Anthropogenic air pollution observed near dust source regions in northwestern China during springtime 2008



[1] Trace gases and aerosols were measured in Zhangye (39.082°N, 100.276°E, 1460 m a.s.l.), a rural site near the Gobi deserts in northwestern China during spring 2008. Primary trace gases (CO: 265 ppb; SO2: 3.4 ppb; NOy*: 4.2 ppb; hereafter results given as means of hourly data) in the area were lower than in eastern China, but still indicative of marked anthropogenic emissions. Sizable aerosol mass concentration (153 μg/m3) and light scattering (159 Mm−1 at 500 nm) were largely attributable to dust emissions, and aerosol light absorption (10.3 Mm−1 at 500 nm) was dominated by anthropogenic pollution. Distinct diurnal variations in meteorology and pollution were induced by the local valley terrain. Strong daytime northwest valley wind cleaned out pollution and was replaced by southeast mountain wind that allowed pollutants to build up overnight. In the afternoon, aerosols had single scattering albedo (SSA, 500 nm) of 0.95 and were mainly of supermicron particles, presumably dust, while at night smaller particles and SSA of 0.89–0.91 were related to pollution. The diverse local emission sources were characterized: the CO/SO2, CO/NOy, NOy/SO2 (by moles), and BC/CO (by mass) ratios for small point sources such as factories were 24.6–54.2, 25.8–35.9, 0.79–1.31, and 4.1–6.1 × 10−3, respectively, compared to the corresponding inventory ratios of 43.7–71.9, 23.7–25.7, 1.84–2.79, and 3.4–4.0 × 10−3 for the industrial sector in the area. The mixing between dust and pollution can be ubiquitous in this region. During a dust storm shown as an example, pollutants were observed to mix with dust, causing discernible changes in both SSA and aerosol size distribution. Further interaction between dust and pollutants during transport may modify the properties of dust particles that are critical for their large-scale impact on radiation, clouds, and global biogeochemical cycles.

1. Introduction

[2] With the Taklimakan and Gobi deserts sprawling over massive areas in China and Mongolia, East Asia is one of the major source regions of airborne dust in the world [e.g., Ginoux et al., 2001; Prospero et al., 2002; Tanaka and Chiba, 2006]. Driven by strong surface winds associated with the midlatitude wave cyclones [e.g., Kurosaki and Mikami, 2003; Qian et al., 2002; Shao and Dong, 2006; Sun et al., 2001; Zhu et al., 2008], the dust storms in East Asia occur most frequently in spring [e.g., Yue et al., 2009]. They are often of synoptic scale [e.g., Hsu et al., 2006; Wang et al., 2007], and influence large areas from near the deserts [e.g., Cheng et al., 2006; Fu et al., 2008] to the west coast of the Pacific [e.g., Fang et al., 1999; Lee et al., 2006; Liu et al., 2009; Sun et al., 2005]. The long-range transport of Asian dust [e.g., Duce et al., 1980], in some cases across the Pacific, has also been well documented [e.g., Huang et al., 2008; Husar et al., 2001; Leaitch et al., 2009; McKendry et al., 2008; Zhao et al., 2008]. Besides modulating macroscale radiative energy [e.g., Cheng et al., 2006; Kim et al., 2005], and modifying microscale cloud properties [e.g., Huang et al., 2006a, 2006b], Asian dust may also play an important role in the global biogeochemical cycles, as a potential source of iron to the marine ecosystems of the North Pacific [e.g., Bishop et al., 2002; Gao et al., 1997; Hsu et al., 2009; Zhang and Gao, 2007].

[3] A key factor determining the large-scale impact of Asian dust is its interaction with various other atmospheric constituents during transport, which for example may mobilize iron in dust particles and make it available to the marine biota [Meskhidze et al., 2005; Solmon et al., 2009]. Black carbon may coat the surface of dust particles and increase their absorption of radiation [e.g., Chaudhry et al., 2007; Guo et al., 2010]. Introduction of compounds such as sulfate and nitrate to dust aerosols can also change their size distribution, optical and hygroscopic properties [e.g., Carrico et al., 2003; Kelly et al., 2007; Sullivan et al., 2009; Tobo et al., 2009; Zhang and Iwasaka, 2004], and their lifetime in the atmosphere. Numerical [e.g., Bian and Zender, 2003; Liao et al., 2003; Tang et al., 2004; Zhang et al., 1994; Zhang and Carmichael, 1999] and laboratory [e.g., Laskin et al., 2005] studies suggest the possibility of heterogeneous reactions taking place on the surface of dust particles, and a number of field experiments have been conducted to provide observational evidence [e.g., Huebert et al., 2003]. Some studies in the industrialized eastern China [e.g., Huang et al., 2010a, 2010b; Sun et al., 2005, 2010; Yuan et al., 2006, 2008; Wang et al., 2007] uncovered that dust and pollution were often mixed coming into the region, while others indicated the two could be separated by frontal systems [e.g., Guo et al., 2004; Zhang et al., 2005]. Traces of chemical modification of dust during transport were found in most experiments farther downwind, off the Chinese coast [e.g., Arimoto et al., 2006; Gao et al., 2007; Geng et al., 2009; Leaitch et al., 2009; Sullivan et al., 2007; Zhang and Iwasaka, 2001]. The extent of the modification varies, and could critically depend on the regions the dust plumes travel through, as well as the duration since the mixing occurred between dust and pollutants [e.g., Maxwell-Meier et al., 2004; Ro et al., 2005; Song et al., 2005]. The longer the dust and pollutants stay together, the more thoroughly the composition and properties of dust particles may evolve.

[4] Frequented by dust storms originating from the Taklimakan and Gobi deserts, northwestern China including the arid provinces of Shanxi, Gansu, Ningxia, Xinjiang, and Qinghai (and the western part of Inner Mongolia) is a main source region of airborne Asian dust [e.g., Laurent et al., 2006; Liu et al., 2004; Shao et al., 2003; Shao and Dong, 2006; Xuan et al., 2000]. Industry and population in this region experienced substantial growth over the past decade [National Bureau of Statistics of China, 2008], leading to large changes in pollutant emissions. For instance, the estimated NOx emissions from Gansu went up from 195 Gg/yr (109 g) in 2000 [Streets et al., 2003] to 323 Gg/yr in 2006 [Zhang et al., 2009], whereas the estimated SO2 emissions dropped from 439 Gg/yr to 338 Gg/yr. The differences at least partly reflect the actual changes in emissions. The influences of anthropogenic pollution on aerosol composition have been identified in several studies in the region [e.g., Arimoto et al., 2004; Cao et al., 2005; Li et al., 2008a, 2008b; Shen et al., 2009; Xu et al., 2004; Zhang et al., 2003a], but the measurements of aerosol precursor gases are scarce, made mainly in cities and reflect chemical and physical changes that occurred during transport over a variety of environments [e.g., Chu et al., 2008; Ta et al., 2005].

[5] Following a pilot field experiment in 2005 in China under the Eastern Asian Study of Tropospheric Aerosols, an International Regional Experiment (EAST-AIRE) [Z. Li et al., 2007], a more intensive field campaign was conducted in 2008 with multiple extensive observation stations across China, as a part of the joint U.S.-China research endeavor for deploying the Department of Energy's Atmospheric Radiation Measurement (ARM) Mobile Facility (AMF) (Z. Li et al., Overview of the East Asian Study of Tropospheric Aerosols and Impact on Regional Climate (EAST-AIRC), submitted to Journal of Geophysical Research, 2010). A suite of surface-based remote sensing and in situ instruments were deployed at the ARM Ancillary Facility consisting of two mobile observatories: SMART (Surface-sensing Measurements of Atmospheric Radiative Transfer) and COMMIT (Chemical, Optical, and Microphysical Measurements of In situ Troposphere), established near the Gobi deserts to study the properties of regional aerosols and their potential climate impact. To our knowledge, this marks the first time such comprehensive observations were made in northwestern China where the dust is generated. In this paper, we analyze the data of trace gases and aerosol properties from the in situ measurements, focusing on the characteristics of air pollution in the area, and its interaction with dust. By exploring a representative dust event observed during the experiment, we demonstrate how anthropogenic pollutants mix with dust, changing the bulk properties of aerosols.

2. Measurements and Ancillary Data

[6] The experiment was carried out from 17 April to 18 June 2008, or day of year (DOY) 108–170, at the Zhangye National Climate Observatory (39.082°N, 100.276°E, 1460 m above sea level) of the China Meteorological Administration, ∼20 km northwest of Zhangye, a city of ∼300,000 residents in China's northwestern province of Gansu. With the Qilian Mountains (elevation: ∼4000 m) to the south and west, and the Beishan Mountains (elevation: ∼2000 m) and the Alashan Plateau to the north, Zhangye is one of the several oasis cities scattered along a narrow northwest-southeast-oriented valley stretching ∼1000 km in northwestern China, known as the Hexi Corridor (corridor to the west of the Yellow River). The Badain Jaran Desert, part of the Gobi deserts, is merely 100 km to the north, rendering the Hexi Corridor a hot spot of dust storms [e.g., Shao and Dong, 2006; Sun et al., 2001]. Airborne dust may also come from the sandy land and the cultivated fields within the valley in April and early May, when little vegetation cover exists prior to the growing season [Zou and Zhai, 2004]. The area was dry with a mild temperature throughout the experiment (relative humidity, RH: 28.1 ± 16.2%, temperature: 16.2 ± 7.0°C, mean ± standard deviation).

[7] During the deployment, the SMART-COMMIT observatories were at least 500 m away from any significant anthropogenic emission sources (Figure 1a). The nearest main roadway, the two-lane National Route 312, runs northwest–southeast about 700 m to the north. Several factories were spotted along the road by the authors, including a few sizable ones near Linze, the local township ∼7 km to the northwest. A major four-lane highway running east-west is about 8 km to the south. A number of rural residences are scattered around the site, between the two roadways. Overall, the Hexi Corridor is a relatively populated and polluted region in northwestern China, as evidenced by the estimated CO emissions [Zhang et al., 2009] shown in Figure 1b.

Figure 1.

(a) The area surrounding the Zhangye site marked with a star. (b) The estimated anthropogenic CO emissions in 2006 for part of China [Zhang et al., 2009]. The Hexi Corridor is evident in the map as a band of relatively strong emissions extending from (40°N, 96°E) to (36°N, 102°E).

2.1. Trace Gases

[8] All trace gas data were acquired from commercial analyzers manufactured by Thermo Environmental Instruments (TEI), Franklin, MA. A modified [Dickerson and Delany, 1988] TEI Model 48C with a detection limit of 35 ppb (signal-to-noise ratio, S:N = 2:1 for ±1σ noise) for 1 min data was used to measure CO. SO2 levels were determined with a modified [Luke, 1997] TEI Model 43C, which had a detection limit of 0.8 ppb at 1 min resolution (S:N = 2:1 for ±1σ noise). To measure O3, we employed a TEI Model 49C, which was calibrated with an in-house primary standard (TEI Model 49 PS) before the experiment and with a transfer standard (TEI Model 146C) afterward. The detection limit was estimated at better than 1 ppb for 1 min averaging time (S:N = 2:1 for ±1σ noise). A chemiluminescence detector (TEI Model 42C) was used to measure NO, with a detection limit (S:N = 2:1 for ±1σ noise) of ∼0.2 ppb for 1 min data. The TEI 42C also has a hot molybdenum converter to reduce NO2 prior to detection as NO, and the efficiency was determined by gas-phase titration [e.g., Dickerson et al., 1984]. Such detectors are known to be sensitive to other oxides of nitrogen such as PAN and alkyl nitrates, while HNO3 is generally lost on the inlet [Dunlea et al., 2007; Fehsenfeld et al., 1987; Poulida et al., 1994]. NOx is defined as NO + NO2. NOy is defined as NOx + NO3 + 2 × N2O5 + HONO + HNO3 + PAN + RONO2 + RONO + NO3. NOz is defined as NOy − NOx [Seinfeld and Pandis, 1998]. Dunlea et al. [2007] indicate that commercial NOx monitors detect NOz with a net efficiency of about 50%, due to the loss of HNO3 gas to the surface of sampling inlets. Thus the reported nominal NOx values are in fact an upper limit for NOx, but a lower limit for NOy, and we will refer to them as NOy*, which are used throughout this paper to qualitatively depict the pollution pattern in the area. For a fraction of the NOy* data acquired under strong influences from local emission sources (spikes in pollutant levels in the morning; see section 3.3), we can make corrections to the NOy* data to estimate NOy concentration, which is used in section 3.3 to characterize those local sources. The correction method is described in section 2.2.

[9] All the trace gas detectors were housed in the COMMIT with temperature control. Sample flows were drawn from the top of the trailer (∼4 m above ground) into the instruments through a 3 m long Teflon sampling line. Daily zero checks of all gas instruments were conducted by introducing scrubbed air (TEI Model 146C). In addition, the instrument backgrounds of the CO and SO2 detectors were determined every few hours. Before, after, and every 3–4 weeks during the deployment, the gas instruments were calibrated with a working standard gas (50.5 ppm CO, 13.5 ppm SO2, and 10 ppm NO mixed in nitrogen, Scott-Marrin Inc., Riverside, California) traceable to NIST (National Institute of Standards and Technology) standard reference materials, diluted with the TEI Model 146C calibrator. Small nonsystematic changes in the derived calibration factors (within ∼15%) suggest that the calibration standards likely remained stable throughout the experiment. Data were stored at 1 min resolution and averaged into 5 min and hourly intervals for this study.

2.2. Correction of the NOy* Data

[10] For the morning data, we can estimate the partitioning between NO and NO2 by assuming the photostationary state [Leighton, 1961],

equation image

where k(NO+O3) is the rate constant of the reaction NO + O3 → NO2 + O2, [O3]ss is the steady state O3 concentration, and j(NO2) is the photolysis rate coefficient for NO2. Note that the oxidation of NO by HO2 is not considered here, and the actual NO2/NO ratio could be ∼10% greater than estimated with equation (1). We use two methods to calculate j(NO2), by linking our measured direct and diffuse ultraviolet irradiance for the band of 285–395 nm to j(NO2) with the relationship proposed by Madronich [1987], and by using an empirical equation given by Dickerson et al. [1982],

equation image

where θ is the solar zenith angle. For the first method, we assumed a relatively high albedo of 0.20 (compared to 0.15 used by Madronich [1987]) to represent the arid land surface near our site. And the resulted j(NO2) values are typically ∼10–15% greater than those derived from the second method or measured in Boulder, Colorado, a site at similar elevation [e.g., Shetter et al., 2003] for downwelling radiation only.

[11] From the NO concentration and the calculated NO2/NO ratio, we can then estimate the NOx concentration. When NOx is a large fraction of the NOy*, the instrument detects NOy with reasonable precision. In this case, we apply a correction factor of 2 (to account for the 50% detection efficiency) to NOz, calculated here as the difference between NOy* and NOx estimated using equation (1),

equation image

If for example NOx/NOy* is 0.70 then the correction gives a value for [NOy] equal to 1.3 × [NOy*]. This correction produces an uncertainty of ∼25% in the NOy values. The NOy data from the correction are used in section 3.3 to characterize pollutant spikes due to local emission sources.

2.3. Aerosols

[12] The mass concentration of the particulate matter with a diameter of 10 μm or less (PM10) was measured with a TEI Model 1400ab Tapered Element Oscillating Microbalance (TEOM) [Patashnick and Rupprecht, 1991]. The minimum detection limit of this instrument is typically 0.06 μg/m3 [Chow et al., 2008]. Loss of volatile and semivolatile aerosol compounds due to heating of the sampling stream (50°C) may lead to a negative bias, which should be negligible in the studied environment largely dominated by dust particles.

[13] An integrating nephelometer (TSI Model 3563, Shoreview, MN) measured particle scattering coefficient (Bsp) at wavelengths of 450, 550, and 700 nm; the detection limits of the instrument at the three wavelengths are 0.44, 0.17, and 0.26 Mm−1 (10−6 m−1), respectively (signal-to-noise ratio S:N = 2:1 [Anderson et al., 1996]). The bias related to the nonideal forward-scattering truncation [Anderson et al., 1996] was accounted for, following Anderson and Ogren [1998]. The RH reduction due to heat inside the integrating volume is not expected to significantly influence the measured aerosol scattering, since the growth in dust particle size is practically zero under the ambient humidity observed during the experiment.

[14] A seven-wavelength (370, 430, 470, 520, 565, 700, 950 nm) Aethalometer (Magee Scientific Model AE-31, Berkeley, California) monitored the change in transmittance as aerosol particles deposited to a filter tape. The multiple scattering of the filter material and the filter loading effects were accounted for, with the method proposed by Arnott et al. [2005], to derive the particle absorption coefficient (Bap). We applied their correcting factors at 520 nm obtained from ambient measurements [Arnott et al., 2005], and extrapolated them to the other wavelengths. Both Bap and Bsp measurements were interpolated to the wavelength of 500 nm. We also estimated black carbon (BC) concentration from Bap at 500 nm for local pollutant spikes (see section 3.3), assuming a specific absorption efficiency of 10 m2/g.

[15] An Aerodynamic Particle Sizer spectrometer (APS, TSI Model 3321) was used to measure the aerodynamic size distribution of particles between 0.5 and 20 μm in diameter, from 17 April until 11 May, before the instrument light source burned out. The performance of the instrument was periodically checked with NIST-traceable monodisperse standards (Duke Scientific, Palo Alto, California). Peters and Leith [2003] compared the same model of instrument to an impactor, noticing that the instrument had a low counting efficiency for the total number concentration, but gave a reasonable shape of aerosol size distribution. All aerosol data were available at 5 min and hourly intervals, and reported for volumes under standard conditions (1013.25 hPa and 20°C).

[16] Aerosol samples were pumped through a metal sampling stack (∼7 m above the ground level) and conductive tubing, before entering the aerosol instruments in COMMIT. The penetration efficiencies [e.g., Baron and Willeke, 2001], calculated assuming particle density of 1000 kg/m3, are 87% for 20 μm sized particles for the APS, 90% for 10 μm particles for the TEOM, 91% for 5 μm particles for the nephelometer, and 75% for 5 μm particles for the Aethalometer. The penetration efficiencies for particles smaller than 2.5 μm are all greater than 95%.

2.4. Meteorology

[17] A Vaisala Model WXT-510 (Helsinki, Finland) weather transmitter, mounted on top of the COMMIT trailer ∼4 m above the surface, logged ambient temperature, relative humidity, ambient pressure, wind speed, and wind direction every minute. Five minute and hourly averages were used in this study.

2.5. Emission Inventory

[18] The anthropogenic emissions of CO, SO2, and NOx in 2006 were estimated by Zhang et al. [2009] for Asian countries. The emission rates of 0.5° × 0.5° resolution were allocated into grid cells of 27 km × 27 km, using population and road network density as surrogates [Du, 2008; Fu et al., 2009]. Emissions in each cell from four sectors (power, industry, transportation, and residential) were further distributed into 12 vertical layers: layer 1 including the transportation and domestic sectors represents the area and mobile sources near the surface; layer 2 accounts for small power plants and industrial emissions; layer 3 and above contain emissions from large point sources. We did not extrapolate the estimated emissions for 2006 to 2008, the year of our experiment, as the focus of this study is on the characteristics of local emission sources. The total pollutant emissions from this rural area could grow significantly in two years, but the properties of different emission sectors (e.g., the CO/SO2 emission ratio for industrial and residential sources) are not expected to change as dramatically. In other words, there could be more factories in the area in 2008 than in 2006, but we assume that the technology and emission characteristics remain largely unchanged.

3. Analysis and Discussion

3.1. Levels of Trace Gases and Aerosols

[19] The time series plots of NO, NOy*, SO2, CO, and O3 at the Zhangye site are given in Figure 2 to illustrate the levels and variability of trace gases during the experiment. Statistics of gases and aerosols are summarized in Table 1. The concentrations of NOy* (4.2 ± 3.4 ppb, hereafter results given as mean ± standard deviation of hourly averaged data), SO2 (3.4 ± 4.8 ppb), and CO (265 ± 161 ppb) are considerably lower than those observed in eastern China. For example, at rural sites near Shanghai [Wang et al., 2004] and Beijing [C. Li et al., 2007], the springtime levels of NOy, SO2, and CO were found to be 13.8–26.0 ppb, 15.9–17.8 ppb, and 680–1090 ppb, respectively. The maxima of NOy*, SO2, and CO in Zhangye, on the other hand, were substantial at 23.1 ppb, 40.3 ppb, and 1287 ppb, implying the influences of local and regional anthropogenic emissions. In Lanzhou, another city in the Hexi Corridor and about 450 km to the southeast, the SO2 levels in April were about 0.5 mg/m3 or ∼200 ppb [Chu et al., 2008]. During our experiment in Zhangye, O3, a product of photochemical reactions in the atmosphere, was 48.5 ± 15.4 ppb with an hourly maximum of 83.0 ppb, suggesting moderate photochemical processes. Much of the temporal variation in the trace gases was controlled by the diurnal change in meteorological conditions, as discussed in section 3.2.

Figure 2.

Hourly average concentrations of (a) NO, (b) NOy*, (c) SO2, (d) CO, and (e) O3 at the Zhangye site from 17 April 17 to 18 June 2008.

Table 1. Statistics of Trace Gases and Aerosol Properties During the Experimenta
 MeanStandard DeviationMedian10th Percentile25th Percentile75th Percentile90th Percentile
  • a

    All aerosol data reported for volumes under 1013.25 hPa and 20°C.

CO (ppb)265161201141159313474
NO (ppb)0.410.600.
NOy* (ppb)
SO2 (ppb)
O3 (ppb)48.515.451.425.537.260.266.0
Bsp (Mm−1)1591911054163196293
Bap (Mm−1)
PM10 (μg/m3)153230954563160264

[20] Aerosol loadings during the Zhangye experiment were high and variable, likely a result of strong and episodic dust emissions in the region (Figure 3). The hourly PM10 was 153 ± 230 μg/m3, comparable to the 145 μg/m3 mean level observed in a rural location near Beijing from January to March [Chaudhry et al., 2007]. A spring field campaign around a sandy area, north of Beijing, found average PM10 levels between 226 and 522 μg/m3 [Cheng et al., 2005]. Some thirty aerosol samples collected in Dunhuang, 500 km to the northwest of Zhangye, yielded an average TSP (total suspended particulate) concentration of 317 μg/m3 in spring 2001 [Zhang et al., 2003b]. The average PM10 was 158 μg/m3 during an almost yearlong experiment in Dunhuang 3 years later [Yan, 2007].

Figure 3.

Same as Figure 2 but for (a) PM10, (b) aerosol absorption coefficient at 500 nm, (c) aerosol scattering coefficient at 500 nm, and (d) aerosol size distribution (unit: cm−3).

[21] For this experiment, the hourly aerosol scattering coefficient (Bsp) at 500 nm was 159 ± 191 Mm−1. The large standard deviations in PM10 and Bsp data are likely due to episodic strong dust events. The aerosol absorption coefficient (Bap, 500 nm) was less variable, averaged at 10.3 ± 9.1 Mm−1, with a maximum of 61.1 Mm−1. The comparison between different studies can be complicated, particularly for measurements of aerosol absorption and instruments with different size cut. However, it may still be meaningful considering that fine particles (PM2.5 or smaller) likely dominate aerosol extinction in most cases. In Dunhuang Bsp was ∼150 Mm−1 from April to June in 2004 [Yan, 2007]. At another desert site in Yulin, roughly midway in between Beijing and Zhangye, Xu et al. [2004] noted that Bsp was 158 ± 193 Mm−1, to go along with a Bap of 6 ± 11 Mm−1, for PM2.5 in April. Bsp and Bap values obtained from eastern China were generally greater: Bsp of 353 Mm−1 and Bap of 23 Mm−1 for PM2.5 were logged in a rural location around Shanghai [Xu et al., 2002]; similarly high Bsp (468 Mm−1) and Bap (65 Mm−1) for bulk aerosols were recorded near Beijing [C. Li et al., 2007]. A recent study coordinating observations at 14 locations in China [Zhang et al., 2008] also suggested stronger aerosol absorption in urban and suburban areas in eastern China.

[22] Several dust events, including three intense ones in early May (2–3, 8–9, and 11 May, or DOY 123–124, 129–130, and 132), stand out in the time series of PM10, Bsp, and aerosol size distribution (Figures 3a, 3c, and 3d). During these dust storms, the hourly PM10 reached well above 1000 μg/m3, and Bsp could exceed 1000 Mm−1. Meanwhile the peak in the aerosol number size distribution shifted from submicron to supermicron particles. The aerosol absorption coefficient, on the other hand, showed much smaller increases. The light absorption of dust particles at around 500 nm appears to be small compared to anthropogenic pollutants, as also suggested by previous studies near the dust source regions in China [e.g., Cheng et al., 2006; Xia et al., 2005; Xu et al., 2004].

3.2. Diurnal Variations

[23] A regular diurnal wind pattern was observed during the experiment (Figure 4): strong northwest wind prevailed in the daytime, from 1100 to 1900 LT (local time), followed by weak and variable wind mainly from south during 2000–2300 LT, and then southeasterly gradually gaining strength from midnight until 0900 LT the next morning, when northwest wind started to take over again. This phenomenon is probably explained by the valley terrain of the area. During daytime, solar heating of the northern slope of the Qilian Mountains warms up the air above, inducing upslope northerly wind, or the valley wind, which generally peaks near midday and dies down before sunset. At night, the radiative cooling of the mountain slope cools down the layer of air in contact with it. The cooler airflows northward down the slope, forming the mountain wind, which usually reaches maximum strength just before sunrise [Arya, 1999]. For this arid area temperature showed a large diurnal range and RH often remained below 50% all day (Figure 4).

Figure 4.

The diurnal variations in (first row, left to right) wind, CO, aerosol size distribution, (second row, from left to right) temperature, NOy*, aerosol absorption coefficient, (third row, from left to right) relative humidity, SO2, aerosol scattering coefficient, (fourth row, from left to right) PM10 level, O3, and aerosol single scattering albedo in Zhangye from 17 April to 18 June 2008 (17 April to 11 May for aerosol size distribution). Median values (solid lines) are shown to give a more distinct diurnal pattern than means, which could be influenced by a few extreme dust events. The dotted lines are the 25th (lower) and 75th (upper) percentiles for each hour of the day. The unit for aerosol size distribution is cm−3.

[24] Closely related to the local meteorology, trace gases and aerosols also demonstrated pronounced diurnal features (Figure 4). Afternoon levels of primary pollutant gases were almost always low, with CO of ∼150 ± 20 ppb, SO2 of ∼1.1 ± 1.0 ppb, and NOy* of ∼2.0 ± 2.0 ppb. This is unlike the eastern United States, where SO2 maxima are normally observed during daytime when vertical mixing brings emissions from upwind point sources to near the surface [e.g., Stehr et al., 2000]. The afternoon CO concentration is close to the lower free tropospheric level measured at Mount Waliguan, a nearby background site (36.29°N, 100.90°E, 3810 m above sea level, NOAA/GMD, webpage:, suggesting generally small influences of local anthropogenic pollution. Strong northwesterly wind from the Gobi deserts and vigorous daytime vertical mixing efficiently dissipate pollutants from the local sources. Light wind and suppressed vertical motion at night favor the accumulation of pollutants near the surface, and nighttime concentrations were generally greater (Figure 4). The maxima of CO (414 ± 164 ppb), SO2 (8.1 ± 6.5 ppb), and NOy* (7.3 ± 6.7 ppb) appeared at around 0800 LT, after hours of southeast wind carrying pollutants from the Zhangye city and the surrounding areas to the site. Increased human activities (more traffic and starting of business) and the downward mixing of emissions from small point sources (e.g., nearby factories) in the early morning may also contribute to the morning peaks in the primary pollutants. Smaller CO and NOy* peaks with little SO2 appearing at 2000–2300 LT likely reflect a mixture of sources near the surface. Emissions of mobile sources and biofuel combustion normally containing CO and NOx but little SO2 could be brought to the site, by southerly wind from the major highway and villages. In section 3.3, we further discuss the influences and characteristics of local emission sources. Unlike the primary gases, O3 had the maximum (63.5 ± 7.0 ppb) in the afternoon (Figure 4) when the photochemical processes are most active. The broad peak during daytime (1000–1800 LT), and the fast increase in the morning (0600–1000 LT) imply that the downward mixing from the residual layer might also be an important source of surface O3. Removed through dry deposition and gas titration, O3 reached its daily low (29.6 ± 12.8 ppb) just before dawn. The mountain wind prevented the formation of a very stable nocturnal boundary layer and O3 seldom dropped below 10 ppb.

[25] The diurnal variation in aerosol absorption coefficient (Bap) was well synchronized with CO (Figure 4). Two comparable peaks of Bap appearing at 0800 LT (17.8 ± 12.0 Mm−1) and 2100 LT (18.9 ± 13.2 Mm−1) were likely associated with anthropogenic emissions. The low in Bap (3.5 ± 1.8 Mm−1), a factor of 5 smaller than the maxima, was found at 1400 UTC when the influence of pollution was minimal. Aerosols (number concentration) comprising mainly submicron particles at night (Figure 4) were characteristic of anthropogenic emissions, although the number of supermicron particles was still substantial. With strong wind from the desert and possible local dust emissions, aerosols in the afternoon were dominated by supermicron particles, presumably dust (Figure 4). The diurnal cycles of PM10 and aerosol scattering (Bsp) were generally in phase with that of Bap but not as large. The daily maxima of aerosol mass (209 ± 463 μg/m3) and scattering (228 ± 307 Mm−1) were found at 0100 LT, probably skewed by the few dust storms mentioned earlier. Slightly smaller peaks in PM10 (morning: 194 ± 240 μg/m3; evening: 200 ± 413 μg/m3) and Bsp (morning: 202 ± 230 Mm−1; evening: 187 ± 294 Mm−1) in the morning and evening hours roughly doubled the daily lows in the afternoon (PM10: 101 ± 80 μg/m3; Bsp: 92 ± 72 Mm−1). Dust particles were likely present in the area throughout a day, responsible for a large fraction of PM10 and Bsp but a relatively small fraction of Bap. The latter mainly controlled by anthropogenic emissions thus demonstrated a more marked diurnal variation.

[26] Under the assumption that the shortwave light extinction is completely due to particles with diameters of 10 μm or less, we estimate the mass scattering efficiency for PM10 aerosols at ∼0.9 m2/g in the afternoon and ∼1.0 m2/g in the morning, comparable to the 1.05 m2/g observed in Dunhuang in spring [Yan, 2007]. The mass scattering efficiencies determined for PM2.5 aerosols in Yulin were 3.0 m2/g for pollution and 1.0 m2/g for dust [Xu et al., 2004]. Likewise, the mass absorption efficiency in the dust-dominant afternoon was only ∼0.03 m2/g, compared to the morning value of ∼0.1 m2/g for the dust-pollution mixture. Our observed diurnal change in the mass extinction efficiency likely mirrors the change in aerosol composition. The single scattering albedo (SSA, at 500 nm), defined as the ratio between aerosol scattering and extinction (scattering + absorption), also showed distinct difference between day and night (Figure 4). The peak SSA values (0.95 ± 0.02) for dust dominant particles in the afternoon were close to another field experiment (0.95 ± 0.05 [Xu et al., 2004]) and several remote sensing studies investigating the airborne dust in northern China [e.g., Cheng et al., 2006; Lee et al., 2007; Xia et al., 2005]. The daily low SSA (0.89–0.91) in Zhangye was still greater than in both urban (0.81 [Bergin et al., 2001]) and rural areas (0.81–0.85 [C. Li et al., 2007]) close to Beijing, probably due to dust particles at night. Another study conducted near Beijing by Yan et al. [2008] gave SSA of 0.88, but their data spanned over a year, including summer when aerosol scattering may be enhanced by hygroscopic growth and secondary aerosols.

[27] The wind roses in Figure 5 provide some further insights into the connection between meteorology and pollution. With wind mainly from southeast and wind speed over 4 m/s most of the time, morning (0600–0900 LT, Figure 5a) measurements probably represent emissions from both local sources and a relatively large area along the valley. Under more stagnant conditions, measurements in the evening hours (2000–2300 LT, Figure 5b) may be largely influenced by local emission sources. Consistent with Figure 4, NOy* (Figure 5c), SO2 (Figure 5d), and CO (Figure 5e) during the whole experiment were in general more abundant in southeast wind. PM10 (Figure 5f) and Bsp (Figure 5g) did not show a strong dependence on wind direction. Northwest wind may bring dust while southeast wind may transport pollutants. Bap, on the other hand, was greater in southeast wind and more closely associated with anthropogenic emissions.

Figure 5.

Wind roses for (a) morning (0600–0900 LT) and (b) evening (2000–2300 LT) hours; shade represents wind speed. Wind roses for all hours, with shade representing levels of (c) NOy*, (d) SO2, (e) CO, (f) PM10, (g) aerosol scattering coefficient, and (h) aerosol absorption coefficient. Date during strong dust episodes are excluded in Figures 5c–5h.

3.3. Characteristics of the Local Anthropogenic Emission Sources

[28] This experiment lends us a good opportunity to characterize the anthropogenic emission sources in a region that so far has been understudied. The diverse local emission sources near the site, on the other hand, can impact our measurements, as revealed in Figure 6, which shows simultaneous 5 min data of CO and SO2. The spread of the data points reflects a variety of source types encountered during the experiment. The upper edge with high CO/SO2 ratio probably represents less efficient combustion sources near the surface such as domestic emissions, whereas the lower edge having relatively low CO/SO2 may reflect more efficient coal burning from nearby small point sources, for example, boilers in factories. The inventory CO/SO2 emission ratios are also given in Figure 6, for layer 1 (residential and transportation; see section 2.5), layer 2 (small power plants and industry), and all (layers 1 and 2) sources within ∼100 km from the site, as estimated in the 2006 Intercontinental Chemical Transport Experiment−Phase B (INTEX-B) emission inventory [Zhang et al., 2009]. The measured CO/SO2 ratio seems to be in reasonable agreement with the inventory, as the majority of the paired measurements are bounded by the layer 1 and 2 inventory ratios.

Figure 6.

The scatterplot of 5 min CO-SO2 data in morning hours (0600–0900 LT) excluding the dust storms. The dash-dotted line stands for the 2006 INTEX-B inventory [Zhang et al., 2009] emission ratios between the corresponding gases for Layer 1 (residential and transportation sectors) sources within ∼100 km from the site. The dotted and dashed lines represent the inventory ratios for layer 2 (industry and power sectors) and all (layers 1 and 2) sources, respectively. An intercept of 150 ppb CO was assumed for the CO/SO2 inventory ratios to account for the daytime background CO level.

[29] The influence of local emission sources can also be identified as spikes in the 5 min time series plots. We looked for these local pollutant peaks in CO, SO2, NOy*, and Bap that satisfy the thresholds in amplitude (peak height) and sharpness (to avoid broad peaks). The choice of the threshold values was somewhat subjective, but could be adjusted to achieve the inclusion of the most prominent peaks and the exclusion of small features. After some iteration, ∼60–90 spikes picked out for each pollutant were further examined. Only those observed under relatively stable wind direction and showing maxima for at least two species were retained for additional analysis.

[30] A number of the identified pollutant peaks appeared between ∼0700 and ∼1100 LT, coincident with the onset of more vigorous vertical mixing in the atmosphere during daytime. As shown in the example in Figure 7, substantial SO2 and NO were observed along with these spikes, suggesting that they were fresh plumes originating from nearby coal-burning sources, such as small industrial boilers. Our estimated NOx/NOy* ratio for these plumes is generally >0.6, and the NOy* concentration can be used to derive NOy level with relatively small uncertainty (typically <30%). Assuming that these pollutant peaks reflect emissions from nearby sources superimposed on a more slowly evolving regional background, we can try to estimate both the background and the local contributions. The regional background concentrations were inferred by linearly interpolating measurements before and after the peak (see dotted lines in Figure 7). The determined background depends on the perception of the start and the end of a particular peak, but the uncertainty should be small for more distinct spikes. The contributions from the local sources to different species, ΔCO, ΔSO2, ΔNOy, and ΔBC, were then estimated as the difference between the inferred background and the ambient levels. The ratios for ΔCO/ΔSO2, ΔCO/ΔNOy, ΔNOy/ΔSO2, and ΔBC/ΔCO for each plume, obtained through linear correlation analysis as illustrated in Figure 8, should reflect the emission characteristics of the corresponding local sources. Table 2 summarizes the ratios derived from the morning pollutant spikes. Only results with R2 (correlation coefficient) ≥0.5 are included. We focus on results based on NOy values estimated using method 1 (see section 2.2). The ratios derived using methods 1 and 2 differ by less than 5% when both are available.

Figure 7.

Time series of (a) NOy (pluses, NO; solid circles, NOy*), (b) CO, (c) SO2, (d) aerosol absorption coefficient, and (e) O3 during a local pollution plume observed in the morning of 27 April 2008. Dotted lines are the baselines of NOy/NOy*, CO, and SO2, and aerosol absorption, inferred by linearly interpolating the concentrations before and after the peak of the plume. Wind was from ESE during the plume before turning to N at ∼0900 LT.

Figure 8.

Scatterplots of (a) ΔCO-ΔSO2, (b) ΔCO-ΔNOy, (c) ΔNOy-ΔSO2, and (d) ΔBC-ΔCO for the local pollution plume shown in Figure 7. ΔCO, ΔSO2, ΔNOy, and ΔBC were derived as the difference between the ambient and the inferred background (dotted lines in Figure 7) concentrations. The solid lines represent the results of linear correlation analysis, with intercepts forced to be zero. Bap data were converted into the black carbon (BC) concentration, assuming a mass absorption efficiency of 10 m2/g. Mass concentration of CO was calculated using simultaneously measured pressure and temperature.

Table 2. Results of Correlation Analysis for Local Pollutant Peaks Observed in the Morninga
Day of Year (LT)ΔCO/ΔSO2bR2ΔCO/ΔNOxbR2ΔNOx/ΔSO2bR2ΔBC/ΔCOcR2Wind direction
  • a

    Only results with R2 ≥ 0.5 are included.

  • b

    Unit is ppb/ppb.

  • c

    Unit is (μg/m3)/(μg/m3).

  • d

    The j(NO2) value were calculated using equation (2) used for NOy data.

  • e

    NOx/NOy* > 0.9; no correction was made.

114.32–114.44    1.10d0.94  SE
116.26–116.3927.70.8533.6e0.720.79e0.515.5 × 10−30.78SE
118.31–118.4041.10.9432.90.890.800.535.3 × 10−30.86ESE
119.26–119.3824.60.83      SE
126.41–126.49    0.810.935.5 × 10−30.64N
130.28–130.3654.20.74    6.1 × 10−30.88SE
143.30–143.3939.70.7625.80.651.520.834.2 × 10−30.82SE
167.34–167.3946.80.9435.9d0.951.31d0.994.1 × 10−30.69NW

[31] The ΔCO/ΔSO2 ratio (by moles) of the morning pollutant peaks ranges from 24.6 to 54.2, and the average industrial emission ratio estimated by Zhang et al. [2009] is 71.9 for the grid cell in which our site was located (39°N–39.5°N, 110°E–110.5°E), and 43.7 for the larger area (38°N–40°N, 99°E–101.5°E). The ΔCO/ΔNOy ratio has a small range between 25.8 and 35.9, greater than the emission inventory for the industry sector (25.7 and 23.7 for the local cell and the larger area, respectively). The ΔNOy/ΔSO2 ratio from the measurements at 0.79–1.52 compares lower than the industrial inventory ratios of 2.79 (local cell) and 1.84 (large area). The ΔNOy/ΔSO2 ratio near unity is likely a result of low-tech combustion of high sulfur coal. The BC/CO emission ratio (by mass) for the industry is estimated at 3.4–4.0 × 10−3 (local cell and the larger area) in the emission inventory [Zhang et al., 2009], compared to 4.1–6.1 × 10−3 for the morning peaks. The differences in CO/NOy and NOy/SO2 ratios between the inventory and measurements point to a possible overestimate of NOx emissions. There is however uncertainty associated with both measurements and emission inventory. Below we discuss the limitations of using our results to evaluate the inventory.

[32] As mentioned above, the analysis of individual plumes is subject to uncertainty in the inferred background concentration. We can further examine the fresh plumes in the early morning (0600–0900 LT), by selecting all the data with NOx/NOy* > 0.7, and NO > 0.2 ppb. A correlation analysis of these data points yields a CO/NOy ratio of 30 (Figure 9), and an intercept (116 ppb CO) close to the minimum CO level observed during daytime. The CO/NOy ratio is consistent with the ΔCO/ΔNOy ratio from the individual plumes, suggesting the robustness of the analysis. As another sensitivity test, we estimate the pollutant ratios for the morning plume peaks (Table 2) assuming fixed background levels (10th percentiles in Table 1, i.e., 141 ppb CO, 0.3 ppb SO2, 0.8 ppb NOy*, and 2.5 Mm−1 Bap). This results in generally consistent ranges of ΔCO/ΔSO2 (25.8–52.3 versus 24.6–54.2), ΔCO/ΔNOy* (21.3–44.2 versus 25.8–35.9), ΔNOy*/ΔSO2 (0.72–2.3 versus 0.79–1.31), and ΔBC/ΔCO (4.0–7.8 × 10−3 versus 4.1–6.1 × 10−3), compared to the previous method using background derived from measurements before and after a local plume. The ratios derived for some individual plumes can be significantly influenced by the selected background, particularly for those superimposed on heavy regional pollution, but the ranges of the ratios from all plumes are relatively insensitive.

Figure 9.

Scatterplot of CO and NOy for 5 min morning (0600–0900 LT) data points with NOx/NOy* > 0.7 and NO > 0.2 ppb. The solid line represents the result of linear correlation analysis.

[33] Besides the assumed background, our estimated ranges may also reflect one or more of the following factors: the sampling of the plumes for the analysis, changes in meteorology and transport, and the actual variations in emissions. Given the relatively small sample size, the derived pollutant ratios are uncertain and need be interpreted with caution. Most of the selected plumes are captured under prevailing southerly winds in the morning, and this may lead to a mismatch between our sampled plumes and the actual emission sources nearby. For comparison, the emission inventory is of coarse resolution and incorporates different industrial sources. Additional analysis involving numerical simulation of meteorological conditions and pollutant transport, in addition to more extensive measurements of ambient pollutants and emission sources, may be necessary before the emission inventory for the region can be evaluated with more confidence.

[34] Results from similar analysis on nighttime pollutant peaks are given in Table 3. As discussed in section 3.2, the O3 concentration was well above zero and NO remained low at night (mostly below 80 ppt), even for local peaks with elevated NOy* levels. ΔNOy* estimated for these peaks, however, may still be regarded as a close lower limit for ΔNOy, as it probably represents contributions from nearby sources that have experienced limited chemical aging at night. Local minima in O3 were also recorded for the majority of these nighttime spikes, further evidence that the plumes were relatively fresh. As expected, a variety of emissions sources were captured at night, with ΔCO/ΔSO2 (10.4–116.7), ΔCO/ΔNOy* (21.5–155.9), ΔNOy*/ΔSO2 (0.12–2.29), and ΔBC/ΔCO (2.4–12 × 10−3) ratios all having wide ranges. In comparison, the CO/NOx inventory emission ratio for the area can be over 100 for the residential sector, and less than 15 for the transportation sources, while the CO/SO2 inventory ratio for the same two sectors is near or over 100 [Zhang et al., 2009]. The inventory NOx/SO2 ratio is more than 60 for traffic emissions and 0.8–0.9 for residential sources. The BC/CO inventory ratio for these two sectors in the area is 7–20 × 10−3 [Zhang et al., 2009]. Both measurements and the emission inventory at least partly reveal the diverse nature of emission sources in the region.

Table 3. Results of Correlation Analysis for Local Pollutant Peaks Observed at Nighta
Day of Year (LT)ΔCO/ΔSO2bR2ΔCO/ΔNOx*bR2ΔNOx*/ΔSO2bR2ΔBC/ΔCOcR2Wind Direction
  • a

    Only results with R2 ≥ 0.5 are included.

  • b

    Unit is ppb/ppb.

  • c

    Unit is (μg/m3)/(μg/m3).

116.96–117.0210.40.9341.60.890.240.973.4 × 10−30.82NW
117.79–117.8622.60.89102.50.930.220.845.5 × 10−30.92NW
118.97–119.04    0.120.562.6 × 10−30.75SE
120.91–120.99  32.50.64    SE
123.04–123.1640.70.78155.90.84  2.4 × 10−30.83SE
125.96–126.02116.70.8175.60.841.540.93  SE
126.77–126.8573.70.8462.30.61  1.2 × 10−20.67N
149.01–149.1114.70.78      SE
152.04–152.1123.80.8381.40.74    E
157.83–157.94  101.00.66    NW
160.09–160.1449.20.9521.50.982.290.966.3 × 10−30.79NW
161.94–162.0445.70.8297.00.92    NW
164.91–165.0134.20.64      NW

3.4. Mixing of Dust and Anthropogenic Pollutants Near the Dust Source Region

[35] As discussed in section 3.2, the mixing between dust particles and anthropogenic pollutants likely occurred regularly at night, when anthropogenic emissions accumulated near the surface. On the other hand, strong episodic dust storms tend to carry dust into large downwind areas. These dust storms are often associated with synoptic or mesoscale disturbance, which would also influence the ambient levels of pollutants.

[36] On the evening of 2 May (DOY 123), a severe dust storm was brought to the Zhangye site by strong northerly wind (Figure 10, 1400–2100 LT). At its peak (1700–2100 LT), the dust storm possessed 5 min PM10 level around 4000 μg/m3 (Figure 10a), and the CO and SO2 concentrations (Figures 10b and 10c) near the low daytime levels. The O3 concentration (Figure 10d) stayed relatively high at ∼60 ppb well into the night, as strong wind dissipating near surface emissions slowed O3 titration. The aerosol single scattering albedo for the dust was about 0.98 (Figure 10e), similar to that measured for “clean” Asian dust particles [Anderson et al., 2003]. The aerosol size distribution was dominated by particles of 1–2 μm in aerodynamic diameter (Figure 10g).

Figure 10.

Time series of (a) PM10 level, (b) CO, (c) SO2, (d) O3, (e) single scattering albedo, and (f) wind vector during a major dust storm on 2–3 May 2008. (g) Average aerosol size distribution during four different periods (time given in the panel) of the dust storm. (h) Aerosol size distribution of fine particles (<2.5 μm) for the latter three periods.

[37] By midnight the strong northerly wind had diminished and been replaced by weaker southerly wind, but dust particles remained in the air, as evidenced by PM10 of ∼2000 μg/m3 (Figure 10a) and the abundant supermicron particles (Figure 10h). Under lower wind speed, a well-defined CO peak of almost 1000 ppb appeared between 0300 and 0600 LT (Figure 10b), as did SO2 of up to 8 ppb (Figure 10c). The NOy* concentration (not shown) also increased by about 2 ppb, but did not show a peak as evident. Good correlation between CO and SO2 (R2 = 0.54) and the CO/SO2 ratio of about 78 (intercept: 303 ppb) implied that the gases were from residential sources. If calcite (CaCO3) makes up 10% of the dust and has a concentration of ∼1 μmol/m3, the peak SO2 of ∼0.2–0.3 μmol/m3 is enough to neutralize 20–30% of the calcite, provided there is complete conversion to H2SO4. The decrease in O3 concentration at the same time was likely due to gas-phase titration, also suggesting the accumulation of pollutants near the surface. Accompanying the CO and SO2 peaks was a dip in SSA from 0.98 to 0.95 (Figure 10e), possibly due to BC from the same emission sources. The aerosols in the trace gas plume, although still mainly consisting of supermicron particles, also saw a substantial increase in submicron particles (Figure 10h). While the data set from this dust episode provided little evidence of the processing of trace gases (SO2 and NOy*) by dust particles, it clearly showed that the anthropogenic aerosols can mix with dust over its source region, modifying the bulk optical properties and size distribution of aerosols. The dust event extended well after the gas plume disappeared (Figure 10a). The observed mixing between dust and anthropogenic pollutants in the area occurred in relatively weak southerly wind, after strong northerly wind had died down. During daytime on 3 May, convective processes may facilitate the upward transport of dust particles, and relatively strong northwesterly winds (up to 6 m/s on the ground, not shown) could transport the dust aerosols mixed with anthropogenic pollutants to other areas. Furthermore, anthropogenic pollutants emitted at the night of 2 May, during the peak of the dust storm, although not detected due to strong ventilation by the northerly wind, likely would still mix with dust particles and can accompany the dust plume to downwind areas.

[38] Meteorological records and lidar measurements at different locations [Z. Huang et al., 2010] on 2 May indicated that this dust event may spread over other locations within the Hexi Corridor, from Dunhuang to the NW to Lanzhou to the SE, and may be of synoptic scale. The spatial coverage and the potential large-scale impact of this dust event will be further studied, with a synergic approach involving satellite data, aerosol model simulation, and radiative transfer modeling. While this particular dust storm is shown as an example here, peaks of gaseous pollutants were observed in most of the several identified dust events. Freshly emitted into the atmosphere over the dust source region, the trace gases and aerosols are likely externally mixed with dust particles. But as anthropogenic pollutants and dust travel together to downwind areas, further interaction is expected and may lead to modification of dust properties. Planning for air quality modeling is underway to simulate the evolution of pollutants and dust along the transport pathway.

4. Conclusions

[39] We have presented the measurements of CO, SO2, NOy*, O3, aerosol mass concentration (PM10), light scattering coefficient (Bsp), light absorption coefficient (Bap), and aerosol size distribution in Zhangye, a rural area near the Gobi deserts in northwestern China, from April to June 2008. The concentrations of primary trace gases for this dust source area (CO: 265 ± 161 ppb; SO2: 3.4 ± 4.8 ppb; NOy*: 4.2 ± 3.4 ppb) were not as high as observed in eastern China, but comparable to the eastern United States, and indicative of marked influence from anthropogenic emissions. PM10 and Bsp were substantial at 153 ± 230 μg/m3 and 159 ± 191 Mm−1, respectively, due to episodic dust emissions in the region. However, as expected near dust source regions, Bap was comparatively low at 10.3 ± 9.1 Mm−1.

[40] The regular alternation between strong northwest wind in the afternoon and relatively weak southeast wind before dawn was induced by the valley terrain of the area, leading to pronounced diurnal variations in trace gases and aerosols. Low afternoon trace gas levels (CO: ∼150 ± 20 ppb; SO2: ∼1.1 ± 1.0 ppb; NOy*: ∼2.0 ± 2.0 ppb) were associated with strong wind from the desert and active vertical mixing, while the peak gas concentrations (CO: 414 ± 164 ppb; SO2: 8.1 ± 6.5 ppb; NOy*: 7.3 ± 6.7 ppb) appeared in the morning, after overnight accumulation of surface emissions. Bap showed a pattern similar to that of the primary trace gases; the daily maxima (17.8 ± 12.0 Mm−1 at 0800 LT, 18.9 ± 13.2 Mm−1 at 2100 LT) related to anthropogenic emissions were much higher than the daily low in the afternoon (3.5 ± 1.8 Mm−1). PM10 and Bsp had less prominent diurnal changes: their peaks during a day (PM10, morning: 194 ± 240 μg/m3, evening: 200 ± 413 μg/m3; Bsp: morning: 202 ± 230 Mm−1; evening: 187 ± 294 Mm−1) were about double of the daily minima at 1400 LT (PM10: 101 ± 80 μg/m3; Bsp: 72 ± 55 Mm−1). Distinct diurnal variations in aerosol optical properties were observed as well. Dust particles influenced by little pollution had single scattering albedo of 0.95 ± 0.02 in the daytime, and the mixed dust-pollution aerosols featured SSA of 0.91 ± 0.04 at ∼0800 LT and 0.89 ± 0.06 at ∼2100 LT. The mass scattering efficiency changed from ∼0.9 m2/g in the afternoon to ∼1.0 m2/g in the morning, as aerosol size distribution shifted to submicron particles.

[41] Emissions from diverse local sources in the area were characterized using the ambient measurements during the experiment. In the morning, small coal-burning point sources such as nearby factories probably had a large impact, as suggested by several pollutant spikes. The measurement-derived CO/SO2, CO/NOy, NOy/SO2, and BC/CO ratios for these sources were 24.6–54.2, 25.8–35.9, 0.79–1.31, and 4.1–6.1 × 10−3, respectively. The same ratios for the industrial sources in the area, estimated in the 2006 INTEX-B emission inventory [Zhang et al., 2009], were 43.7–71.9, 23.7–25.7, 1.84–2.79, and 3.4–4.0 × 10−3. While the comparison may suggest a possible overestimate of NOx and uncertainties in other species, the results need to be interpreted with caution due to large uncertainty in the measurement-derived ratios and the emission inventory, as well as the sampling mismatch between the two. Local emission sources observed at night demonstrate more diverse properties: the CO/SO2, CO/NOy*, NOy*/SO2, and BC/CO ratios for nighttime pollutant peaks were 10.4–116.7, 21.5–155.9, 0.12–2.29, and 2.6–12 × 10−3, respectively. More field experiments are needed to better characterize the anthropogenic emission sources in the region, and to reduce uncertainties in the emission inventory.

[42] Besides the mixing between dust and pollutants regularly observed at night, we also looked into episodic regional dust storms that could potentially influence large areas. During one strong dust storm on 2–3 May, a nighttime pollutant plume with up to 1000 ppb CO and 8 ppb SO2, probably from residential sources, was observed at the Zhangye site. The anthropogenic aerosols emitted along with the trace gases mixed with dust particles, causing the SSA to drop from 0.98 to 0.95, and detectable change in aerosol size distribution. Elevated pollutant levels were also observed in several other dust events during the experiment. The mixing between pollution and dust could be ubiquitous over the area, where local and nearby dust emissions frequently encounter anthropogenic emissions. Initially, external mixing is more likely between fresh anthropogenic pollutants and dust. Once driven away from the source area, further interaction between pollutants and dust can possibly modify the optical and hygroscopic properties of the dust particles, key factors determining their effects on radiation, clouds, and biogeochemical cycles.

[43] In summary, our results indicate that substantial anthropogenic air pollutants exist in the Hexi Corridor and can mix with dust particles regularly at night or during strong dust episodes. The dust sometimes moves out of this source region already associated with pollutants, in some cases as much as several ppb of SO2 and ∼1 ppm of CO. The observed nontrivial amount of soot near the deserts may also enhance the light absorption by dust particles. Dust storms passing over cities (e.g., the Zhangye city) and industrial districts in the area may possess even higher levels of pollutants. Anthropogenic emissions in this important dust source region, bound to further increase as the economy and population continue to grow in the future, may play a crucial role in the chemistry and large-scale impact of Asian dust. For the uninhabited desert areas in NW China, the anthropogenic impact on dust composition is likely still small. Further study is necessary to estimate the fraction of the dust leaving northwestern China already mixed with pollutants, and our findings merit more extensive measurements in this region.


[44] We thank Jianrong Bi, Xianjie Cao, Jinsen Shi, Songtao Song, and Jiliang Xu of Lanzhou University for their help during the field experiment. We also thank two anonymous reviewers for their insightful comments that helped improve the original manuscript. This research was partially supported by MOST (2006CB403706), DOE (DEFG0208ER64571), and NASA (NNG04GE79G, NNX08AH71G).