and aging aerosols.A subset of mechanisms and observations have linked aqueous chemical reactions (e.g., Carlton and Turpin, 2013) and organosulfates and other tracers with SOC production (e.g., Takahama et al., 2006;Gao et al., 2006;Surratt et al., 2007;Froyd et al., 2010;Kleindienst et al., 2010;Budisulistiorini et al., 2013).Elucidation of these mechanisms is an important element of the goals of the SOAS experiments.

Gas and particle emissions
The principal resource for post-2000 emissions and their trends in the SEARCH region is EPA's National Emissions Inventory (NEI).The NEI is formally revised and updated every 3 years, but its estimates have been extrapolated on an annual basis (e.g.Xing et al., 2012;Blanchard et al., 2013c).Revision and extrapolation since the 1990s have involved intermittent changes in emission modeling, revised interpretation of emission testing, and the addition of major fugitive sources such as open burning, NH 3 , and biogenic NMOC.These methodological changes are incompletely documented, which makes interpretation of differences difficult to follow (e.g., Xing et al., 2012;Blanchard et al., 2013c).In any case, the NEI provides a starting point for interpretation of changes in ambient air chemistry in response to annual emission changes.
Annual trends in the region are given in the Supplement Fig. S2.Annual emissions from the states of Alabama, Georgia, Mississippi, and the 12 western counties of Florida are adopted for comparison with ambient data (Supplement Fig. S2).Analysis of the NEI data indicates that emissions of essentially all anthropogenic species of interest have decreased or remained nearly constant since 2000.Between 1999 and 2013, SO 2 emissions decreased by 75 % (Fig. S2a), NO x emissions decreased by ∼ 57 % (Fig. S2b), and anthropogenic VOC emissions declined by ∼ 32 % (Supplement Fig. S2c).Whereas NH 3 emissions have remained roughly constant during this period (13 % decrease -Supplement Fig. S2d), CO emissions decreased by 39 % (Supplement Fig. S2e).Fine particle emissions decreased by 7 % (Supplement Fig. S2f), and primary OC and EC emissions decreased by 0.3 % and 36 %, respectively (Fig. S2g and S2h in the Supplement).

Role of meteorology
Once present in the atmosphere, constituents are dispersed by meteorological processes and react chemically with one another, with varying temperature, humidity, and oxidant levels.It is widely known that meteorological factors have a strong influence on ambient concentrations of chemicals, including those produced in the atmosphere.This knowledge is imbedded in source-based air quality models that estimate spatial and temporal concentration distributions from sources (e.g., Seinfeld and Pandis, 1998;McMurry et al., 2004).While emissions of precursors are important, meteorology controls a major fraction of O 3 and SO 4 variability in multi-spatial studies from the 1970s through the 1990s (e.g., Rao et al., 1997;Hidy, 1994).Analysis of measurements shows relationships with meteorology independent of modeling.Relevant examples for the study of meteorological influences on secondary species include sulfate (SO 4 ) and O 3 (e.g., Hidy, 1994;Chameides and Cowling 1995;Meagher et al., 1998;Hidy, 2000;Blanchard et al., 2010bBlanchard et al., , 2014a)).Hidy (1994) summarized the key 1978-1980 findings of multiple spatial-and temporal-scale meteorology in the eastern US that affect SO 4 and O 3 concentrations.Rao et al. (1997) discuss the importance of synoptic-scale processes in affecting ground-level concentrations at temporal and spatial scales of 1-2 days and ∼ 600 km in the Southeast.The SOS investigators describe meteorology and the multiscale O 3 climatology in the Southeast; Hidy (2000) summarized the role of meteorology in modulating O 3 chemistry, and Blanchard et al. (2010bBlanchard et al. ( , 2013a, b;, b;2014a) described pollutant species in the light of meteorological parameters in the SEARCH region.
The case of SO 4 on a regional scale in the eastern US, including parts of Tennessee, Virginia and North Carolina, was explored extensively in the 1980s, as reviewed by Hidy (1994).Using principal component analysis, Henry (e.g.Hidy, 1994, Chapter 5) showed that meteorological factors, primarily light winds, temperature, and humidity, accounted for much of the SO 4 variability in the eastern US during 1977-1978.As a part of SOS, Vukovich (1994Vukovich ( , 1998) ) found that meteorological factors, including temperature, humidity and winds, accounted for much of the variability in O 3 concentrations in the Southeast based on data from 1980 to 1991.SOS experiments in the Atlanta, Georgia (1991)(1992) and Nashville, Tennessee areas (1994)(1995)1999) confirmed the strong influence of meteorological factors, especially temperature, on daily O 3 maxima (e.g., Banta et al., 2002;Frost et al., 1998;White et al., 2002).
A recent analysis of summer rural and urban mean O 3 concentrations using the Camalier et al. (2007) generalized additive model (GAM) aimed to quantify statistically the impact of meteorology on O 3 formation in recent years (Blanchard et al., 2010b).The results indicated that meteorological factors accounted for ∼ 70 % of the variability in daily peak 8 h O 3 concentrations over the 1999-2007 time frame.Application of this method to 2002-2011 data from multiple sites confirmed this influence (Blanchard et al., 2013b(Blanchard et al., , 2014a)).Temperature and relative humidity (RH) were the strongest factors influencing high O 3 concentrations (solar radiation correlated with RH and was a factor when RH was excluded).Other meteorological factors potentially important to O 3 concentrations include light winds or, alternatively, transport distances < 200 km, signifying air mass stagnation with slow moving synoptic weather systems over the Southeast.

Ambient concentration trends
A conventional chemical climatology uses certain "macro" indicators for representing ambient chemistry at the ground (SO 2 , NO y , NMOC, CO, O 3 , and particle mass and component concentration).The following discussion of these chemical indicators extends beyond 1999-2010 to 2013 the recent downward trends in ambient concentration trends analyses of Blanchard et al. (2013b).

Sulfur oxides
Historically, the evolution of atmospheric conditions in the Southeast involved major urban expansion superimposed on a base of agriculture and silviculture.This evolution resulted in a sulfur dioxide (SO 2 )-rich atmosphere inland from the Gulf of Mexico through the early 1990s (e.g., Irving, 1991;Hidy, 1994).After this time, SO 2 emissions and ambient concentrations declined dramatically, both regionally and in cities such as Birmingham and Atlanta.A regional decline after 2000 is shown for SO 2 emissions and for annual-average ambient SO 2 and SO 4 concentrations in Fig. 2 (see also Figs.S2a and S3 in the Supplement).Regional ambient SO 4 concentrations follow the emissions throughout this period, and are consistent with emissions and ambient concentration changes over a longer period of time documented in other studies (e.g., pre-1990s, Irving, 1991post-1990post- , Xing et al., 2012)).SO 2 emissions from electric generating units (EGUs) represented 76 % of total SO 2 emissions, on average, between 1999 and 2013.SO 4 tracks the SO 2 emission reductions (Supplement Fig. S2a).Statistically significant linear relations between annual SO 2 emissions and ambient SO 2 and SO 4 concentrations are indicated in Table 1.The SO 2 intercept is not statistically significant, while the statistically significant SO 4 intercept is ∼ 1 µg m −3 .We identify the SO 4 intercept with a continental background similar to that found to the west at this latitude (e.g.Hidy and Blanchard, 2005).
Seasonal averages of SO 4 are included in Fig. S3.An example of seasonally averaged changes for SO 4 is shown in Fig. 3 relative to changes in organic and elemental carbon (OC and EC).The concentrations of SO 4 in CTR reach maxima in summer and minima in winter over a 15-year period, with a decline in seasonal amplitude after 2007.This is consistent with the known sulfur oxidation chemistry involving temperature, oxidant production, and elevated moisture content in summer.The seasonal amplitude in SO 4 is greater than that of the carbon components, which correlate during The presence of organic sulfur compounds is hypothetically a factor in SOC formation as postulated, for example, by Surratt et al. (2007).An upper estimate can be made of the mass of sulfur bound in such compounds through differencing methods.It is known from historic studies (e.g., Novakov, 1972;Hoffman et al., 1980) that particle S can include species other than SO 4 as acid, ammonium, or other salts.The fraction of non-(inorganic) sulfate (NIS) component can be estimated by comparing sulfur measured by X-ray fluorescence (XRF) and the water-soluble component (WSSO4) measured by ion chromatography.This difference could reflect NIS compounds such as sulfite salts, methane sulfonate, or other organosulfur compounds, including isoprene-related sulfate esters.A comparison of the Atlanta (JST) daily av-erage values by season for the period 1999-2000 is shown in Fig. 4. The comparison between XRF and WSSO4 in this urban area suggests small differences between the measures, in the range of −6 to +4 % of total sulfur as NIS, depending on season.The months with the highest apparent contribution were May through September.If the differencing is an indicator of NIS, its S fraction in particles is small relative to inorganic sulfate, even in summer.Given the uncertainty in sampling and measurement expected from a differencing method (see also Supplement Table S3), the presence of NIS is not definitive for detecting small mass fractions of NIS in particles; however, the mass of organic sulfur particles would be a larger fraction of OC than of sulfate.
Recent studies of SOC formation from biogenic NMOC suggest that condensed material is facilitated with acid nuclei (e.g., Hallquist et al., 2009).Since SO 4 is a major contributor to particle acidity, the degree of SO 4 neutralization can be used as a qualitative surrogate for apparent acidity.The acidity in particles at rural Southeast locations, including CTR, and its relation to OC, was investigated several years ago.Tanner et al. (2009) found no unambiguous evidence of acid-catalyzed SOC formation from intercomparison of filter-based SO 4 and OC under varying conditions.However, the Lin et al. (2013) plume study of chemistry in high vs. low SO 2 and NH 3 conditions indicated a small enhancement in isoprene epoxydiol (IEPOX)-related tracers in high SO 2 conditions.Analysis of the SEARCH particle data, including NH 4 , SO 4 , and pNO 3 , indicates that, stoichiometrically, sulfate can be interpreted, on average, as a mixture that lies between (NH 4 ) 2 SO 4 and NH 4 HSO 4 (e.g., letovicite), both in urban and rural locations.Even though the average sulfate neutralization is equivalent to letovicite, there are excursions in daily values of neutralization ranging from ammonium sulfate and bisulfate to (rarely) sulfuric acid.Analysis of the data by year from 2000 to 2013 (Fig. 5) suggests that acidic SO 4 neutralization has not changed appreciably over this period.Since neutralization has not changed, but sulfate concentrations have declined, the hydrogen ion concentrations associated with SO 4 would have decreased.This trend suggests that acid-promoted reactions of biogenic NMOC could be present but have potentially changed with SO 4 concentrations unless other nucleophiles act as reactants (as indicated in experiments of Nguyen et al., 2013).

Nitrogen oxides, NMOC, and ozone
The nitrogen oxides and VOC are important elements of the oxidant photochemistry cycle in the troposphere and are components involved in aerosol particle chemistry.Because of its importance, O 3 and its precursors have been studied extensively in the Southeast in major programs, including the SOS campaign in the Atlanta, Georgia and Nashville, Tennessee areas in the 1990s (e.g.Meagher et al., 1998).
The urban-rural contrasts for O 3 production and atmospheric transport were investigated under the O 3 precursor conditions at that time.The SEARCH urban-rural site pairs are analogous to the 1994-1995 Nashville study, covering similar regional conditions inland, but distinctly different along the Gulf Coast.Thus, the Nashville study, combined with longer-term SOS measurements, yields insight into oxidant chemistry for 1990s conditions, with higher O 3 and precursor concentrations than those seen since 2000 in the SEARCH region.NO and NO y have been measured at SEARCH sites since 1992; NO 2 has been measured at some SEARCH sites since 1992 and at all sites since 2002.NMOC was measured at JST through 2008 and is supplemented by the PAMS observations, including from the non-urban site at YRK.The sources of nitrogen oxides are mainly combustion from transportation, EGUs, and to a lesser extent, industrial, commercial, and residential sources.Anthropogenic VOC emissions, including NMOC, deriving from transportation, industrial, and commercial sources, are superimposed as a small fraction on a much larger regional emissions component of natural sources (see Supplement Fig. S2c).In Table 1, linear regression models indicate statistically significant responses of ambient NO y and NMOC concentrations to decreasing NO x and (anthropogenic) VOC emissions, respectively, without statistically significant intercepts.The SEARCH regional annual NO x emissions are shown along with rural annual-average NO y concentrations in Fig. 6.Reductions in ambient NO y levels tracked the NO x emission reductions between 2000 and 2013.The trends in CO emissions for transportation (Supplement Fig. S2e) have a similar pattern to NO x emissions (Fig. S2b in the Supplement).Ambient CO concentrations have decreased at both urban and rural sites in the SEARCH region, with statistically significant linear relationships to CO emissions (Table 1).Blanchard et al (2013b) showed that CO concentrations have an extra-regional "offset" of about 120 ppbv apparently associated with a regional or larger continental background.Such a "background" was not found for SO 2 or NO y .This is supported by the linear model for ambient CO and emissions in Table 1, where the CTR relationship has an intercept of ∼ 100 ppbv; the Birmingham data (BHM) show a strong local total CO emissions relationship without a statistically significant intercept.
Emissions of VOC are indicated in Supplement Fig. S2c; the anthropogenic component is superimposed on a large natural component such that the decline in anthropogenic VOC emissions is a small increment of the total.Decreases in rural ambient NMOC are not well documented, but the urbannon-urban contrast is established with the PAMS and Atlanta (JST) data.The annual averages of 24 h average NMOC at JST are shown with Georgia motor vehicle emissions in Fig. 7.The decline in urban NMOC is similar to the on-road VOC emissions reduction -as expected from a principally urban source; motor vehicles, NMOC trends, and OH reactivity are discussed, for example, in Blanchard et al. (2010a).Evaluation of trends in NMOC at JST indicates that mean concentrations of anthropogenic species decreased between 1999 and 2007; however, the average OH reactivity of the NMOCs decreased less, if at all.One of the important contributors to the JST OH reactivity was isoprene, and in summer at the non-urban Yorkville and Conyers sites, where it contributed half of the estimated reactivity (Blanchard et al., 2010a).The dominant NMOC in the oxidant cycle at CTR is isoprene.
In the absence of long-term NMOC data at CTR, we rely on the PAMS data in the Atlanta area for potential qualitative indications of ambient NMOC characteristics or trends.The contrasting speciated NMOC data in summer from the PAMS sites in Atlanta and the site at YRK indicate the magnitudes of the decreases in species concentrations with distance from the urban area.The YRK concentrations are expected to be similar to those at CTR, with CTR possibly exhibiting still lower anthropogenic species concentrations relative to a nominally similar level of natural NMOC.Trends in the summer PAMS data (Fig. 8) are considered for selected species having measurements that appear robust (few values below detection limits; no obvious extreme outliers).These species include the anthropogenic groups of C 2 -C 4 compounds, aromatics such as benzene, toluene, ethyl benzene, and xylenes (BTEX), pentanes, and isoprene (representing biogenic species).The concentrations of BTEX have decreased over the years at the Atlanta urban sites but not at the non-urban sites (YRK and Conyers).Mean urban pentane concentrations began to decrease in 2004, followed by a 3-year increase in 2007.The concentrations of the anthropogenic groups are substantially lower at YRK, about 60 km WNW "upwind" from Atlanta, while isoprene tends to be larger in concentration away from the city and increasing slightly by year.The reason for this seasonal increase in isoprene concentration has not been explained but could be related to changes in biogenic emissions and chemistry in response to decreasing O 3 concentrations (e.g., Sharkey et al., 2008) since they do not appear to be consistent with changes in T and RH (e.g., Supplement Fig. S13 for a 6-week period in early summer).
Ozone (with NO 2 and NO z ) can be used as an indicator of the oxidation state of the air.Average 8 h maximum O 3 concentrations in the SEARCH region have declined along with decreasing emissions of its precursors, NO x and NMOC (e.g., Fig. 9 and Fig. S2 in the Supplement).Linear regression results for decreasing O 3 concentrations as a function of decreasing NO x and (anthropogenic) VOC emissions are shown in Table 1.The O 3 decrease is less than a 1 : 1 proportionality with the precursor emission reductions, as indicated by the presence of intercept terms that are a large fraction of ambient O 3 concentrations.The intercepts of ∼ 29-37 ppbv are similar to background levels projected for the Southeast by Lefohn et al. (2014).The O 3 proportionality with its precursors would not necessarily be expected due to the nonlinear photochemistry involved, but given the substantial reduction in precursor emissions, a relatively small change in ozone of about 1 % yr −1 has been observed.
Rural NO z (NO y -NO x , a photochemical reaction product indicator) concentrations have tended to level out since 2005, as indicated in Fig. 6, paralleling O 3 concentrations.The linear NO z relationship with NO x emissions is statistically significant (Table 1) with a variance higher than for other gases, and without a statistically significant intercept.Ozone concentrations vary with meteorological conditions, which need to be accounted for in long-term averages.Camalier et al. ( 2007) have developed a statistical approach (GAM) employing meteorological variables to adjust mean summer O 3 concentrations locally and regionally, which has been adopted by EPA for inspecting O 3 trends (e.g.EPA, 2013).We have extended the GAM method to use air quality variables as well to improve the model for O 3 concentrations in the SEARCH region (Blanchard et al., 2014a).Evidence from JST compared with YRK indicates that Atlanta has frequent periods of NMOC-sensitive O 3 formation, but the regional O 3 formation continues to be NO x -sensitive as found earlier in the SOS studies, including the Atlanta and Nashville areas (e.g., Blanchard et al., 2010b;Chameides et al., 1988;Chameides and Cowling, 1995) (Meagher et al., and NMOC-sensitive).Evidence for O 3 as an oxidant indicator derives from comparing diurnal O 3 concentrations with OH and HO 2 data.For example, at a suburban site 8 km NE of downtown Nashville, OH and HO 2 observations tracked the diurnal buildup and decline of O 3 (Martinez et al., 2003).This linkage also has been reported from observations in New York City (e.g.Ren et al., 2003).From the data taken in the New York City area, OH concentrations empirically are roughly proportional to O 3 concentrations during the day (e.g., Supplement Fig. S4).If this urban-suburban pattern applies to rural conditions in the Southeast, one could expect to see a decline in daytime OH concentrations between 2000 and 2013.This assumption is tempered by recent evidence for the regeneration of OH from isoprene reactions yielding an excess of OH from chemical model calculations (e.g.Fuchs et al., 2013).The morning rise in O 3 concentrations at the ground also corroborates the potential importance of air mass transport with down-mixing of O 3 from a residual (upper boundary) layer during the breakup of the nocturnal inversion layer 2-3 h after sunrise (Baumann et al., 2000).As a part of the analysis of the SOAS data, a close look at relationships between OH, HO 2 , O 3 and VOCs would be of interest not only for SOC formation, but also for distinguishing rural from urban O 3 photochemistry.

Aerosol particles
Suspended particles and their origins are a key element of the SOAS experiments.The SO 4 component was discussed above with SO 2 .Trends in regional and CTR particle mass concentration and composition are shown in Supplement Fig. S5.Mass concentrations of fine particles (PM 2.5 ) have decreased with primary emissions and secondary components (SO 4 and carbon).Sulfate has decreased in fine particles relative to EC and OC; NH 4 has not decreased as much; and pNO 3 decreased, even though increases in pNO 3 concentrations potentially could have occurred as declining SO 4 made an increasing fraction of NH 3 available for reaction with HNO 3 (Blanchard et al., 2007).The major metal oxides (MMO) represent a small fraction of fine particle mass.Retained water (associated with species such as SO 4 , and represented as one of the unmeasured components within "other") is relatively large.As shown above in Fig. 3, SO 4 , EC, and OC display somewhat different seasonal variations.Regional OC is identified with primary sources such as vehicle transportation and biomass burning (e.g., Lee et al., 2010;Zhang et al., 2010).Biomass burning is believed to be a significant portion of PM 2.5 concentrations especially in the rural Southeast.Various methods have been used to separate secondary (SOC) from primary (POC) (e.g., OC / EC ratio method, Lim andTurpin, 2002, Saylor et al., 2006; organic tracer-chemical mass balance, Zheng et al., 2002;Ding et al., 2008; regression, carbon mass balance, Blanchard et al., 2008;Blanchard Yu et al., 2007).While these methods provide insight into the two components of OC, they do not yield a consistent quantitative picture of the fractions of each.The mean fraction of SOC estimated by different methods varies from ∼ 20 % to > 50 % depending on the time of year data were obtained, and the method used.
The trends in rural annual average EC concentrations compared with EC emissions are shown in Fig. 10.At the rural SEARCH sites, ambient EC decreases over the decade beginning in 2000, and tracks the EC emission decline occurring between 2005 and 2013.EC concentrations declined approximately coinciding with declining mobile source emissions (Fig. S6 in the Supplement).Mean OC concentrations followed similar trends as mean EC at both urban and rural SEARCH sites on an annual basis (Supplement Fig. S7).These and other results indicate that POC is a major part of average OC concentrations (e.g., Blanchard et al., 2008;Zheng et al., 2002;Blanchard et al., 2013cBlanchard et al., , 2014b)).
Linear regression models for relating decreases in rural EC and OC concentrations to decreasing emissions are shown in Table 1.Included are relationships between ambient concentrations with total EC and OC emissions and with mobilesource EC and OC emissions.The slopes for total and mobile EC are significant, but the intercepts are not.The slope and intercept for ambient OC versus mobile source OC emissions are each significant, but neither the slope nor the intercept for regressions of rural ambient OC versus total OC emissions is significant.Mobile OC comprises less than 10 % of the total POC emissions in the region (Blanchard et al., 2013c), but SOC derived from mobile VOC emissions could augment the mobile POC contributions; OC derived from a combination of motor vehicle emissions of POC and of low-volatility VOC potentially represents a large enough fraction for the changes in mobile-source emissions to be detectable at CTR.
Birmingham EC and OC regression models are included in Table 1 for comparison with the rural case.This urban case is dominated by motor vehicle and industrial carbon emissions.The models for EC and OC vs. annual regional total carbon emissions and motor vehicle emissions show a much stronger annual rate of decline than the rural cases do, without statistically significant intercepts.Regression models (not shown) for EC and OC in Atlanta (JST) are similar to those found for BHM.The linkages between motor vehicle OC emissions and ambient OC are more evident at BHM and JST than at CTR because the higher motor vehicle emission density within the urban areas results in a stronger signal that is more readily observed within the variability associated with meteorology and with year-to-year variations in biomass burning on small spatial scales (e.g., ∼ 10-100 km).
From source apportionment, wildfires and prescribed burns are a major factor in OC and EC concentrations in the SEARCH region, including at CTR (e.g.Blanchard et al., 2013c).The post-2005 NEIs indicate that the major contributors to EC and POC derive from vehicle transportation emissions and biomass burning (Supplement Fig. S2g and h; Blanchard et al., 2013c).For the period from 2002 to 2013, annual wildfires and prescribed burning are shown in Fig. S8.There is substantial variability in fires reported by year and between states (e.g., Larkin et al., 2013).The prescribed burns in each of the states in the SEARCH regime are much more frequent than the wildfires over this period.EPA (2014b) has tabulated major seasonal differences by state in the occurrence of fires for use in the preparation of the NEI.According to the EPA, the Georgia fire season is in late winter (January-March) while the Alabama season is apparently later, mainly in May-June; the information on which these temporal profiles were based is not documented.The implications of the temporal differences are unclear for prescribed burn EC and POC contributions relative to annual or seasonal ambient averages.
The SEARCH data have also given insight into the urbanrural contrasts in ambient EC and OC.For example, carbon isotope analysis has indicated that most of the total carbon -TC (OC + EC) at CTR is modern carbon compared with nearby Birmingham, which has a major fraction of fossil carbon (e.g., Blanchard et al., 2008).The modern carbon fraction can be derived from primary emissions of smoke from fires, cooking, or from detritus and/or secondary atmospheric reactions of VOC.The secondary component of modern carbon comes from a host of potential biogenic reactions.The urban-rural contrasts between BHM and CTR indicate a large fraction of SO 4 and OC has a regional character, with SO 4 having a small increment and OC a larger increment associated with the urban emissions (Blanchard et al., 2013a).Urban and rural contrasts are readily seen in Table 2, illustrating differences for the urban and rural pairs, and between inland and coastal sites.Strong urban-rural differences are found in the inland pair, CTR-BHM, for particles and gases.Interestingly, mean O 3 concentrations are higher at CTR than BHM.The urban-rural contrast is weaker for the coastal pair, OAK and Gulfport, Mississippi (GLF).The secondary species tend to have weaker urban-rural gradients than the primary species.The data suggest the regional character for SO 4 (also NH 4 ) and O 3 , while other species have lower mean concentrations along the coast than inland.The conditions along the Gulf Coast show lower concentrations, consistent with smaller cities, fewer major point sources along the coast than further inland, and mesoscale meteorological effects associated with marine conditions.Seasonal similarities and differences are shown in Table S4.f The SEARCH database adopts the convention OM / OC = 1.4.

Aerosol light extinction
Ground-level estimates of light extinction have been made at the SEARCH sites since 1999.Extinction apportionment gives additional insight into the role of chemical components including retained water.Extinction has tended to decrease over the 2000-2013 period, as expected from the decline in fine particle concentrations.A time series of observations for b ext at CTR is shown in Supplement Fig. S9, including the dry light scattering component and the absorption component.The bulk of the extinction is from light scattering.
Figure 11 shows a trend in combined fine and coarse particle (PM 10 ) contributions to light extinction coefficient using a weighting relationship with particle composition, in-cluding Rayleigh scattering and also water through a hygroscopicity relationship with RH (Pitchford et al., 2007).On a dry basis, major contributors to "dry" extinction coefficient include SO 4 and organic matter (OM as 1.8 OC) at the SEARCH sites, including CTR.Water content, shown separately in Fig. 11, based on hygroscopicity of SO 4 and NO 3 , is a major component of extinction at the average RH seen annually in CTR.Hygroscopicity of OM is not included; the inorganic component accounts for most of the water retention.This result is consistent with findings from other studies, including those in the Shenandoah, Virginia region of the Southeast (e.g., Malm et al., 2000), in the Great Smoky Mountains (Lowenthal et al., 2009), and also with laboratory studies (e.g., Brooks, et al., 2004).The black carbon component (light-absorbing carbon or LAC) represents a relatively minor portion of the extinction coefficient annually, while the contribution of OM is larger than ammonium sulfate.For comparison, the contributions to the extinction coefficient based on chemical constituents without separation of water from SO 4 and NO 3 is given in Supplement Fig. S10.

Representativeness of 2013
The previous section provides a long-term context of air quality changes in the SEARCH domain.For the representativeness of the 2013 summer experiments, we turn to the long-term ambient data associated with the period of June 1 to July 15.The discussion includes consideration of the representativeness of meteorological conditions as well as the ambient chemistry at CTR.

Diurnal emission characteristics
The annual emission trends derived from the NEIs provide no information about short-term emission patterns.Two major contributors to regional, diurnal emissions relevant to the SOAS experiments are SO 2 and NO x from EGUs, and NO x and NMOC from transportation.Natural emissions for isoprene and terpenes are also of interest.The June-July diurnal EGU SO 2 and NO x emissions for Georgia, Alabama, and Mississippi are shown in Supplement Fig. S11.A national average transportation emission pattern applied to NO x and NMOCs is shown in Supplement Fig. S12 for comparison with the EGU pattern.The EGU emissions of SO 2 and NO x show a "common" increase in daytime emissions, rising at about 06:00 LST through the daytime followed by a decrease in evening after 18:00 LST (with stated exceptions, all data are given in local standard time).The annual trends in the reduction in emissions across the states are consistent with the annual changes for SO 2 and NO x in Fig. S2.The average diurnal changes in the on-road motor vehicle emissions (NO x and NMOC) show an increase around 06:00 LST with morning urban commutes, with steady levels thereafter with a rise in traffic, followed by a decrease around 17:00 LST.As indicated in the figure, there is a weekday-weekend difference in emission timing.Natural VOC emissions of isoprene are influenced by rising and declining temperature and solar radiation, with a rise at ∼ 06:00 LST in summer, modulated by O 3 concentrations (e.g.Sharkey et al., 2008).Terpene emissions are mainly temperature dependent and tend to rise at 06:00 LST, before decreasing somewhat during midday and increasing again in the evening (Supplement Fig. S19; e.g., Guenther et al., 1993).NO x emissions from soil emanations vary with temperature and soil moisture, with a less-known average diurnal character; lightning-derived NO x production is obviously intermittent depending on stormy conditions.

Meteorological conditions
Surface meteorological conditions are described in terms of box plots of the CTR data, directional pollution roses, and diurnal plots of parameter variation.Box plots of wind speed, temperature, relative humidity, and solar radiation are shown in Supplement Fig. S13 for 1 June to 15 July 2013.The period generally is characterized by light winds (air mass stagnation), warm daytime temperatures, high relative humidity, and reduced solar radiation relative to other years.A comparison of the box plots by year indicates that the period in 2013 was moist and cool relative to previous years, with slightly less solar radiation compared with earlier years.Meteorological conditions favorable to regional pollution episodes around CTR in current years are analogous to those found in the SOS Nashville experiment (McNider et al., 2002;Banta et al., 2002) and identified for the SEARCH region (Blanchard et al., 2013a).
Conditions for elevated pollution involve large-scale, multi-day periods of light winds, high solar radiation (SR), and high RH, with limited nocturnal vertical mixing in a temperature inversion in the lower boundary layer.In the morning, the inversion tends to break up, permitting increased vertical mixing with pollution aloft reaching the ground.This is followed in the evening by re-formation of the inversion layer suppressing mixing.Occasional winds aloft, in the boundary layer, develop around the high pressure regime creating conditions for dispersion of pollution across the region, accompanied by decreases in ground-level concentrations and concentrations aloft.The diurnal changes in summer EGU and transportation emissions, as well as natural VOC emissions tend to parallel changes associated with vertical mixing and changes in inversion conditions, providing increased daytime reactants.The morning increases tend to "fuel" the photochemistry of oxidants and SOC formation even with daytime dilution, while reducing nocturnal chemical reaction rates.

Chemical characterization
Box plots of trace gas and particle composition in Supplement Figs.S14 and S15 indicate the downward trend in pollutants at CTR in June-July, consistent with the trends in annual averages and emissions in the region discussed above.In some years, the range of concentrations is large relative to the mean, but in recent years the range, based on the 25th and 75th percentiles, appears to have decreased along with the mean values.
Pollution-wind direction roses for the 1 June-15 July 2013 period are indicated in Fig. 13.Wind speed and frequency are also shown, indicating that northerly winds were rare in 2013; instead westerly winds occurred more frequently with higher wind speeds than usual.The pollution roses suggest that SO 2 has some directionality that varies from year to year during the early summer, with a distinct source influencing the site from the southwest during this period.NO y , CO, and EC show less directionality but broad enhancements in northern sectors, pointing to urban (traffic) influence from Birmingham and Tuscaloosa.O 3 , PM 2.5 , SO 4 , NH 4 , and OC show more regional character, indicated by near-symmetrical concentration distributions with direction.The OC / EC ratio has a distinct directional relationship.Although elevated EC derives mainly from the BHM and Tuscaloosa (northern) directions, OC / EC ratios show larger values from the west and south in association with low EC, while OC / EC ratios are lower from the north and east.These differences suggest that urban sources influence CTR from the northern and eastern sectors, but non-urban sources (including biomass burning and secondary material) are strongest from the southwest.In contrast to SO 2 concentrations, which peak in the morning and evening at CTR, sulfate shows a weak enhancement between midday and early evening (Fig. 15), with declining overall concentrations since 2000.The reactive species concentrations display a character consistent with recent lower hourly reaction rates associated with lower SO 4 and OH concentrations compared with earlier years.The diurnal EC and OC concentrations over the years have settled into low levels slightly enhanced around 09:00 CST -corresponding to a time similar to NO y , thus possibly associated with the same sources such as a local traffic increase.The weak diurnal OC concentration pattern does not indicate a strong daylight driven component with rapid rise in oxidant concentrations, but does not rule out the potential for SOC formation at night, as well as during the day.Periods prior to 2013 show an apparent spiky character of EC concentrations, possibly identified with local yard waste or trash burning, and less associate with the previously discussed regional biomass burning (e.g.prescribed fires for forest management).Prescribed burning is a ubiquitous, widespread source for OC and EC with OC / EC emission ratios of 14 or larger (Lee et al., 2005) in the forests of the Southeast, including Alabama.In 2013, the Alabama Forestry Commission issued 14 137 burn permits for over 1 million acres, the largest total burn area since 2010 (B.Elliott, personal communication, 2014).Prescribed burning is regulated differently by state, making the NEI estimates of emissions for this source highly uncertain.Satellite observations are used to locate burns, but they do not reliably identify small burn areas below dense forest canopies.
The importance of aerosol water content in SOC reaction chemistry is hypothesized in Shen et al. (2013) and Carlton and Turpin (2013).A sample diurnal pattern of liquid water suggested in the difference between dry particle light scattering and ambient particle scattering is seen in Fig. S17.A maximum water uptake is at night with the highest RH, while the aerosol tends to lose water systematically with equilibration to lower midday RH.While ample water is present in the aerosol particles during the day, they are much wetter on average at night.The retained water in particles provides an opportunity for both gas and aqueous-phase oxidation reactions  of NMOC to take place in the CTR environment.OC does not increase with drying out of the aerosol at morning to midday, which would indicate release of SOC from nocturnal reaction products.
Diurnal variations in NMOC concentrations are important for characterizing a portion of the potential for SOC production at CTR.While NMOC data were not obtained at CTR as part of the basic SEARCH observations, a long-term record of summer concentrations found at YRK can be compared qualitatively with results taken at a rural sites in the Southeast (Supplement Table S5) at locations relevant to SOS, including CTR (e.g., SOS studies of Goldan et al., 1995;Hagerman et al., 1997;Riemer et al., 1998).The data cited in Table S5 indicate the importance of isoprene as a reactive species in the southeastern rural environment.The data suggest a varied but persistent regional presence of anthropogenic NMOC.In each of the locations, anthropogenic NMOC is mixed with the natural NMOC.These data suggest that the production of SOC in the rural sites may be dominated by natural NMOC, but the formation chemistry is complicated by low concentrations of anthropogenic species.
Figure S18 in the Supplement shows the diurnal changes at YRK in several NMOC groups: C 2 -C 4 , BTEX, pentanes, and isoprene.The first three groups show a concentration peak in the morning followed by a decrease through midday, with "recovery" of concentrations during the night.As noted earlier, this pattern is interpreted to be associated with the accumulation of emissions in a shrinking nocturnal inversion combined with the morning emissions increase and inversion breakup, followed by vertical mixing and chemical loss during the day.The four groups vary in terms of reactivity, but show similar morning to midday change; the change is likely to be mainly dilution during the day.BTEX and pentanes concentration have decreased over the years shown, while C 2 -C 4 hydrocarbons seem less consistent in annual change.Isoprene has a distinctly different diurnal pattern, with a modest morning increase in concentration followed by depletion, then a strong peak in the late afternoon/evening.Hypothetically, annual diurnal change in isoprene concentrations appears to reflect variations in June-July meteorological influence combined with oxidation chemistry rather than a systematic trend over the years.A comparison of the diurnal change in isoprene concentration with OC diurnal change (Fig. 15) suggests no direct or parallel link between midday OC associated with SOC production.Data taken at a rural mixed deciduous and evergreen forest site in western Alabama (Kinterbish State Park) could be similar to CTR for biogenic NMOC concentrations.A comparison between isoprene concentrations at YRK and 1990 concentrations at Kinterbish (Goldan et al., 1995) is shown in Supplement Fig. S19 along with Kinterbish α-pinene concentrations.The Kinterbish pattern is similar at night and morning through midday, but displays a weaker afternoon/evening peak compared with YRK.The α-pinene pattern shows a morning peak, followed by a decrease from oxidation and dilution, but is much lower in concentration than isoprene.Interpretation of these diurnal patterns could give perspective to the understanding of oxidation chemistry and SOC formation.

Summary
Inspection of long-term data in the SEARCH region and specifically at CTR indicates that the experiments of summer 2013 generally represent aerometric conditions within the range of those in the last decade, although some parameters were lower relative to previous years.Meteorologically, the early summer 2013 data at CTR were biased towards a range of conditions that was marginally wetter and cooler, with lower solar radiation, than in most previous years.Chemically, air quality at CTR in 2013 exhibited a mix of natural and anthropogenic influences, but the magnitude of the anthropogenic component was significantly lower than during the earlier years of the last decade or before.The decline in ambient concentrations of gases and particles is not anomalous since emission reductions after 1999, extending beyond 2010-2011 through 2013, reflect regulatory actions intended to lower emissions permanently.The implications of observed changes are uncertain as to whether the ambient response to emission reductions results in slowing reaction rates due to reduced reactant concentrations, or in changes in the relative importance of the hundreds of photochemical reactions involved.This aspect of the change in the southeastern air chemistry needs to be investigated in detail, and SOAS will assist in achieving this understanding.The current conditions are likely to foretell the future conditions, which perhaps will increasingly return to a quasi-natural state as anthropogenic emissions continue to decline.
The 1 June-15 July period of 2013 displayed a range of pollution concentrations ranging from clean air to conditions of substantial intrusion of pollution over the region with evident plumes from urban or point sources.Tracking trends of trace gases indicates important reductions in ambient concentrations since 2000, resulting from regional reductions of CO, SO 2 , NO x , and NMOC emissions as indicated by statistically significant regressions of ambient concentrations against emissions.The emission reductions also reduced extreme ground-level O 3 concentrations, but not at the rate seen in precursor reductions.Fine particle concentrations have declined since 2000, with declines in concentrations of SO 4 and both EC and OC.There is enough NH 3 in the air to effectively neutralize the average acidity of rural SO 4 and pNO 3 , but the summer tends to contain less NH 4 neutralized SO 4 than winter.Average acidic SO 4 has remained approximately the same since 2000; stoichiometrically, SO 4 is present as an ammonium salt, and is approximately neutralized as letovicite or bisulfate in summer.Water-soluble (inorganic) sulfur as sulfate appears to account for almost all (> ∼ 95 %) the total S in particles.Gas and particle concentrations tend to be lower along the Gulf Coast compared with inland.There are also urban-rural differences in air quality, with stronger gradients associated with primary pollutants in contrast to secondary species, although weaker secondary species gradients do exist.
Evidence from the long-term record and source apportionment suggests that much of the OC is modern in origin at CTR, but OC in total derives in part from primary sources, including biomass burning and vehicle transportation.This result needs to be accounted for in apportionment of the SOC fraction.Evidence from Zheng et al. (2002) and Blanchard et al. (2008), for example, suggests that the majority of the regional annual or seasonal average OC can be accounted for from primary sources.
A major focus of the SOAS experiments was to evaluate the mechanisms for SOC enhancement through interaction of biogenic emissions with anthropogenic pollution.While it is clear that various SOC formation mechanisms are active (as measured with the chemical tracers), the magnitude of the OC enhancement and when it occurs are in question.SEARCH data coupled with trends in anthropogenic emissions provide the ability to test various hypotheses over an extended time period.The detailed chemical species measurements from the SOAS experiments, combined with the constraints of the long-term SEARCH observations, provide an important framework to add a perspective on how, when and where SOC occurs in the southeastern US.The Supplement related to this article is available online at doi:10.5194/acp-14-11893-2014-supplement.

Figure 2 .
Figure 2. Mean annual SO 4 concentrations (all SEARCH sites) and SO 2 concentrations (rural SEARCH sites -CTR, OAK, YRK) compared with emissions of SO 2 from EGUs located in Alabama, Georgia, northwest Florida, and Mississippi vs. year.Uncertainties are 1 standard error of the inter-site means.EGU emissions are updated from Blanchard et al. (2013c).

Figure 3 .
Figure 3. Mean seasonal concentrations of SO 4 , OC, and EC at CTR. Seasons are by calendar month (1) December-February, (2) March-May, (3) June-August, (4) September-November.Vertical divisions are calendar year between Q4 and Q1.OC and EC tend to track one another and the trends are downward at approximately the same annual rate, while SO4 shows a sharper decrease until 2008.OC tends to have winter minima, but OC maxima vary between Q2 and Q4.

Figure 4 .
Figure 4. Comparison of the US federal reference method (FRM) SO 4 , FRM X-ray fluorescence S (multiplied by 3 to represent SO 4 ), and ARA-SEARCH sampler (PCM1) SO 4 .The data are from JST, 1999 and 2000 (19-54 days per monthly mean).The PCM1 sampler is configured with a denuder to remove SO 2 and thereby reduce or eliminate positive artifacts, whereas the FRM sampler is not.FRM SO 4 measurements at other sites and during other years were too limited for comparison (n < 4 values per month except at BHM, where n = 8-11 values during 6 months).XRF S measurements are made on FRM filters.

Figure 6 .
Figure 6.Mean rural NO y and NO z concentrations at the three rural SEARCH sites (CTR, OAK, YRK) and emissions of NO x from all emission sources located in Alabama, Georgia, northwest Florida, and Mississippi vs. year.Uncertainties are 1 standard error of the inter-site means.Emissions are derived from the NEIs and are updated from Blanchard et al. (2013c).

Figure 7 .
Figure 7. Mean annual NMOC concentrations at JST and emissions of VOC from on-road mobile exhaust within Georgia vs. year.Uncertainties are 1 standard error of the inter-site means.The 1998 NMOC measurements were made from August through December.Emissions are updated from the NEIs, and from Blanchard et al. (2013c).

Figure 8 .
Figure 8. Mean summer (June-August) concentrations of NMOC species at four PAMS sites within the Atlanta metropolitan area.South DeKalb and Tucker are urban sites; YRK is considered an "upwind" non-urban site, while Conyers is a suburban (semi-rural) "downwind" site.(Note: 2012 only July data are available; 2013 only YRK July data are available).

Figure 9 .
Figure 9. Mean peak 8 h O 3 at rural SEARCH sites during the months of March through October compared with anthropogenic NO x and VOC emissions in Alabama, Georgia, Mississippi, and northwest Florida.The 2012-2013 NO x emissions were estimated beginning in 2011 using the 2011 and 2012-2013 CEMS data for EGU emissions.

Figure 10 .
Figure 10.Mean annual EC concentrations at rural SEARCH sites compared with emissions of EC from all emission sources in Alabama, Georgia, Mississippi, and northwest Florida.Emissions data are from extension using the method of Blanchard et al. (2013c).

Figure 11 .
Figure 11.Mean annual light extinction computed from daily species concentrations at CTR calculated using the new IMPROVE algorithm (a).Mean light extinction computed from daily species concentrations at CTR during 1 June-15 July of each year (b).The aerosol components include ammonium salts, OM (= 1.8OC), lightabsorbing carbon (LAC), soil dust, coarse material (CM), and water.The black line shows the measured RH averaged over different periods represented in the two panels of the figure.RH dependence was computed using the new IMPROVE f(RH) formula with hourly RH measurements at each site.Effects of water in particles, associated with SO 4 and NO 3 , are shown separately as the bH 2 O term (See also Fig. S10).

Figure 13 .
Figure 13.Comparison between average 1 June-15 July, 2000-2012 and the SOAS period of 1 June-15 July 2013 for directional wind speed frequency and direction, trace gas concentrations, and concentrations of PM 2.5 mass, EC, OC, SO 4 , NH 4 , and OC / EC at CTR.

Figure 14 .
Figure 14.Mean diurnal concentrations of trace gases SO 2 , NO y , NO z , and O 3 at CTR from 1 June through 15 July of each year.
Geographical locations of the SEARCH monitoring sites and the nominal region of influence for the site measurements.PNS, OAK, and GFP were retired after 2009, 2010, 2012, respectively.The grey areas in the figure represent the counties corresponding to the urban metropolitan areas of Birmingham, Atlanta and Pensacola, Florida.

Table 1 .
Linear regression trends models relating 1996-2013 annual average ambient concentrations (ppbv for CO, NO y , NO z , NMOC, and O 3 ; µg m −3 for SO 2 , SO 4 , EC, OC) to annual, regional a chemical emissions (MMT yr −1 ).Statistically significant (p < 0.05) results are indicated by bold-face p values b .
a Total annual emissions from Alabama, Georgia, Mississippi, and northwest Florida covering 1996-2013 (including emissions from biomass burning but excluding emissions from vegetation and soils).b The p value is a measure of statistical significance, defined as the probability of obtaining a nonzero result as large (small) or larger (smaller) than observed under the null hypothesis (i.e., when the true parameter value is zero).c Georgia on-road mobile source VOC emissions.d Mobile-source emissions.

Table 2 .
Mean concentrations by inland and coastal sites determined for1999 -2002  and 2009 -2012.PM 2.5 .PM 2.5and PM 10−−2.5 mass and species concentrations are means of daily measurements (24 h filter-based samples).O 3 measurements are means of peak daily 8 h maxima.Concentrations of gases are means of days determined from hourly data.Means were determined by year, and the yearly means were then averaged.
a Gas measurements at BHM began in 2000.bPM10−−2.5 measurements atGFP and OAK started in 2003.cMeasurements after 2010 are not available.d NO 2 measurements took place over 1999-2002 e NH 3 measurements commenced in 2004 as 24 h filter-based measurements; hourly measurements began in 2011.

Table A1 .
Glossary.refractory,insoluble,chemically inert, light-absorbing carbon, operationally defined by measurement protocol (SEARCH filter samples utilize CSN/IMPROVE thermal optical reflectance method) IEPOX isoprene epoxydiol, a product of isoprene oxidation LAC light-absorbing carbon -operationally defined concentration using light absorbance as an estimate of black (as elemental) carbon NMOC non-methane organic carbon, a part of VOCs OC organic carbon made up of organic species directly from sources (primary organic carbon -POC) and secondary organic carbon (SOC) from chemical reactions of volatile organic species in the atmosphere, including photochemically induced oxidation products OM (particle) organic matter -defined as x × OC, where "x" is an empirical constant ranging from∼1.4 to ∼ 2.3.SEARCH data are reported with x = 1.4 OVOC oxygenated volatile organic compounds, including alcohols, aldehydes, ketones, and acids TC total carbon normally without the carbonate carbon component VOC volatile organic compounds Other terms EGU electric generating unit -power plant, frequently coal fired in the Southeast GAM generalized (statistical) additive model MMO major metal oxides (defined in SEARCH data as the sum of Al 2 O 3 , SiO 2 , K 2 O, CaO, TiO 2 , and Fe 2 O 3 ) NEI (USEPA) National Emissions Inventory NIS non-inorganic sulfate; e.g., sulfate esters, methane sulfonate, sulfite salts pNO 3 particulate nitrate, often assumed to be mainly NH 4 NO 3