1. Introduction

Anthropogenic pollutants and dust storms originatingfrom East Asia not only impact on local environments and human health but also exert a broadrange of influences on the climate(Wake et al., 1994;Gao et al., 1997; Streets and Waldhoff, 2000; Akimoto, 2003; Streets et al., 2003a; Xia et al., 2005;Arimoto et al., 2006; Ramanathan and Carmichael, 2008; IPCC, 2013). The long-range transport of duststorms and anthropogenic pollutants from their continentalsources creates serious problems in the quantificationof regional and global climate change(Duce et al., 1980; Prospero, 1999a, b; Husar et al., 2001; Arimoto et al., 2006; Chen et al., 2010; Wu et al., 2010;Prospero and Mayol-Bracero, 2013; Wang et al., 2013).As a major contributor to the aerosol loading in thetroposphere, the outbreak of dust storms may affectthe radiation balance of the atmosphere by absorbing and reflecting solar radiation(Tegen et al., 1996; Dickerson et al., 1997; Xu et al., 2004), and cause a reductionin the photolysis rate, thereby inhibiting ozoneproduction(Bonasoni et al., 2004). Mineral dust alsoaffects cloud processes and alters hydrological cycles(Rosenfeld et al., 2001). An additional effect causedby dust particles is the modification of atmospheric gaseous composition through heterogeneous reactions(Dentener et al., 1996). Dust particles mixed with airpollutants create a brownish haze, which absorbs and scatters sunlight, reducing sunlight at the surface and resulting in so-called “global dimming”(Ramanathan et al., 2001).

It has been shown that East Asian air pollutionis a large and growing source of NO_x, SO₂, CO, and other atmospheric pollutants(Jaffe et al., 1999; vanAardenne et al., 1999; Streets and Waldhoff, 2000;Streets et al., 2001; Akimoto, 2003; Streets et al., 2003a). Emission levels of various air pollutants inEast Asia, particularly in China, have been increasingrapidly since 1980 as a consequence of industrialdevelopment, economic growth, and the large-scale migrationof rural residents to cities, causing various airpollution problems(van Aardenne et al., 1999; Streets et al., 2001; Zhang et al., 2003a; Wang et al., 2004; Che et al., 2007). The widespread dust and local air pollutionover northern China makes it one of the largestpollution-affected regions in the world, especially inspring. This transport is sufficiently fast. Severalstudies have reported that transport across the Pacificfrom Asia to North America contributes significantlyto the atmospheric loading of mineral dust, CO, SO₂, NO_x, and O₃(Streets and Waldhoff, 2000;Streets et al., 2000; de Gouw et al., 2004; Jaffe et al., 2005; Huang et al., 2008a; Wang et al., 2010).

Examples of ground-based and airborne-basedmeasurement campaigns recently conducted to examinethe transport of anthropogenic emissions and duststorms include: the Asian Pacific Regional AerosolCharacterization Experiment(ACE-Asia)(Anderson et al., 2003; Conant et al., 2003; Arimoto et al., 2006; Sullivan et al., 2007); the East Asian Studyof Tropospheric Aerosols–An International RegionalExperiment(EAST–AIRE)(Anderson et al., 2003;Huebert et al., 2003; Zhang et al., 2003b, d; Arimoto et al., 2004; Kim et al., 2005; Arimoto et al., 2006; Li et al., 2007b; Xia et al., 2008); the IntercontinentalTransport and Chemical Transformation(ITCT)aircraft campaigns(Brock et al., 2004; de Gouw et al., 2004; Heald et al., 2005; van Curen et al., 2005); the Transport and Chemical Evolution over the Pacific(TRACE-P)aircraft mission(Jacob et al., 2003; Tu et al., 2003; Woo et al., 2003; Allen et al., 2004; Mari et al., 2004); the Intercontinental ChemicalTransport Experiment-Phase B(INTEX-B)campaign(McKendry et al., 2008; Zhang et al., 2009); and thePacific Dust Experiment(PACDEX)(Huang et al., 2008b). Although these experiments have enrichedour knowledge on dust storms and anthropogenic pollution, there have been few attempts to measure tracegas emissions close to the dust sources. The mixingprocesses between trace gases and dust particles and their emissions in East Asia, particularly in China, arestill poorly understood.

To better address the air pollution problems inEast Asia, a field campaign was conducted from lateApril to mid June 2007, which focused specifically onthe direct measurement of aerosol optical properties and the levels of air pollutants near dust sources, suchas the Taklimakan and Gobi deserts. The main objectiveof this study is to identify the initiation ofdust storms and thus to deepen our underst and ingof regional emission sources, as well as to characterizeair pollution using facilities at the Semi-Arid Climate and Environment Observatory of Lanzhou University(SACOL), including both dust particles and trace gases. In Section 2, we describe our experiments and the instrumentation used. In Section 3, we identifythe possible signatures of the mixing of mineraldust with anthropogenic air pollutants both from remoteemission sources during transport. Conclusionsare given in Section 4. The study woll examine the(1)seasonal variations in aerosols and anthropogenicpollutants, (2)properties of mineral dust mixed withanthropogenic pollutants, and (3)emission factors ofair pollutants.

2. Measurements 2.1 Measurement site

An international research experiment facility(Fig. 1)was established in 2005 at SACOL(35.57°N, 104.08°E, 1965.8 m above sea level(a.s.l.)), 48 kmsoutheast of the city center of Lanzhou(Huang et al., 2008b). The site is close to the geometric center of the Chinese mainl and . It is located in the Chinese LoessPlateau(CLP), the largest arid and semi-arid regionin China. This extensive arid and semi-arid regionis experiencing severe shortage of water resources and rapid increase in population, both of which appearto have a significant anthropological impact onthe regional climate and environment, due to the intensificationof l and -use and increasing greenhouse gasemissions.

2.2 Instrumentation 2.2.1 Trace gases

An air quality monitoring system(AQMS-9000)was used to measure levels of ambient air pollutantssuch as carbon moNO_xide(CO), ozone(O₃), nitrogenmoNO_xide(NO), nitrogen dioxide(NO₂), nitrogenoxides(NO_x), sulfur dioxide(SO₂), and particulatematter with diameters < 10 microns PM₁₀. Ambientair samples for trace gases were collected througha Teflon tube equipped with 5-mm Teflon particulatefilters. Samples were analyzed with O₃(EC9810), SO₂(EC9850), and CO(EC9830)monitors(Thermo EnvironmentalInstruments Inc., Franklin, MA, USA). The mixing ratios of NO₂ and NO were analyzed by using acommercial non-dispersive infrared instrument(Jaffe et al., 1999), with an external Mo converter operatedin a time-sharing mode to measure the two tracegases separately. The mixing ratio of O₃ was measuredby UV absorption. Gases used for calibration and daily quality assurance/quality control(QA/QC)activities were obtained from EPA Protocol SO₂, CO, O₃, and NO_x gas cylinders(Scott-Marin, Riverside, CA, USA). All trace gas analyzers were located in aroom at SACOL. To protect the instruments, particlefilters were installed ahead of the sampling lines forthe trace gases. The filters were replaced twice everymonth during this field experiment.

2.2.2 Aerosol optical properties

The mass concentration of PM₁₀ was measuredcontinuously with a Tapered Element Oscillating Microbalance(TEOM: 1400a analyzer, Rupprecht &Patashnick, Albany, NY, USA), using the appropriatesample inlet. The collector operated at 50℃ todry the aerosol, with a flow rate of 16.67 L min⁻¹. Ithas been well documented that such gradual heatingproduces a loss of volatile compounds. The scatteringcoefficient was measured with a single-wavelength totalscatter integrating nephelometer(M9003: Ecotech, Victoria, Australia)designed for general environmentalmonitoring. The nephelometer is available at wavelengthsof 450, 520, and 700 nm. The performance ofthe M9003 was previously assessed in inter-comparison and calibration studies(Heintzenberg et al., 2006;Muller et al., 2009). We applied correction factors(Anderson and Ogren, 1998), using the ratios of trueto-measured nephelometer values for both total scattering and backscattering. An Aethalometer(AE-31:Magee Scientific, Berkeley, California, USA)was usedto measure black carbon(BC)mass concentrations inunit of ng m⁻³. It operated at wavelengths of 370, 470, 520, 590, 660, 880, and 990 nm, with 5-min measurementintervals. Reductions in aerosol light absorptionare determined by measuring multiple scatteringenhancement factors before and after a filter change(Arnott et al., 2005).

Recent studies have reported a wide range of values(2–25 m² g⁻¹)of mass absorption efficiency(MAE)(Horvath, 1993; Petzold et al., 1997; Penner et al., 1998; Sharma et al., 2002; Arnott et al., 2003;Bond and Bergstrom, 2006; Schwarz et al., 2008).A recent review by Bond and Bergstrom(2006)suggestedan MAE value of 7.5 ± 1.2 m² g⁻¹ at 550 nmfor uncoated soot particles. Barnard et al.(2007)indicatedthat the average values of MAE are either 8.9or 8.2 m² g⁻¹ at 500 nm, depending upon the physical and optical parameters assumed for BC, whichare reasonably consistent with the value of 9.5 m² g⁻¹reported by Schuster et al.(2005) and 7.0 m² g⁻¹ reportedby Baumgartner et al.(2002), both of whichwere measured at 550 nm.

In this study, the light absorption coefficients(B_ap)at a wavelength of 520 nm were calculated byusing B_ap = mBC×10 m² g⁻¹, under the assumptionof an average mass absorption efficiency of 10 m²g⁻¹. The vertical distribution of aerosol-attenuatedbackscattering coefficients was derived from a Micro-Pulse Lidar system(MPL-4; Sigma Space, Lanham, MD, USA). The MPL-4 has one measurement channelthat records backscatter signals up to approximately20 km. The primary quantity from this signal is thelowest detected cloud base. The MPL data can beused to calculate cloud scattering cross-sections, opticalthickness, planetary boundary layer height, and aerosol extinction profile(Welton et al., 2001; Huang et al., 2010). The aerosol-attenuated backscatteringcoefficient had already been calibrated by using an overlap factor, so the lower backscattering coefficientwas a higher concentration. The detailed procedurefor processing raw data has been described by Campbellet al.(2002).

The Aerosol Robotic Network(AERONET; version2, level 1.5)measured aerosol optical depths usinga CE318 Cimel sun photometer obtained fromthe National Aeronautics and Space Administration(NASA)Goddard Space Flight Center(GSFC)(http://aeronet.gsfc.nasa.gov). The automatic trackingsun and sky scanning radiometer provides directsun measurements with a 1.2-deg full field ofview, every 15 min in 8 spectral channels, at wavelengthsof 340, 380, 440, 500, 675, 870, 940, and 1020 nm(Holben et al., 1998). Back trajectories fromthe National Oceanic and Atmospheric Administration(NOAA)Air Resources Laboratory(ARL)HybridSingle-Particle Lagrangian Integrated Trajectory(HYSPLIT)model, version 4(Draxier and Hess, 1998)(http://ready.arl.noaa.gov/HYSPLIT.php)have beenused in the experiments undertaken at SACOL to locatethe sources contributing to the relevant air mass and their origins(Huang et al., 2008b; Wang et al., 2010). In this study, a cluster analysis is employed totrack possible dust and local air pollution sources byair mass trajectories. This is considered the most effectivestatistical technique for analyzing a set of trajectoriesdivided into the same, or similar clusters(Kalkstein et al., 1987; Zhang et al., 2013).

2.3 Meteorological parameters

An automated meteorological tower was locatedin the center of the field site to measure ambient temperature, relative humidity, and wind speed at sevenlevels(1, 2, 4, 8, 12, 16, and 32 m above ground level(AGL)). The wind speed(WS), air temperature(T), relative humidity(RH), and air pressure(p)were measured continuously throughout the study period and reported every 30 min. Note that only the meteorologicalparameters measured at 2 m were selectedto represent the near-surface conditions.

3. Results and discussion 3.1 Variations of aerosols and trace gases 3.1.1 Temporal variability

The hourly wind speed varied within a wide rangeof 0.1–12.8 m s⁻¹, with an average of 4.99 ± 3 m s⁻¹(hereafter results are given as an average plus st and arddeviation)at SACOL from 25 April to 15 June2007. Statistical analyses of the mixing ratios for tracegases and aerosol optical properties measured duringthe fieldwork are summarized in Table 2. The averagemixing ratios were 694 ± 486, 6.2 ± 10, and 6.8 ± 3.3ppb for CO, SO₂, and NO_x, respectively. Generally, the mixing ratios for SO₂ and NO_x were much lowerthan those observed in northeastern China because ofhigh levels of anthropogenic pollution(Wang et al., 2004; Sun et al., 2005; Li et al., 2007a; Yuan et al., 2008). For example, Wang et al.(2004)found that theaverage mixing ratios for CO, SO₂, and NO_x at Lin’anwere 677 ± 315, 15.9 ± 14.6, and 13.8 ± 7.2 ppb, respectively(NO_x was compared with NO_y in Wang et al., 2004). The maximum values of CO, SO₂, and NO_xwere 2098, 82.5, and 59.5 ppb at rural sites, which werepossibly influenced by local or regional anthropogenicair pollutants. Similar results were also found at aremote observation site at Zhangye in 2008; however, the levels of CO, SO₂, and NO_x were 12%, 14%, and 16% higher at Zhangye than those at SACOL(Li et al., 2010).

Aerosol optical properties can also be used to determinelocal and regional anthropogenic air pollutants.We found a very low aerosol loading, with anaverage concentration of 172 ± 180 μg m⁻³ for PM₁₀during non-dust storm periods. In contrast, a highaerosol loading was observed during dust storms, as reportedin Section 3.2. During the field campaign, theaverage values of aerosol scattering coefficient(B_sp) and aerosol absorption coefficient(B_ap)were 164 ±89 and 11.7 ± 6.6 Mm⁻¹ at 520 nm. Li et al.(2010)reported similar results at the National Climate Observatoryof the China Meteorological Administration(39.082°N, 100.276°E; 1460 m a.s.l.), located nearly 20km northwest of Zhangye, Gansu Province. However, the B_sp(488 ± 370 Mm⁻¹) and B_ap(83 ± 40 Mm⁻¹)values were much lower than the mean values inBeijing(Bergin et al., 2001). Table 2 shows the finemode and coarse-mode aerosol optical depth(AOD)derived from the sun tracking photometer, whichranged from 0.20 ± 0.14 to 0.32 ± 0.49. The highercoarse mode AOD indicates that mineral dust or localsoil dust was dominant at SACOL during the springseason because of the strong cold fronts frequently arrivingfrom the north and northwest. We also notedthat anthropogenic pollutants, including BC, had animportant effect on the regional climate during thisperiod.

Figure 2 shows a time series of NO_x, NO, O₃, SO₂, and CO levels during this field campaign. Generally, high mixing ratios of NO_x and SO₂ occurred duringperiods dominated by local air pollution. The longtermtrends of the mixing ratios of NO and SO₂ decreased, while the mixing ratio of CO increased duringthe experiment. In addition to the changes in meteorologicalconditions(increased WS and RH, decreasedp), the emission sources of SO₂ and NO_x also changedfrom spring to summer. The mixing ratio of O₃ alsodisplayed a slightly increasing trend because of theincreases in surface temperature and solar radiation.The ratio was much higher at SACOL than in easternChina, such as Xianghe(Li et al., 2007a). During theshort period of this experiment, high CO/NO_x and CO/SO₂ ratios were detected due to biomass burningevents. Descriptive statistics for the major air pollutantsare shown in Table 3, and are further discussedin Section 3.3. The relatively high level of CO in mid May may be due to the agricultural activities in thisregion.

The average hourly mixing ratios of NO_x, NO, O₃, SO₂, and CO in Fig. 2 indicate that of all these tracegases, O₃ was the only pollutant displaying a diurnalcycle that was not sensitive to anthropogenic pollutants.We noted that the high surface temperaturesin late spring, combined with strong solar radiation atan altitude of 1965.8 m, may be conducive to enhancementin the photochemical production of O₃(Abdul-Wahab and Bouhamra, 2004; Elminir, 2005), whichwas likely to have originated primarily from reactionsinvolving volatile organic compounds and oxides of nitrogen(Fowler et al., 2009). As shown in Fig. 2c, theaverage O₃ concentration at SACOL was 50.7 ± 13.1ppb during this experiment, with an hourly maximumof 86.75 ppb. Similar results for O₃(48.5 ± 15.4 ppb)were also found at Zhangye during a field campaignin 2008(Li et al., 2010). However, concentrations ofO₃ were much higher at Xianghe(Li et al., 2007a), indicatingmore moderate photochemical O₃ productionprocesses at SACOL.

The hourly averaged concentrations of PM₁₀, B_sp, B_ap, aerosol single scattering albedo(SSA)at 520 nm, and AOD at 500 nm are shown in Fig. 3. The most obviousfeature is the generally decreasing trend in PM₁₀levels. The concentration of PM₁₀ at SACOL waseven lower in May than in March and April 2007(figuresomitted). Because dust plumes mainly occur duringspringtime, especially March and April(Littmann, 1991; Wang et al., 2008), we noted that the decreasingtrend of PM₁₀ was closely correlated with dust storms and local air pollutants due to the strong winds inspring over northwestern China. The hourly averageconcentration of PM₁₀(Fig. 3a)was 168.0 μg m⁻³, with a peak value of 2484.7 μg m⁻³. B_sp was closelycorrelated with dust storms and locally suspended soildust during the study period(Fig. 3b). The temporalvariation of B_sp at 520 nm was in the range of 41.76–893.39 Mm⁻¹ during air pollution episodes(with B_spvalues obtained from 13 air pollution episodes). An experimentat Xianghe in 2005 at the same wavelengthrecorded 940 Mm⁻¹ at 550 nm during an air pollutionepisode. However, we noted that B_ap(11.7 ± 6.6 Mm⁻¹)at 520 nm reached its peak during the periodsdominated by local air pollution, and that the longtermB_ap values, which decreased slightly during theexperiments, may be two-fold as indicated in Fig. 3c.

In addition to the changes in meteorological conditions, emission source intensities also changed withthe season. For example, from late spring to early fallthere was less biomass burning than in spring. Furthermore, variations in B_ap values not only fluctuatedwith levels of local air pollutants, but also were affectedby dust storms because of the light absorptionby mineral dust. The most prominent feature of theair pollution measured at SACOL, as expressed in Fig. 3, was the very high concentration of pollutants duringa dust storm on 2 May 2007. The value of the scatteringcoefficients gradually increased with the PM₁₀concentration during the dust plume. The durationof the experiment was not sufficiently long for a seasonalcharacterization of the changes in air pollutionat SACOL. The AOD displayed a slight decreasingtrend in the range from 0.10 to 2.00, with the exceptionof some cases of higher AOD(> 1.00)during airpollution events in the last few days of this field campaign.Defined as the ratio between aerosol scattering and extinction(scattering+absorption), SSA at 520nm was an important parameter representing the opticalproperties of aerosols. During this experiment, the SSA displayed changes on a daily basis from 0.84to 0.98.

3.1.2. Diurnal cycles The diurnal cycles of regular meteorological parameters, aerosol optical properties, and mixing ratiosof trace gases are shown in Fig. 4. The mixing ratioof O₃ displayed a large diurnal variation, whichwas closely correlated with the NO_x concentration.The wind speed had three peak values at 0200, 1500, and 2100 LST(local st and ard time). The temperature and relative humidity displayed large diurnal changes, while there was less than 50% variation in the RHduring daylight time. Generally, the diurnal changesin the levels of primary pollutants displayed a similartrend to the low mixing ratios, with CO concentrationsof 566 ± 371 ppb, SO₂ concentrations of 3.2 ± 2.6 ppb, and NO_x concentrations of 5.1 ± 1.6 ppb in the after noon. The maximum values of CO(934 ± 699 ppb), SO₂(13.3 ± 19.3 ppb), and NO_x(9.6 ± 5.1 ppb)wererecorded around 0800 LST because of the low windspeed and anthropogenic emissions in the mornings.The changes in the levels of these gaseous pollutantsdiffered from the diurnal changes of the same pollutantsin eastern China(Li et al., 2007a, 2010).

The concentration of PM₁₀ displayed a significantdiurnal cycle during the experiment. Two comparablepeaks of daily PM₁₀(morning: 205 ± 156 μg m⁻³;evening: 191 ± 177 μg m⁻³)were found at 0800 and 2000 LST. These peaks were affected by the aerosolsoriginated from the local soil dust and air pollutants.The diurnal variation of B_sp was closely synchronizedwith B_ap and CO. The B_sp peak(219 ± 111 Mm⁻¹)was recorded at 0800 LST in the morning, which mayreflect the addition of anthropogenic air pollutants, given that the lowest B_sp(120 ± 55 Mm⁻¹)occurred at 1800 LST. The variations in B_ap were similar to B_sp(high: 16.3 ± 6.9 Mm⁻¹ and low: 7.9 ± 4.7 Mm⁻¹).The anthropogenic emissions at SACOL occurred inthe morning and were dispersed in the afternoon inresponse to air pollution and meteorological parameters, such as WS, RH, and T. The low wind speed(4.3 ± 2.6 m s⁻¹)at 0800 LST and the peak value(4.7± 2.7 m s⁻¹)at 1600 LST is likely to be associatedwith B_ap and B_sp. There was relatively little variationin the AOD, with only small changes in the daytime(lower than 0.04). However, the highest values of theAOD 90th percentile(ranging from 0.56 to 1.82)wererecorded during the presence of local air pollutants ordust storms during this experiment. These results areconsistent with previous studies(Wang et al., 2010; Bi et al., 2011; Che et al., 2014).

3.1.3. Back trajectory cluster analysis

To capture the possible sources of dust, local airpollution, and the regional air mass, a cluster analysiswas applied with a 60-h HYSPLIT back trajectory model(starting on 25 April 2007, and including 204total trajectories)at 1000 m AGL. It was used to capturethe transport path from the source locations toSACOL(Fig. 5). The total trajectories were dividedinto four clusters, which were considered a reasonable number. More than 10% of the trajectories were followedby each of the clusters. Of the total trajectoriesarriving at SACOL, 71.6% consisted of clusters 1, 2, and 3 from northern and northwestern China.

The pathways of these trajectories were mostlyfrom dust source regions, such as the Taklimakan(Li et al., 2007a; Wang et al., 2008), and Mongolian Gobideserts(Zhang et al., 2013). The pathways of thesetrajectories were determined by two major factors.First, the location of SACOL is surrounded by dustsources to the north, including the Badan Jaran, Taklimakan, Tengger, and Gobi deserts(Zhang et al., 2003c, d; Wang et al., 2008). Second, the WSWESE-trending Hexi Corridor runs along the northernedge of the Tibetan Plateau to the south of the TianshanMountains. The north and northwest pathwaysare prevalent during the spring season in this region, with the dust aerosol from the Taklimakan and Gobideserts suspended from the ground and transportedto remote regions(Huang et al., 2010; Wang et al., 2010). The movement of the two dust storms on 2 and 10 May 2007 during the experiment closely followedthese paths at 1000 m AGL(figures omitted).In contrast, approximately 28.4% of the descendingair flows from the southeast consisted of relativelyclean air masses with very low levels of trace gases and aerosols, although at SACOL there was some local airpollution. Therefore, the apparently high concentrationsof dust aerosol and local air pollutants may havedifferent compositions.

3.2 Properties of mineral dust during a dust event 3.2.1. Aerosol optical properties

Prior to the outbreak of one particular strong duststorm, a synoptic low-pressure system that had developedin the northeast of Mongolia generated high windspeeds over the region, including north-central China and Mongolia, and the wind direction subsequentlychanged from northwest to west(figure omitted). Duringthis dust storm, the wind speed increased sharplyto 8.8 m s⁻¹, and maintained at about 6.6 m s⁻¹ duringthe dust storm. The surface temperature rapidlydecreased from 22.5 to 8.2℃. However, the variation in RH was not sensitive to dust storms or local airpollution conditions, given that it related only to thewater content in the atmosphere.

The 5-min average variation of attenuatedbackscattering coefficients, PM₁₀, B_sp, B_ap, SSA at520 nm, and AOD at 500 nm were recorded duringthe strong dust storm on 2 May 2007(Fig. 6). Anthropogenicpollutants were observed near the l and surface before the dust storm occurred in the morningof May 2(Fig. 6a). The pollution layer was observedbelow 500 m AGL. When the dust storm occurred, thenormalized relative backscatter signal increased to 3, indicating that the atmospheric aerosol was dominatedby mineral dust. The top layer of the dust storm suddenlyincreased at an altitude of approximately 1.5km, three times the height of the local air pollutionlayer. The strongest return signal for the dust layerobserved by the MPL system occurred at around 1300LST. The cold front during the dust storm also resultedin regional rainfall at 1600 LST, leading to wetdeposition of mineral dust. After the rainfall, the signalsof the attenuated backscattering coefficients beganto weaken, and the dust storm had dissipated bymidnight.

It is known that mineral dust can serve as cloudcondensation nuclei(CCN)leading to rainfall(Huang et al., 2010). The most prominent feature was thatthe mass concentration of PM₁₀ increased to approximately4072 μg m⁻³, approximately 21-fold higherthan the value in non-dust-storm periods. The approachingcold front also delivered a period of heavyrainfall after the 1600 LST dust storm, which washedout the suspended particles from the atmosphere(Fig. 6b). The peak B_sp reached 1408 Mm⁻¹ because ofthe scattering effect of mineral dust closely related toPM₁₀. The correlation coefficient for the relation betweenB_sp and PM₁₀(R² = 0.86)was statistically significantat the 1% level. SSA also increased sharplyfrom the typical values of 0.92–0.94 to exceeding 0.97.It was interesting that the coarse mode AOD wasfound to increase gradually to 0.84 on arrival of thedust storm, whereas no significant changes were observedin the fine mode(range from 0.2 to 0.3). The5-min average concentrations of B_sp, B_ap, SSA at 520 nm, and AOD at 500 nm in the morning before thedust storm arrived were 160 Mm⁻¹, 14.5 Mm⁻¹, 0.92, and 0.39, compared to 341.3 Mm⁻¹, 11.0 Mm⁻¹, 0.97, and 0.87 during the dust storm, respectively(Figs.6c–f). The results were in good agreement with thosefrom experiments measuring aerosol properties and trace gases during the 2008 China–U.S. joint dust fieldexperiment(Li et al., 2010; Wang et al., 2010).

3.2.2. Ratios of dust aerosol mixed with anthropogenicpollutants

Most of the local air pollution episodes occurredin the morning at SACOL during the experiment, when the wind speed was lower than 3.0 m s⁻¹. Usinga detailed emission inventory developed for Asia, we identified the trace gases and dust aerosols thatcharacterize emissions from a dust plume(Fig. 7).The most prominent feature of air pollution at SACOLwas the close relation between PM₁₀ and B_sp duringthe dust plume, which was evident from the ratios of B_sp/PM₁₀. During the dust plume, the correlation coefficient(0.89)for the relation between B_sp and PM₁₀was much higher than in the non-dust plume period.However, during non-dust plume periods, the value ofB_sp was either higher or lower than that during dustplume periods(Fig. 7a). We conclude that the likelyreason for this is that the values of B_sp are related notonly to the dust aerosol, but also to local air pollutantssuch as sulfate and nitrate aerosols. The B_ap/PM₁₀ ratiosreflected the relative impact of sources that wererich in BC(Fig. 7b). The B_ap/PM₁₀ ratios takenfrom data at the rural background site before the dust storm were found to be much lower than the ratiosduring the dust storm. The observed high B_ap values and low PM₁₀ concentrations revealed that local finemodeparticles dominated at SACOL before the arrivalof the dust plume. The slope of B_ap/PM₁₀ before thedust storm was 0.001, compared to 0.008 during thedust storm. The lower B_ap/PM₁₀ ratio reflected thesudden increase in large mineral dust particles duringthe dust storm, compared with the high B_ap/PM₁₀ratio due to local air pollution. A correlation coefficientof R² = 0.83 for the relation between PM₁₀ and B_ap(aerosol absorption coefficient at 520 nm)was obtained, suggesting that the high B_ap/PM₁₀ ratio wasattributed to local emissions from nearby urban areas.This could be explained partly by the general increasein B_ap in the local region rather than the emission ofBC.

Generally, the anthropogenic trace gases originatedfrom different emission sources. For example, CO is an important indicator of combustion sources and has a relatively long tropospheric lifetime, whereasNO_x is produced by the combustion of diesel oil, gasoline, and coal. It is considered that SO₂ is mainly a product of burning coal, which contains sulfur. Theratios of CO/NO_x(Fig. 7c), compared with a ratioof 27.2 in Xianghe(Li et al., 2007a), were 45.4 and 58.2 in the dust and non-dust storm periods, implyinga large relative contribution from combustion sourcesacross the larger region. Thus, the dust plume fromthe Lanzhou area contributed to a lower CO/NO_x ratioin the wider region under ambient weather conditions.As shown in Figs. 7c and 7d, the correlationcoefficients between CO/NO_x and CO/SO₂ weremuch lower in the dust storm period, than during thenon-dust storm period. During these periods the emissionof NO_x was suppressed and the emission of COwas uninhibited. Accordingly, simultaneous measurementsof NO_x and CO could help to characterize theemission sources. Once emitted into the atmosphere, NO quickly reacts with other gases and is transformed.Thus, the ratio between CO and NO_x was used in thisstudy.

The correlation coefficient between CO and SO₂(Fig. 7d)was slightly lower than that between CO and NO_x. This might reflect the fact that CO and SO₂ have more emission sources in common than CO and NO_x. Efficient combustion activities are likely tobe important sources of SO₂ but not CO. The CO/SO₂ratio in regional signals is lower than that in localsignals, reflecting the difference in source composition.The CO/NO_x and CO/SO₂ ratios also helpedto identify a major agricultural burning event. Agriculturalburning is characterized by high CO/NO_x ratios(Streets et al., 2003b). The correlations betweenCO/NO_x and CO/SO₂ were high(R2 = 0.89 and R2= 0.8, respectively)before the dust storm, and withCO/NO_x and CO/SO₂ ratios of nearly 0.31 and 0.34, respectively, during the dust storm. The CO/NO_x and CO/SO₂ ratios were similar to the values estimatedfor biomass burning emissions(Crutzen and Andreae, 1990), implying that biomass burning indeed occurredin the study region.

3.3 Properties of aerosols and air pollutants in pollution episodes 3.3.1 Local anthropogenic pollutants During the experiment, we also detected severalanthropogenic emissions in the vicinity of SACOL. Generally, the local anthropogenic emissions were observedbetween 0700 and 1000 LST in the morning, and were coincident with the vertical profiles of attenuatedbackscattering coefficients measured with theMPL system(Fig. 8a). We noted that these pollutantscould significantly affect the measurements(Table 3).To identify the influence of anthropogenic emissionsources, a 5-min time series of B_ap(520 nm) and themixing ratios of NO_x, O₃, SO₂, and CO were observedat SACOL during one local-pollution episode on 7May. The similar characteristics of the vertical profilesof attenuated backscattering coefficients recordedduring the air pollution episode on 7 May indicatedthat the pollutant layer occurred at altitudes of 0–600m AGL. The signals of the attenuated backscatteringcoefficients were much lower than those recorded duringthe dust storm, confirming that fine particles weredominant in the local air pollution episode at SACOL.

The relatively high value of B_ap shown in Fig. 8bduring the local pollution episode was affected by thepresence of light-absorbing impurities such as BC and OC, which originated from the incomplete burning ofbiomass. The mixing ratios of NO_x(Fig. 8c) and SO₂(Fig. 8e)had wide ranges(approximately 10–20 and 30–40 ppb, respectively)during this period, indicatingthat these pollutants originated from coal-burningsources nearby, such as industrial factories. The mixingratio of O₃(Fig. 8d)decreased significantly duringthe period because of the photochemical production ofNO_x. However, the mixing ratios were not closely correlatedwith the increase in SO₂ and NO_x during thelocal pollution period. We found that CO originatedfrom other sources such as incomplete biomass burning.

Another local air pollution episode on 13 May alsohad similar characteristics as shown in Fig. 8. Thepeak of total AOD was approximately 0.49. The timeseries of AOD at 500 nm also shows the arrival of anthropogenicemissions at SACOL early in the morningof 13 May, a conclusion supported by a rise in the finemodeAOD from 0.12 to 0.26, compared with a minorfluctuation in the coarse-mode AOD during the daytime.The SSA also changed, rising from 0.88 to 0.97 and 0.87 to 0.98, with peak values of 0.97 at 0225 LST7 and 1735 LST 13 May, respectively(figure omitted).

3.3.2. Emission ratios of anthropogenic air pollutants

Different types of emission source often have distinctemission profiles. Therefore, the ratios amongCO, NO_x, and SO₂ could reveal the possible sourcesof these compounds at SACOL. In Fig. 9, to identifythe emission sources of the major air pollutantsat SACOL, 5-min average ratios of CO and NO_x, CO and SO₂, and NO_x and SO₂ during 13 days with pollutantemissions(circle)were compared with those withno pollution(triangle)from 0700 to 1000 LST in themorning, as shown in the scatter plots. The CO/NO_xratio(Fig. 9a)was –0.12 for the days with anthro pogenic pollution and 38.25 for the non-pollution days, implying a higher relative contribution from industrialsources across the wider region. The observedCO/NO_x and CO/SO₂ slopes were much lower thanthe NO_x/SO₂ slope on both anthropogenic pollution and non-pollution periods. The emission of NO_x wassuppressed and the emission of CO was uninhibitedduring non-pollution periods. The correlation coefficientfor the correlation between CO and SO₂(Fig. 9b)was similar to that between CO and NO_x. Thismight reflect the fact that NO_x and SO₂ have similaremission sources(e.g., coal burning, gasoline, and diesel combustion)on the days with anthropogenicpollution. These high-efficiency combustion activitiesare likely to be important sources of SO₂ and NO_xbut not CO. The CO/SO₂ ratios in Fig. 9b were alsovery high(approximately 9.9)during non-pollutionperiods, compared with a CO/SO₂ ratio on thedays with anthropogenic pollution of approximately–0.46, reflecting complex SO₂ emissions. TheCO/NO_x and CO/SO₂ ratios were similar to the valuespreviously estimated for biomass burning emissions(Wang et al., 2004; Li et al., 2007a, 2010), implyingthe presence of biomass burning emissions inthe location of this study. The concentrations of SO₂ and NO_x were strongly correlated on the days withanthropogenic pollution and during non-pollution periods(R2 = 0.86 and R2 = 0.78, respectively), with anSO₂/NO_x ratio of approximately 0.21 during periodsof anthropogenic pollution(Fig. 9c).

Comparing the observed pollutant ratios to theestimated ratios from inventory studies provides anevaluation of the inventories. Table 3 lists the ratios ofCO/SO₂, CO/NO_x, and NO_x/SO₂ during eight heavyanthropogenic pollution episodes and a case study duringa dust storm. All of the results shown in Table 3 were significant at the 5% level. The observedCO/SO₂, CO/NO_x, and NO_x/SO₂ ratios during airpollution episodes in this experiment were 4.2–18.3, 13.7–80.5, and 0.15–1.24, respectively. Similar ratiosobtained from the industrial sources estimated in the2006 INTEX-B emission inventory(Zhang et al., 2009)were 43.7–71.9, 23.7–25.7, and 1.84–2.79. In order toreflect the characteristics of air pollution for the larger region, pollutant ratios from this experiment werecompared to results from other studies. CO has amuch longer lifetime(approximately 60 days)thanSO₂ and NO_x(approximately 2–3 days), making theobserved ratios higher than the actual emissions. However, the effects of chemical reactions were likely smallfor this study, since plumes from some major emissionsources could reach SACOL in just a few hours.There was also an inventory study covering East Asia and a high-resolution inventory for Xianghe in easternChina. The inventory study covered a relatively largeregion, while measurements often experienced the str ongest signals from nearby sources. This is evidentin the differences between the two experimental studies;for example, a large number of modern mobilesources in downtown Beijing emitted both CO and NO_x, but with a lower CO/NO_x ratio than residentialsources(Wang et al., 2005). It is also possible that SO₂emissions were overestimated, because the NO_x/SO₂ratio from this experiment(approximately 1.18)wasslightly higher than the inventories of 0.78(Streets et al., 2003b) and 0.89(Wang et al., 2005). It must benoted that these measurements were obtained at twosites that are only 70 km apart and thus characterizedthe emissions for only a relatively small region(Beijing, Xianghe, and surrounding areas). Nevertheless, observational studies at another rural site in easternChina(Wang et al., 2002, 2004) and over the westernPacific(Carmichael et al., 2003a, b)also suggestedhigher CO emissions from China than those estimatedin inventories. Future intensive field campaigns as wellas direct emission factor measurements are essential toreduce the uncertainties in the differences revealed bythis comparison.

4. Conclusions

In the field experiment at SACOL, we found avery low aerosol loading, with an average concentrationof 172 ± 180 μg m⁻³ for PM₁₀ during non-duststorm periods, and with the average values of B_sp and B_ap of 164 ± 89 and 11.7 ± 6.6 Mm⁻¹ at 520 nm. Theaveraged mixing ratios of CO(694 ± 486 ppb), SO₂(6.2 ± 10 ppb), NO_x(6.8 ± 3.3 ppb), and O₃(50.7 ±13.1 ppb)were lower than those observed in urban areas, but comparable to previous studies in Zhangye, arural area near the Gobi desert in northwestern China.

In the initial stage of a heavy dust outbreakoriginating from the remote dust sources, local anthropogenicpollutants were dominant at SACOL.When the dust storm arrived, the 5-min averagedmass concentration of PM₁₀ sharply increased up toapproximately 4072 μg m⁻³, approximately 21-foldhigher than that in non-dust-storm days. The lowerB_ap/PM₁₀ reflected the sudden increase in large mineraldust particles during the dust storm. During thepollution episodes, compared with the high B_ap/PM₁₀ratio due to urban air pollution during the dust event.During the pollution episodes, compared with dustevents, the mixing ratios of air pollutants were muchhigher. For instance, the mixing ratios of NO_x and SO₂ showed wide ranges of approximately 10–20 and 30–40 ppb during anthropogenic emissions, indicatingthat these pollutants originated from coal-burningsources nearby, such as industrial factories. Our resultsalso indicate that there were other types of emissionsource, such as coal burning, gasoline and dieseloil on anthropogenic pollution days at SACOL.

As the sources of air pollution at SACOL are complicatedin late spring, we note that further field experimentsfor measuring dust events reacting with urbanpollutants near the dust sources need to be performed.Our findings could also be used to improve model simulations.

Acknowledgments. The authors thank theAERONET website(http://www.arl.noaa.gov/ready.html)for providing real-time AOD datasets. The authors also thank the NOAA Air Resources Laboratoryfor provision of the HYSPLIT transport and dispersion model and the READY website(http://www.arl.noaa.gov/ready.html)used in thispublication.