Open Access

This study characterizes the spatial and temporal patterns of aerosol and precipitation composition at six sites across the United States Southwest between 1995 and 2010. Precipitation accumulation occurs mostly during the wintertime (December-February) and during the monsoon season (July-September). Rain and snow pH levels are usually between 5-6, with crustal-derived species playing a major role in acid neutralization. These species (Ca2+, Mg2+, K+, Na+) exhibit their highest concentrations between March and June in both PM2.5 and precipitation due mostly to dust. Crustal-derived species concentrations in precipitation exhibit positive relationships with [Formula: see text], [Formula: see text], and Cl-, suggesting that acidic gases likely react with and partition to either crustal particles or hydrometeors enriched with crustal constituents. Concentrations of particulate [Formula: see text] show a statistically significant correlation with rain [Formula: see text] unlike snow [Formula: see text], which may be related to some combination of the vertical distribution of [Formula: see text] (and precursors) and the varying degree to which [Formula: see text]-enriched particles act as cloud condensation nuclei versus ice nuclei in the region. The coarse : fine aerosol mass ratio was correlated with crustal species concentrations in snow unlike rain, suggestive of a preferential role of coarse particles (mainly dust) as ice nuclei in the region. Precipitation [Formula: see text] : [Formula: see text] ratios exhibit the following features with potential explanations discussed: (i) they are higher in precipitation as compared to PM2.5; (ii) they exhibit the opposite annual cycle compared to particulate [Formula: see text] : [Formula: see text] ratios; and (iii) they are higher in snow relative to rain during the wintertime. Long-term trend analysis for the monsoon season shows that the [Formula: see text] : [Formula: see text] ratio in rain increased at the majority of sites due mostly to air pollution regulations of [Formula: see text] precursors.


Introduction
The southwestern United States is experiencing rapid population growth, land-use change, drought, and variability in precipitation and water availability (Woodhouse et al., 2010;Cayan et al., 2010;Seager and Vecchi, 2010;Harpold et al., 2012), which both affect and are affected by the region's aerosol particles and precipitation.Ongoing changes in the Southwest's climate are reducing the relative contributions of winter snow versus summer rain to the annual water balance (Cayan et al., 2010) and shortening the duration of snow cover and melt (Harpold et al., 2012).Although chemical relationships between particulate matter and precipitation have been studied in a wide range of environments, few locations exhibit as wide a range of sensitivity to atmospheric chemistry as the Southwest.For example, dust deposition in seasonal snowpacks increases melt rate during spring in the mountains of Colorado (Painter et al., 2007).The amount of fine and coarse aerosol particles may also alter the amount and spatial distribution of potential rain or snow via their role as cloud condensation nuclei (CCN) and ice nuclei (IN), respectively (e.g Rosenfeld and Givati, 2006).In both desert and montane ecosystems, the deposition of nitrate and sulfate have been shown to be acidifying agents for aquatic ecosystems resources (e.g.Fenn et al., 2003), while excess nitrogen in precipitation has altered plant-soil nutrient relations and induced directional biological shifts in ecosystems (Fenn et al., 1998;Baron et al., 2000;Wolfe et al., 2003;Neff et al., 2008).Consequently, the composition and acidity of wet deposition in the Southwest have critical effects on terrestrial and aquatic ecosystems.
Published by Copernicus Publications on behalf of the European Geosciences Union.

A. Sorooshian et al.: Aerosol and precipitation chemistry in the southwestern US
Precipitation chemistry is governed largely by the composition of the seeds of warm cloud droplets (CCN) and snow (IN), and gases and particles that deposit to these hydrometeors.There have been limited attempts to examine precipitation chemistry in relation to air mass source origins and particulate matter composition in the Southwest.Hutchings et al. (2009) focused on monsoon clouds near Flagstaff, Arizona and suggested that windblown soils serve as CCN and can be found in cloud water.It is widely accepted that dust particles act as both CCN (Levin et al., 1996;Rosenfeld et al., 2001;Koehler et al., 2007) and IN (Isono and Ikebe, 1960;Kumai, 1961;Twohy and Gandrud, 1998;Heintzenberg et al., 1996;DeMott et al., 2003a, b;Sassen et al., 2003;Cziczo et al., 2004;Koehler et al., 2007;Prenni et al., 2009;Zimmermann et al., 2008), which is important for the Southwest as it has the highest dust concentrations in the United States (e.g.Malm et al., 2004).This is assisted by disrupted soils from agricultural activity, vehicles, construction, grazing, and mining operations (Schlesinger et al., 1990;Neff et al., 2005;Fernandez et al., 2008;Csavina et al., 2012).Atmospheric dust not only originates from regional sources in the Southwest and Mexico, but it can also be transported from distant regions such as Asia, especially in spring months (VanCuren and Cahill, 2002;Jaffe et al., 2003;Wells et al., 2007;Kavouras et al., 2009).In addition to dust, the region is impacted by diverse anthropogenic and biogenic sources with the relative strength of each of these sources being sensitive to meteorological and seasonal factors.
The goal of this work is to examine co-located aerosol and wet deposition chemical measurements at six Southwest sites with an aim to characterize their spatiotemporal trends and interrelationships.The analysis specifically aims to address the following questions: (i) What is the annual profile of rain/snow water accumulation, precipitation pH, and composition of precipitation and aerosol particles?; (ii) What species are best correlated with each other in rain and snow?; (iii) What species are most influential towards rain and snow water pH?; (iv) How well-correlated are common species measured in aerosol and precipitation samples?; (v) What is the nature of the nitrate:sulfate ratio in precipitation and aerosol particles?and (vi) How have aerosol and precipitation species concentrations changed between 1995 and 2010? 2 Data

Aerosol data
Aerosol composition data were obtained from the Interagency Monitoring of Protected Visual Environments (IM-PROVE) network (Malm et al., 1994(Malm et al., , 2004;; http://views.cira.colostate.edu/web/).IMPROVE aerosol monitoring stations are located primarily in National Parks and Wilderness Areas and contain samplers that collect ambient aerosol on filters over a period of 24 h, typically every third day.
Prior to 2000, sampling was conducted twice each week with a 24 h duration per sample.The change in sampling frequency in 2000 is not expected to bias the results over the monthly and seasonal time scales of interest in this study.Collected samples are analyzed for ions, metals, and both organic carbon (OC) and elemental carbon (EC).Ammonium is not routinely measured in the IMPROVE program and thus its concentrations in precipitation are only discussed.Sampling protocols and additional details are provided elsewhere (http://vista.cira.colostate.edu/improve/Publications/SOPs/UCDavis SOPs/IMPROVE SOPs.htm).Nitrate is vulnerable to measurement artifacts and this issue is minimized via the use of an annular denuder (to remove nitric acid, HNO 3 ) and nylon filters as compared to Teflon to prevent NO − 3 loss via recapture of volatilized HNO 3 (Ames and Malm, 2001;Yu et al., 2005).This study uses data from six sites summarized in Table 1 and Fig. 1 in terms of location, altitude, and range of dates for which data are examined.Specific species concentrations discussed in this study are from the "fine" fraction of aerosol, PM 2.5 , while total mass concentrations are also reported for the "coarse" fraction, defined as PM 10 − PM 2.5 .Among the elemental measurements, x-ray fluorescence (XRF) is used for iron (Fe) and heavier elements while particle-induced x-ray emission (PIXE) is used for elements ranging from sodium (Na) to manganese (Mn).Fine soil is discussed in this work and is calculated from IM-PROVE tracer concentrations using the following equation (Malm et al., 2004): +2.42[Fe] + 1.94 [Ti] (1) Statistical methods used to analyze IMPROVE and the precipitation data below are briefly summarized in the Supplement.

Precipitation data
Precipitation chemistry and pH data are reported from six sites (Table 1 and Fig 3 ), potassium (K + ), sodium (Na + ), and sulfate (SO 2− 4 ).Data that were obtained from the NADP data repository have undergone quality control and assurance protocols (http://nadp.sws.uiuc.edu/data/ntndata.aspx).Data have been categorized to separate rain and snow, with no instances of rain-snow mixtures included in the analysis.Since sample handling procedures at all NADP/NTN sites changed substantially on 11 January 1994, data are only used beginning in 1 January 1995 or the first day of January in another year if data collection began in the middle of a year.

Remote sensing data
Regional maps of ultraviolet aerosol index (UV AI) were developed using data from the Ozone Monitoring Instrument (OMI) for the period between 2005-2008.Data were obtained at a resolution 1 • × 1.25 • using a minimum threshold value of 0.5 (Hsu et al., 1999).The UV AI parameter serves as a proxy for absorbing aerosol particles (Torres et al., 1998), which are predominantly comprised of smoke and dust.UV AI is used here as a proxy for dust owing to its greater abundance relative to smoke in the region over the time scales examined in this work.

Site descriptions
The six sites studied represent areas throughout the southwestern United States influenced by varying degrees of pollution and meteorological conditions (Fig. 1).Organ Pipe National Monument is the lowest altitude site (∼ 500 m a.s.l.) and the closest to marine-derived emissions from the Pacific Ocean.Organ Pipe is approximately 16 km north of the US-Mexico border in southern Arizona.Anthropogenic pollution sources include the towns of Sonoyta, Mexico (population ∼ 15 000, ∼ 10 km south; http://www.inegi.org.mx/default.aspx) and Ajo, Arizona (city population ∼ 3500, ∼ 36 km north; US Census Bureau, 2010).Chiricahua National Monument (∼ 1560 m a.s.l.) is located in the Chiricahua Mountains in southeastern Arizona, approximately 18 km west of the Arizona-New Mexico border.Willcox, Arizona (city population ∼ 3800; US Census Bureau, 2010) is located 55 km west of Chiricahua and contains the Willcox Playa and the Apache Power Station, which is a coal-fired power station.Sierra Vista, Arizona (city population ∼ 44 000; US Census Bureau, 2010) is located 97 km to the southwest of Chiricahua.The largest source of major urban pollution is Tucson, Arizona (city population ∼ 520 000; US Census Bureau, 2010), which is 150 km to the west of Chiricahua.This site can also be influenced by copper smelter emissions from the Mexican towns of Cananea and Nacozari (140 km and 180 km south of Chiricahua, respectively).
The Gila stations (∼ 1775 m a.s.l.) are in southwestern New Mexico.

Air mass source regions
Figure 2 summarizes the representative air mass source regions for each site as a function of season using threeday back-trajectory data from the NOAA HYSPLIT Model (Draxler and Rolph, 2012).Four seasons are defined in this study as follows: December-February (DJF), March-June (MAMJ), July-September (JAS), October-November (ON).The MAMJ season is meant to include the months with strongest dust influence, while JAS represents the monsoon season.Air masses from the Pacific Ocean influence all sites, with the strongest influence on Organ Pipe due to its proximity to the ocean.The three southernmost sites (Organ Pipe, Chiricahua, Gila) tend to exhibit similar trajectory frequency patterns relative to the three sites that are farther north.The former three stations that are closest to the US-Mexico border are most influenced by crustal emissions from the Sonoran Desert, dry lake beds such as Laguna Salada (southwest of Yuma, Arizona), the Chihuahuan Desert and a network of playas and alluvial, lacustrine, and aeolian sediments near the Mimbres Basin by southwestern New Mexico.The major seasonal difference at the easternmost sites is that the MAMJ trajectories originate farthest from the west, while JAS tends to coincide with more influence from towards the Gulf of Mexico.This is consistent with the arrival of monsoon moisture from the Gulf of Mexico during this time of year (Adams and Comrie, 1997;Higgins et al., 1997).Mesa Verde and Bryce Canyon exhibit similar trajectory frequency maps and receive more influence from the northwest direction as compared to the other sites.The DJF and ON seasons are characterized by being influenced by air with the smallest range of distance away from the study sites owing to meteorological conditions suppressing transport relative to the other two seasons.The majority of the backtrajectories include the Phoenix metropolitan area, which have previously been linked to enhanced levels of anthropogenic species (e.g.sulfate, lead, copper, cadmium) in cloud water more than 200 km to the north in Flagstaff, Arizona (Hutchings et al., 2009).

Aerosol data
The majority of the aerosol mass at the study sites resides in the coarse fraction, which is due to the strong influence of dust (Fig. 3).The two lowest altitude sites (Organ Pipe and Chiricahua) exhibit the highest coarse aerosol concentrations on an annual basis with their concentration peaks in July (9.55 ± 7.41 µg m −3 ) and May (8.97 ± 3.74 µg m −3 ), respectively.Owing to Organ Pipe's lower altitude and closer proximity to dust and sea salt sources, it exhibits higher concentrations year-round with fairly sustained average coarse aerosol concentrations between April and September (8.25-9.55 µg m −3 ).The spatial and temporal patterns in coarse aerosol concentrations across the Southwest are consistent with seasonal UV AI maps (Fig. 4).The highest regional values occur during MAMJ, followed by JAS, ON, and then DJF.The sites co-located with the highest and lowest year-round UV AI levels are Organ Pipe and Bryce Canyon, respectively.A consistent feature at all sites except Organ Pipe is that the ratio of coarse : fine aerosol mass is highest during MAMJ (Fig. 5); this ratio can be used as a measure of when coarse dust aerosol influence is strongest from local sources (Tong et al., 2012).The average coarse : fine ratio at Organ Pipe is highest in DJF (1.98); the different behavior of this ratio at this site may be due to its proximity to marine-derived sea salt emissions (Fig. 2).PM 2.5 concentrations peak between May and July for the six sites, indicative of sources and production mechanisms (i.e.gas to particle conversion) that differ from coarse aerosol in the region.The most abundant contributors to PM 2.5 are fine soil, organic carbon (OC), SO 2− 4 , and NO − 3 (Fig. 6).Fine soil levels are highest in the spring months (April-May) owing largely to dry conditions, high wind speeds, and also the highest frequency of transported Asian dust (VanCuren and Cahill, 2002;Jaffe et al., 2003;Wells et al., 2007;Kavouras et al., 2009;Tong et al., 2012).The contributions of Ca, Mg, and Na to PM 2.5 are highest during MAMJ due most likely to fine soil emissions (Fig. 5).Potassium is associated with crustal matter and biomass burning emissions, and its highest concentrations and mass fractions occur during MAMJ.Although no direct measurement of organic carbon (OC) is available in the precipitation datasets, OC in the PM 2.5 fraction is still examined owing to its significant contribution ranging from 10-29 % depending on the site and season (Fig. 5); note that the inorganic aerosol constituents examined account for between 28-47 % of PM 2.5 .Organic carbon has a variety of sources in the Southwest where it is produced via both direct emission and secondary production processes from sources including biomass burning, biological particles, biogenic emissions such as isoprene, combustion, meat cooking, plant debris, and dust (Bench et al., 2007;Schichtel et al., 2008;Holden et al., 2011;Sorooshian et al., 2011;Cahill et al., 2013;Youn et al., 2013).Although the atmospheric mixing height is largest between May-July in the region (Sorooshian et al., 2011), OC concentrations are the highest at all the sites during this time suggestive of the influence of biomass burning and secondary OC production.Sulfate production is enhanced during moist conditions, which occurs during the monsoon months in the Southwest.As a result, maximum concentrations (Fig. 6) and mass fractions (Fig. 5) for SO 2− 4 are observed during JAS.Nitrate is a marker for anthropogenic emissions as it often increases in concentration with decreasing mixing height in the winter months and because it is thermodynamically more stable in colder conditions; however, it is also associated with larger particles in the fine mode owing to reactions of HNO 3 (or precursors) with dust and sea salt (Malm et al., 2003;Lee et al., 2004Lee et al., , 2008)).As a result, NO − 3 exhibits a bimodal concentration profile with a peak in the winter months Fig. 5. Average monthly mass fractions of selected PM 2.5 constituents for all six IMPROVE sites and for four seasons.The labels for each color in the top left pie are the same for the other pies.Also reported are average PM 2.5 and coarse aerosol concentrations in units of µg m −3 , the concentration ratio of OC to PM 2.5 , and the concentration ratio of the sum of the seven inorganic components of the pies ("Inorg") relative to PM 2.5 .These results are based on data ranges in Table 1 for each site.and during the spring months when soil dust is most abundant.Nitrate mass fractions are usually highest in DJF.Chloride exhibits peak concentrations in various months (March, May, June, October-December) depending on the site.Maximum concentrations observed at the majority of sites between March and June likely originate from a combination of crustal-derived particles and other sources such as biomass burning (e.g.Wonaschütz et al., 2011).Chloride is especially enhanced at Organ Pipe due to marine-derived sea salt, which is supported by higher mass fractions of Cl − and Na at this site relative to others (Fig. 5).

Annual rain and snow accumulation profiles
Precipitation falls in two major modes (Fig. 7).The first is during DJF mostly as a result of Pacific Ocean frontal storms.These storms provide snow to high altitude sites and warm rain to lower altitude sites.The second mode is the summertime monsoon rainfall that typically occurs between July and October.The lowest altitude site, Organ Pipe, was the only one to have no snow data recorded.The next lowest altitude site, Chiricahua, has relatively similar amounts of snow and rain during the DJF period.This site also is characterized by major enhancements in precipitation during the monsoon season, with the two highest amounts in July and August (71 mm and 90 mm, respectively).The relative amount of snow in DJF relative to rain during JAS increases as a function of altitude and distance to the north for the other sites: Bryce Canyon > Mesa Verde > Bandelier > Gila Cliffs.Table S1 (Supplement) reports more specific statistics for precipitation data for each month and site.July and August are the months with the most frequent rain days (∼ 5-12 depending on the site).The month with most frequent snow days (∼ 1-7 days, depending on the site) varied between December and February.
To more closely examine when dust impacts precipitation in the Southwest, Ca 2+ and Mg 2+ are used as rain tracer species (e.g.Stoorvogel et al., 1997;Reynolds et al., 2001;Rhoades et al., 2010); other crustal-derived rain constituents such as K + and Na + /Cl − are not used as they likely have contributions from biomass burning and sea salt, respectively.The rain water concentration sum of Ca 2+ and Mg 2+ is highest at all sites during the months of April-June (Fig. 8), which coincides with the highest levels of dust according to IMPROVE and satellite data (Figs.3-6 between April-May, which presumably explains why they also have the highest rain pH in those months.Rain Cl − and K + concentrations are also highest during MAMJ, likely due to crustal emissions (dust and sea salt); Cl − is most abundant at Organ Pipe for nearly the entire year due to sea salt from marine-derived air masses that impact the site year-round (Fig. 2).Nitrate and SO 2− 4 exhibit different annual concentration profiles in precipitation as compared to PM 2.5 for reasons that will be discussed subsequently.
Figure S1 (Supplement) shows annual cycles for snow water constituent concentrations.Annual snow pH values range between 5 and 6 at the various sites, similar to rain water.Snow pH and the concentration sum of Ca 2+ and Mg 2+ are highest between March and May for three sites (Gila Wilderness, Chiricahua, Mesa Verde), and between September and October for Bryce Canyon and Bandelier.The rest of the species exhibit their highest concentrations in a wide range of months depending on the site.

Precipitation species mass fractions
Either Cl − , SO 2− 4 , or NO − 3 is the dominant rain anion on a mass basis depending on the site and season (Fig. 9).Chloride exhibits the highest anion mass fraction in Organ Pipe rain during DJF (29 %) due largely to sea salt.Nitrate is the dominant anion at Organ Pipe during JAS (44 %) and ON (39 %), while all three anions are nearly equivalent contributors during MAMJ (20-24 %).Sulfate and NO − 3 exhibit the highest anion mass fractions in rain at the other sites with a consistent trend being that NO − 3 accounts for the highest mass fraction in JAS and MAMJ.The highest cation mass fraction in rain was usually for Ca 2+ (6-27 %) at all six sites and seasons with the following exceptions: NH + 4 (10-13 %; Bandelier DJF, Chiricahua DJF/ON, Organ Pipe JAS); Na + (14-18 %; Organ Pipe DJF/MAMJ).Snow mass fraction data are only shown for DJF in Fig. 9 due to insufficient data in other months.The highest snow cation mass fraction in DJF was always for Ca 2+ (9-19 %), followed by either NH +

4
(5-7 %), K + (8 %), or Na + (9 %).The anion with the highest mass fraction in snow was usually NO − 3 (28-49 %), followed by SO 2− 4 (19-29 %), and Cl − (4-14 %).In other regions such as those associated with the Acid Deposition Monitoring Network in East Asia (EANET; EANET Executive Summary, 2011), the Tibetan Plateau, Canada, Spain, India, and Israel, the dominant precipitation cation has been reported to be either Ca 2+ , Na + , or NH + 4 (Avila et al., 1998;Herut et al., 2000;Kulshrestha et al., 2005;Zhang et al., 2007 and references therein;Aherne et al., 2010;Yi et al., 2010;Zhang et al., 2012).Those studies also showed that SO 2− 4 was the dominant anion, which may be due to significant anthropogenic influence in those studies; the one exception was in western Canada where marine-influenced air promoted Cl − to be the dominant anion.Calcium and Cl − were shown to be the dominant cation and anion, respectively, in Jordan rain water (Al-Khashman, 2009).Consistent with our results, Hutchings et al. (2009) showed that NO − 3 was frequently more abundant than SO 2− 4 in northern Arizona monsoon cloud water; however, they also showed that NH + 4 was the dominant cation.San Joaquin Valley and Sacramento fog water in California exhibited high NO − 3 : SO 2− 4 concentration ratios (equivalent/equivalent) of 4.8 and 8.6, respectively, due to the influence of agricultural emissions (Collett et al., 2002).It is cautioned again that such comparisons are sensitive to the time span of data examined due to reasons such as varying air quality regulations at different locations and times.Significant changes in the relative amounts of SO 2− 4 and NO − 3 have been observed in the United States since the 1980s (e.g.Butler and Likens, 1991;Lynch et al., 1995;Nilles and Conley, 2001;Butler et al., 2001;EPA, 2003).

Interrelationships between precipitation species concentrations
Correlation matrices for rain and snow chemical concentrations are used to provide more support for common sources of species, using Organ Pipe and Bandelier as representative examples for rain and snow, respectively (Table 2).Tables S2-S3 report the rest of the matrices for the six sites, which show the same general relationships as those in Table 2.The crustal-derived species (Ca 2+ , Mg 2+ , K + , Na + , Cl − ) exhibit statistically significant correlations (95 % confidence using a two-tailed Student's t test; this condition applies to all correlations reported hereinafter) with each other in both rain and snow (r = 0.48-1.00,n = 90-107), suggesting that their common source is dust or sea salt depending on the site.Sodium and Cl − are strongly correlated at the site closest to marine emissions, Organ Pipe (r = 1.00).These two species exhibit high correlations for both rain and snow at the other sites too (r = 0.66-0.97).Sulfate, NH + 4 , and NO − 3 are highly correlated with each other relative to other species in rain and snow reflecting noncrustal sources, specifically anthropogenic emissions in the form of SO 2 , nitrogen oxides (NO x ), and ammonia (NH 3 ).Sulfate, NO − 3 , and NH + 4 in precipitation originate from scavenging of these species in the aerosol phase and also from transfer of their vapor precursors: SO 2− 4 from SO 2 ; NO − 3 from nitric acid (HNO 3 ), which originates from NO x emissions; NH + 4 from NH 3 .Ammonium typically serves as a base for sulfuric and nitric acids and originates from NH 3 , which is emitted from livestock waste, fertilizer applications, biomass burning, motor vehicle emissions, and coal combustion (e.g.Apsimon et al., 1987;Asman and Janssen, 1987;Kleeman et al., 1999;Anderson et al., 2003;Battye et al., 2003;Sorooshian et al., 2008).The dominant route by which SO 2− 4 becomes associated with drops is thought to be aerosol scavenging (e.g.van der Swaluw et al., 2011).Other work has shown that the close relationship between SO 2− 4 and NO − 3 in rain and snow is mainly linked to anthropogenic inputs (e.g.Wake et al., 1992;Legrand and Mayewski, 1997;Schwikowski et al., 1999;Preunkert et al., 2003;Olivier et al., 2006;Dias et al., 2012).Ammonia from anthropogenic sources has also been linked to soluble ion measurements in ice and rain (Kang et al., 2002;Hou et al., 2003).
The crustal cation species (Ca 2+ , Mg 2+ , K + , Na + ) exhibit statistically significant correlations with SO 2− 4 , NO − 3 , and Cl − at all sites.This is suggestive of reactions of acids (e.g.nitric, sulfuric, hydrochloric acids) with crustal surfaces such as dust and sea salt (e.g.Matsuki et al., 2010).This link is supported by a large inventory of previous work: (i) measurements in Asia indicate that dust is a significant source of SO 2− 4 , largely of anthropogenic origin which comes together with dust, in snow and glaciers (Wake et al., 1990;Kreutz et al., 2001;Zhao et al., 2011); (ii) a close association of SO 2− 4 with crustal matter was argued to explain the close relationship between SO 2− 4 and Ca 2+ in rain water in India (Satyanarayana et al., 2010); (iii) Zhang et al. (2007) suggested that acids such as HCl react with windblown crustal particles to yield a high Mg 2+ /Cl − correlation in China; and (iv) dust surfaces have been shown to become coated with soluble species such as SO 2− 4 , NO − 3 , and Cl − (Desbouefs et al., 2001;Sullivan et al., 2007;Matsuki et al., 2010) leading to enhanced hygroscopic properties (Levin et al., 1996;Koehler et al., 2007;Crumeyrolle et al., 2008;Sorooshian et al., 2012).Correlations between similar subsets of species (crustal species, SO 2− 4 /NH + 4 /NO − 3 , and the combination of the latter two) have also been observed in other regions such as the Mediterranean, Turkey, India, Brazil, Mexico, and China (Al-Momani et al., 1997;Basak and Agha, 2004;Safai et al., 2004;Mouli et al., 2005;Baez et al., 2007; Zhang et al., 2007;Teixeira et al., 2008;Yi et al., 2010;Raman and Ramachandran, 2011).

Interrelationships between aerosol and precipitation species
It is of interest to examine the extent to which aerosol and precipitation species concentrations are related.As SO 2− 4 and fine soil represent the most abundant PM 2.5 constituents of interest in this work (excluding other constituents such as carbonaceous species), their particulate concentrations are compared to all precipitation species concentrations in Table 3.The following factors could bias the interpretation of these results: (i) gases that partition to hydrometeors; and (ii) different air masses affecting altitudes at which the IM-PROVE measurements take place and where precipitation is produced.With the exception of Organ Pipe, crustal-derived species in rain (Ca 2+ , Mg 2+ , K + , Cl − , Na + ) exhibit statistically significant correlations with fine soil.Although not shown in Table 3, particulate Cl − was only correlated with rain Cl − (r = 0.29; n = 105) at one site (Organ Pipe) because of the proximity of Organ Pipe to the Pacific Ocean; particulate Cl − was also correlated with Na + at this site (r = 0.29, n = 105).Interestingly, NH + 4 , SO 2− 4 , and NO − 3 in rain are also correlated with fine soil at four sites including Organ Pipe.This result is consistent with these same anthropogenically-related species being related to the crustal species in the rain data.Fine soil levels exhibit statistically significant correlations with those of crustal-derived species in snow at Bryce Canyon, Mesa Verde, and Gila.
Particulate SO 2− 4 exhibits a statistically significant correlation with SO 2− 4 in rain at all sites except Chiricahua.Particulate SO 2− 4 was also correlated with NO − 3 and NH + 4 in rain at four sites including Organ Pipe and Chiricahua.Particulate SO 2− 4 exhibits few statistically significant correlations with snow species: it only exhibited positive correlations with SO 2− 4 and NO − 3 at Bryce Canyon.The different and SO 2− 4 both exhibit higher overall mass fractions in rain relative to snow during DJF.One explanation is the efficient adsorption of gaseous NO − 3 precursors such as HNO 3 to snow (e.g.Jacobi et al., 2012); however, the relative strength of partitioning of HNO 3 to rain drops and snow is uncertain and requires additional investigation for this region.Another explanation could be the preferential role of different particle types in serving as CCN versus IN, which was already suggested to explain why particulate SO 2− 4 was mainly correlated with SO 2− 4 in rain rather than snow.More effective nucleation scavenging of hygroscopic particles containing SO 2− 4 at lower altitudes in the form of CCN would limit their ability to reach higher altitudes where deeper clouds produce snow.At those higher altitudes, dust particles can serve as effective IN (Isono and Ikebe, 1960;Kumai, 1961;Twohy and Gandrud, 1998;Heintzenberg et al., 1996;DeMott et al., 2003a, b;Sassen et al., 2003;Czizco et al., 2004;Koehler et al., 2007;Prenni et al., 2007;Zimmermann et al., 2008), and as noted already, they contain enhanced levels of NO − 3 due to reactions with HNO 3 (e.g.Malm et al., 2004;Lee et al., 2008).This speculation is partly supported by the finding that the coarse : fine aerosol ratio was positively correlated with snow pH at more sites (Bryce Canyon, Gila, Mesa Verde) than with rain pH (Mesa Verde).But a conflicting re-sult is that the snow ratio of NO − 3 : SO 2− 4 does not exhibit a statistically significant relationship with the coarse : fine aerosol mass ratio at any site.It is unclear as to whether this is due to dissimilar air masses influencing altitudes where snow is produced relative to the IMPROVE stations.More detailed investigations would assist with explaining the findings above related to the NO − 3 : SO 2− 4 ratios, especially examining HNO 3 partitioning behavior and the role of different particle types in serving as CCN and IN in the Southwest.

Interannual variability in aerosol and precipitation chemistry
Previous analyses of NADP/NTN concentration data over the United States between 1985 and 2002 showed general increases in ammonium, reductions in sulfate, and mixed changes in nitrate depending on location (Lehmann et al., 2005); furthermore, reductions in sulfate have been shown to be more significant as compared to nitrate (Lehmann and Gay, 2011).As JAS is the season with the most available precipitation data across all sites, a long-term trend analysis for this season shows that the only species in rain exhibiting a statistically significant concentration change is SO 2− 4 (Table 4).This species exhibited a decreasing trend at Bryce Canyon (−0.062 mg L −1 yr −1 ) and Gila Cliff (−0.057 mg L −1 yr −1 ).The decreasing trend is ubiquitous across the region in the fine aerosol fraction, with the largest reduction at Organ Pipe (−0.109 µg m −3 yr −1 ); the reduction at other sites ranged between −0.029 and −0.047 µg m −3 yr −1 .This reduction in the region can be explained by air regulations of SO 2− 4 precursors (e.g.Matichuk et al., 2006;Sorooshian et al., 2011).Nitrate does not exhibit a statistically significant change in concentration in rain or in particles, except relatively small reductions as compared to SO 2− 4 at Chiricahua (−0.006 µg m −3 yr −1 ) and Organ Pipe (−0.016 µg m −3 yr −1 ).Other work in the Southwest has suggested that a lack of a change of NO − 3 over the last decade in at least one part of the Southwest (i.e.southern Arizona) may be due to competing factors: (i) land-use changes (e.g.agricultural land to urban areas) can reduce NH 3 emissions and particulate NO − 3 formation; and (ii) higher NO x emissions linked to population growth and reductions in SO 2− 4 allow for more NH 3 to neutralize HNO 3 to promote ammonium nitrate (NH 4 NO 3 ) production (Sorooshian et al., 2011).While the NO − 3 : SO 2− 4 ratio in the fine aerosol fraction only increased at one site (Mesa Verde), there was an increase in rain at all sites except Chiricahua and Organ Pipe.Rain pH has also increased at all sites except Mesa Verde and Organ Pipe; the increase at four of the sites is due to reductions in SO  specifically, Organ Pipe was the only site to show an increase in the coarse : fine aerosol mass ratio in JAS, with an increasing rate of 0.084 yr −1 .This result is suggestive of the presence of more coarse particle types, mainly sea salt and dust, that can react with HNO 3 to form particulate NO − 3 , simultaneous with reduced fine aerosol SO 2− 4 over time.

Conclusions
This study characterized aerosol and precipitation composition at six sites in the US Southwest.The main results of this work are as follows, following the order of questions posed in Sect.1: i. Precipitation accumulation is concentrated in a wintertime mode (DJF) and a monsoon mode (JAS), with only warm rain associated with the latter.The relative amount of rain and snow during DJF depends on geography and altitude, with rain being more abundant farther south near the international border and at lower altitudes.All aerosol and precipitation species concentrations typically were highest during MAMJ (including precipitation pH) due to increased dust concentrations.
ii. Statistically significant relationships in the regional rain and snow are observed for numerous crustal-derived species (Ca 2+ , Mg 2+ , K + , Na + ), mainly from dust, and a subset of species with anthropogenic sources (NH + 4 , SO 2− 4 , NO − 3 ).Species in the crustal group also exhibit positive relationships with SO 2− 4 , NO − 3 , and Cl − , suggesting that acidic gases likely react with and partition to either coarse crustal particles or hydrometeors enriched with crustal constituents.Organ Pipe, the site closest to the Pacific Ocean, shows an especially strong relationship between Na + and Cl − in rain water due to sea salt influence, indicating that this aerosol type more strongly affects precipitation in parts of the Southwest closest to the ocean.
iii.Rain and snow pH levels were usually between 5-6.
Rain pH was highest during MAMJ, which was coincident with the highest rain and particulate concentrations of crustal-derived species (Ca 2+ , Mg 2+ , K + , Na + ).Rain and snow pH were generally well-correlated with these species showing that dust in the region is highly influential in acid-neutralization.
iv. Crustal-derived species in both rain and snow (Ca 2+ , Mg 2+ , K + , Cl − , Na + ) exhibit statistically significant correlations with particulate fine soil.The coarse : fine aerosol mass ratio was correlated with snow concentrations of crustal species (Ca 2+ , Mg 2+ , Na + , K + , Cl − ) and NO − 3 , suggestive of a preferential role of coarse particles (mainly dust) as IN in the region.Particulate SO 2− A. Sorooshian et al.: Aerosol and precipitation chemistry in the southwestern US in rain was Organ Pipe, which exhibited the only longterm increase in the particulate coarse : fine mass ratio.Increasing relative amounts of coarse particles as compared to fine particles is thought to increase rain pH due to reduced influence from fine particulate SO 2− 4 and increased influence from basic particulate species that are concentrated in the coarse fraction.Furthermore, reactions of HNO 3 with coarse particle types and potential partitioning of this species to rain and snow can promote higher NO − 3 : SO 2− 4 ratios.
Future research is needed to test hypotheses used in this work to explain some of the results for the Southwest, including (i) the role of different particle types in serving as CCN and IN and (ii) the partitioning behavior of gases such as HNO 3 to particles and hydrometeors.While this work has looked at factors influencing precipitation chemistry, it is noted that another major issue in the Southwest is deposition of aerosol particles to high altitude areas that reside in the snowpack or fall as summer rain and release nutrients into downstream ecosystems (Psenner, 1999;Lawrence and Neff, 2009).For example, mineral dust is thought to be among the strongest sources of atmospheric phosphorus (Okin et al., 2004;Mahowald et al., 2008) and its deposition at highelevation sites represents a major nutrient source for lakes (Morales-Baquero et al., 2006;Vicars and Sickman, 2011).Case studies in the Southwest have shown that dust events can influence the composition of snow water, specifically leading to enhancements in snowpack pH and calcium levels (Rhoades et al., 2010).Similar findings have linked dust to elemental composition of both precipitation and snow and changes in surface water chemistry (e.g.Landers et al., 1987;Turk et al., 2001).Other work has suggested that aerosol deposition can be a source of harmful contaminants such as lead (Liptzin and Seastedt, 2010).
Dust particles can also have a large impact on the melt rate of mountain snowpacks in Colorado by lowering the albedo, from 0.7 to 0.4 on average, and thereby increasing shortwave radiation inputs to the snowpack (Painter et al., 2010;Skiles et al., 2012).We observed the highest coarse aerosol mass concentrations and other proxies of dust during MAMJ when snow is on the ground at most of the mountains surrounding the study sites.Recent work from Colorado has shown that the advancement in the loss of snow cover from dust, due to faster melts, is linearly related to the amount of dust in the snowpack, despite variability in irradiance and the timing of dust deposition (Skiles et al., 2012).Predicting the amounts of wet and dry dust deposition to and from the Southwest is therefore critical to predicting snowmelt rates and downstream water resources of the Colorado River Basin (Painter et al., 2010).More research is necessary to combine information on dust sources and deposition, as done in the current study, with regional variability in hydroclimate and snow processes (Harpold et al., 2012) in the mountains of the western US.

Fig. 2 .
Fig. 2.Seasonal HYSPLIT data showing the approximate source regions for air parcels ending 10 m AGL at each of the six study sites that are represented by red open markers.The colored borders represent a minimum trajectory frequency of 1 % using three-day back-trajectory data, where frequency is defined as the sum of the number of trajectories that passed through each point on the map divided by the number of trajectories analyzed.

Fig. 9 .
Fig. 9. (Top four rows) Summary of pH and chemical mass fraction data for rain during different periods of the year.(Bottom row) Snow pH and chemical mass fraction data for DJF, which is the season with the most snow data available.The labels in the top left pie are the same for the other pies.Note that during DJF there is no rain data for Bryce Canyon or snow data for Organ Pipe.
to increases in NO − 3 .A potential reason as to why Organ Pipe does not show increases in either the NO − 3 : SO 2−

Table 1 .
Summary of co-located aerosol (IMPROVE) and precipitation (NADP/NTN) data used with coordinates, altitudes, and range of full years in which data are analyzed.The location of sites is shown in Fig.1."NP" and "NM" refer to National Park and National Monument, respectively.Altitudes are ASL.
Fig. 1.Spatial map of co-located EPA IMPROVE and NADP/NTN stations used in this study.
The nearest town is Silver City, New Mexico (city population ∼ 10 000; US Census Bureau, 2010), which includes a number of large open-pit copper mining Average monthly fine (PM 2.5 ) and coarse (PM 10 − PM 2.5 ) aerosol mass concentrations at six EPA IMPROVE sites.These results are based on data ranges shown in Table1for each site.
Atmos.Chem.Phys., 13, 7361-7379, 2013 www.atmos-chem-phys.net/13/7361/2013/ . Examples of regions with higher pH values (> 6) than those in the Southwest, mostly due to alkaline species (e.g.ammonium from agriculture and calcium Average monthly PM 2.5 constituent mass concentrations at six EPA IMPROVE sites.Shaded regions represent when maxima are observed for individual or groups of sites.These results are based on data ranges shown in Table1for each site.Average monthly precipitation accumulation at the six NADP/NTN sites over the data ranges shown in Table1.
Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Fig.8.Annual pH and concentration profiles for rain in the Southwest.Shaded regions represent when maxima are observed for individual or groups of sites.These results are based on data ranges shown in Table1for each site.

Table 2 .
Correlation matrix (r values) for rain water constituent concentrations measured at Organ Pipe between 2003 and 2010 and snow water constituent concentrations measured at Bandelier between 1995 and 2010.Values are only shown when statistically significant (95 %) with a two-tailed Student's t test.Refer to Supplement for all data for the six sites.

Table 3 .
Correlation (r) of aerosol mass concentrations (fine soil, sulfate) and the coarse : fine mass concentration ratio with precipitation species mass concentrations.Values are only shown when statistically significant (95 %) with a two-tailed Student's t test.There are no snow data at Organ Pipe.The sample range for data below is 39-240.

Table 4 .
Long-term trend analysis for the Southwest monsoon season (JAS).Slopes of each parameter versus year are shown with correlation coefficients (r 2 ) of the linear best fit line in parenthesis.Units are µg m −3 yr −1 for the aerosol species, mg L −1 yr −1 for the rain species, and yr −1 for the coarse : fine ratio, NO − 3 : SO 2− 4 ratio, and pH.No other common aerosol and rain water species are shown as they do not have statistically significant changes over the durations shown in Table1.