The Gobi desert in northwest China is an important source of mineral aerosols over both eastern Asia and the northern Pacific Ocean. In order to determine the chemical, physical, and radiative properties of aerosols originating from the Gobi desert source region, field measurements were performed in Yulin, China, in April 2001 as part of the Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia) campaign. The means and standard deviations of the measured aerosol light absorption coefficient σap, scattering coefficient σsp, and single-scattering albedo ω are 6 Mm−1 (11 Mm−1), 158 Mm−1 (193 Mm−1), and 0.95 (0.05), respectively. A clear diurnal pattern is observed in both σap and σsp, resulting from diurnal changes in the mixing height as well as from local combustion sources in the morning and dust sources in the afternoon. Two distinct populations of aerosol mass scattering efficiencies Escat_2.5, one for aerosols dominated by desert dust (∼1.0 m2 g−1) and the other for aerosols composed primarily of local pollutants (∼3.0 m2 g−1), are observed. During the field study there were three significant dust events that occurred for, on average, several days at a time. The most significant dust storm resulted in a 24-hour-average PM2.5 concentration (mass concentration of particles having aerodynamic diameters less than 2.5 μm) of 453 μg m−3 and a peak σsp of 2510 Mm−1 on 8 April. The mean PM2.5 mass concentration during the dust storm periods is approximately 169 μg m−3, about 4 times greater than the mean value of 44 μg m−3 observed during local pollution periods. When local pollution is the dominant source of fine particulate mass, organic matter (OM) is the major chemical component, contributing 41% to the PM2.5 mass, followed by crustal material (29%), sulfate (17%), and elemental carbon (EC) (13%). During sand storm periods, ∼51% of PM2.5 mass is crustal material, followed by CO32− (11%) and OM (9.5%). The element enrichment factors indicate that coal combustion, biomass burning, and mobile source emissions are important local pollution sources. Overall, our results indicate that in addition to dust, local pollution also has a significant influence on aerosol properties in the region.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Atmospheric mineral dust is believed to have a significant influence on the radiation balance of the Earth and hence climate [Duce, 1995; Li et al., 1996; Sokolik et al., 2001]. It can both scatter sunlight back to space (leading to a cooling effect) and absorb solar and infrared radiation (leading to a warming effect) [Sokolik and Toon, 1999]. Mineral dust can also influence the microphysical properties of clouds and indirectly perturb the Earth's energy balance [Schwartz, 1996]. The sign and magnitude of the climate forcing of dust aerosols depends on the aerosol loading as well as chemical, physical and radiative properties that are largely controlled by the dust source [Xuan and Sokolik, 2002; Xu et al., 2003].
 Aerosol samples collected from the northwest China desert region (Minqin), a coastal suburb (Qingdao) and the interior of the Yellow Sea (Qianliyan) in the spring and summer of 1995 and 1996 show that concentrations of total suspended particles (TSP) change considerably in time and space, and the coastal ocean (e.g., Yellow Sea) atmosphere responds to the spring episodic dust storms in northwest China by a dramatic increase in aerosol levels associated with dust [Zhang et al., 2001]. On the basis of case studies of dust particle characteristics and their long-range transport from China to Japan in April 1993, Zhou et al.  report TSP concentrations of ∼800–900 μg m−3 during a heavy dust storm on 12 April 1993 in Beijing. These values are roughly an order of magnitude greater than typical TSP concentrations in Beijing. The study reports enrichment factors (the enrichment of the element relative to crustal material) for many elements that are distinctly lower during the dust storm period than during other days. Air quality data gathered in Taiwan in spring 2000 also provide clear evidence of the long-range transport of yellow sand dust froMmongolia, the Gobi desert, and the Loess Plateau, with the concentration of aerosols with aerodynamic diameters less than 10 μm (PM10) sharply increasing from ∼50 μg m−3 to more than ∼400 μg m−3 during the sand storm periods [Lin, 2001]. This suggests that aerosols from the Gobi desert may have a significant influence on climate both regionally and globally. In order to estimate the regional influence of aerosols on climate, it is important to determine the relative contributions of both dust and local anthropogenic sources to pertinent aerosol properties. It is also necessary to characterize these properties in order to assess the influence of Gobi desert aerosol emissions on more distant parts of Asia and the rest of the world.
 In order to characterize the aerosols emitted from the Gobi desert region, a field campaign was conducted in April 2001 in Yulin, China, as part of the Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia). Yulin is located in northern Shaanxi Province, close to the border with Inner Mongolia. It is located at the junction of the Gobi desert and the Loess Plateau (Figure 1). The aerosol light-scattering coefficient, σsp, and absorption coefficient, σap, were measured continuously. Daily PM2.5 filter samples were also collected and analyzed for concentrations of PM2.5 mass, major ions, trace elements, and elemental, organic, and carbonate carbon (EC/OC/CC). Two distinct populations of aerosols are identified on the basis of the values of aerosol mass scattering efficiencies, one dominated by desert dust and the other dominated by local pollutants. The chemical properties and major sources of the two populations of aerosols are discussed.
2. Experimental Methods
 Measurements were made at the Zhenbeitai aerosol monitoring station from 30 March to 1 May 2001. This station is set up and operated by the Institute of Earth Environment of the Chinese Academy of Science (IEECAS) to study the emission of dust from the desert. No important local pollution sources are close to the monitoring station. It is located about 5 km north of the township of Yulin at an elevation of 1118 m (latitude 38°20′N, longitude 109°43′E). The size of Yulin is about 14.2 km2, with an urban population of ∼145,000. The measurements include the aerosol light-scattering and absorption coefficients and also 24-hour-average PM2.5 mass concentration and chemical composition. All instruments and sampling inlets were installed approximately 10 m above the ground on the second level of a monitoring tower. Additional routine measurements made at the meteorological station include temperature, pressure, wind speed, and wind direction.
2.1. Measurements of Aerosol Radiative Properties
 Aerosol light-scattering and absorption coefficients were measured at the Zhenbeitai Aerosol Monitoring Station. A Radiance Research nephelometer and a Radiance Research Particle Soot Absorption Photometer (PSAP) were used to measure the light-scattering coefficient σsp at 530 nm and the light absorption coefficient σap at 565 nm. The nephelometer was calibrated prior to going into the field using clean-filtered air and HFC-134a as the span gas. Two nephelometers were run side by side in our laboratory before the field campaign to estimate the instrument precision. Sampling laboratory air, the instruments agreed to within ∼5%. Bond et al.  suggest that the PSAP has a unit-to-unit variability of ∼6%, and that the instruments may overestimate light absorption by as much as 20–30% because of light scattering and other instrumental factors. In this paper, the absorption coefficient was corrected for both light scattering and instrument overestimation as described by Bond et al. . Air was sampled at a flow rate of 16.7 L min−1 at ambient temperature (14 ± 9°C) and pressure (895 ± 19 mbar) through a URG Corporation cyclone inlet that removed particles having aerodynamic diameters greater than 2.5 μm. The air then passed through 1.5 m of 3/8″ i.d. black conductive tubing and was split into two separate sample streams using a URG Corporation flow splitter. One of these lines (flow = 0.13 L min−1) was sampled by the PSAP. The other (flow = 16.57 L min−1) was sampled by the nephelometer. The nephelometer has a capacity-type RH/temperature sensor (Vaisala, Inc., Humicap, model 50 Y) to measure the sample temperature and RH. The mean sample temperature was 15°C with a standard deviation of 10°C. The mean RH was 35% with a standard deviation of 21%. All of the flows were generated using a vacuum pump (Gast, Inc.) and maintained by critical orifices manufactured by the O'Keefe Controls Company. It is important to point out that for relatively large dust particles, nephelometers are subject to truncation errors because of the inability of these instruments to measure the scattered light at all angles from 0° to 180°. This results in the underestimation of the light-scattering coefficient. In this study, we measured the aerosol scattering coefficient of PM2.5. According to Molenar , the truncation error for a particle diameter of 2.5 μm is ∼15–20% for the Radiance Research nephelometer. Since our inlet excluded particles greater than 2.5 μm, this suggests that the light-scattering coefficient measurements may be underestimated by at most ∼20%. This error is most pronounced during the dust storm events, when particles are primarily composed of supermicron sized particles. Although, it is likely that the magnitude of the underestimation is less than 20% since the tail of the dust size distribution includes particles in the submicron mode. This may also lead to an overestimation of the absorption coefficient when PSAP data are corrected for light scattering. Therefore an underestimation of the single-scattering albedo is expected during the peak of dust events.
2.2. Measurements of Aerosol Chemical and Physical Properties
 Daily PM2.5 filter samples were collected from ∼0900 local time (LT) to 0900 LT the following day, and analyzed for PM2.5 mass and specific chemical compounds (it should be noted that the filters collected on 6 April are not included in the data presented here because of sample contamination). Air was sampled through a URG Corporation PM2.5 cyclone installed about 10 m above the ground on the second level of a monitoring tower. The flow was then separated into three sample lines by a URG flow splitter, each with a flow rate of 5.6 L min−1. One sample line was used to measure PM2.5 ionic and mass concentrations, another to measure the mass concentrations of major elements, and the third to measure elemental, organic, and carbonate carbon concentrations. The flows were generated using the vacuum pump (Gast, Inc.) and maintained by critical orifices. The air flow of each line was measured using Dwyer Rotameters, with the total air volume through the sampling systeMmonitored by a Gallus 2000 dry gas meter. The flows calculated from the readings of the rotameters and from the readings of the air meter agreed to within 8% over the entire sampling period. The pressure drop of each line was monitored by Magnahelic pressure gauges to insure that the filters were seated correctly in the filter holders and that the sampling lines had no leaks. Pall Gelman Teflon Zefluor filters (2 μm pore size and 47 mm diameter) were used to collect PM2.5 samples for mass and ionic analyses. Filter masses were determined using a Mettler Toledo MT5 electronic mass balance located in a clean room (temperature 21° ± 1°C, RH 35 ± 3%) at the Georgia Tech Southern Center for Integrated Studies of Secondary Air Pollutants (SCISSAP) analytical laboratory. The details of the analysis procedure of PM2.5 mass concentration are described by Xu et al. . Ionic components were analyzed with ion chromatography following the procedure described by Ricard et al. [2002a, 2002b] at the Laboratoire de Glaciologie et Géophysique de l'Environnement (LGGE), Université Joseph Fourier, Grenoble, France. Pallflex #2500 QAT-UP quartz fiber filters (47 mm diameter) were used to collect aerosol particles for EC, OC, and CC analyses. The quartz filters were prepared and prebaked at the University of Wisconsin, Madison [Schauer et al., 2003], and elemental, organic and carbonate carbon mass concentrations were determined using the thermal evolution technique described by Birch and Cary  and Schauer et al. . The organic carbon concentrations were then multiplied by a factor of 1.4 [White and Roberts, 1977; Countess et al., 1980; Japar et al., 1984] to determine the concentrations of the organic matter (OM), and the CC concentrations were multiplied by a factor of 5 to determine the carbonate (CO32−) concentrations. Fine particulate matter samples for element analysis were also collected on Pall Gelman Teflon Zefluor filters. The elements were analyzed with by inductively coupled plasma mass spectrometry (ICP-MS) after a microwave-assisted acid digestion in a class 200 trace metal clean laboratory at the University of Wisconsin, Madison, using the method described by Lough et al. . For 22 of the sample digestates, the elements Al, Ca, Fe, K, Mg, Mn, Na, Si, and Zn were also determined using a Perkin Elmer 4300 DV inductively coupled plasma optical emission spectrometer (ICP-OES). This is a CCD-based, simultaneous reading instrument with a typical optical resolution of 0.007 nm. All emission data were acquired in axial-viewed mode for significantly enhanced sensitivity. Multiple emission lines for each element were viewed to ensure spectral purity. The optical bench was purged with argon to improve low-UV performance. The spectrometer was interfaced with an acid resistant, polymer-based, Scott-type spray chamber and an alumina GEM-TIP nebulizer, operated at room temperature. The demountable quartz torch was fitted with a sapphire injector. Six-milliliter subsamples of the 30 mL total volume digestates from the microwave-assisted acid digestion of the aerosol samples were utilized for the analysis. Samples and standards were transferred to trace metal clean polypropylene tubes for automated analysis (AS-91 autosampler). Excellent agreement was found between ICP-MS and ICP-OES measurements (mean difference was less than 8%), with the exception of the elements Al, Ca and Mg. Although high correlation (R2 > 0.99) was found between ICP-MS and ICP-OES measured Al, Ca and Mg concentrations, ICP-MS underestimated their concentrations by about 20–30% most likely because of nonideal plasma conditions. So for these 9 samples when ICP-OES measurements are not available, Al, Ca and Mg concentrations are adjusted on the basis of the mean ratios of ICP-OES to ICP-MS measurements. Si was only measured in selected pooled samples by ICP-OES because of volatilization of silicon fluorides during microwave digestion for some samples. For this reason, the measured Si and Al concentrations in these selected samples were used to estimate the ratio of Si to Al for dust-dominated and local-pollution-dominated aerosols, which were used to estimate the Si concentrations for the remaining samples from the measured Al concentrations. The concentration of crustal material was estimated by summing the concentration of aluminum, silicon, calcium, iron, titanium, potassium, and manganese oxides (Al2O3, SiO2, CaO, Fe2O3, TiO2, K2O, and MnO2) [X. Zhang et al., 1993; He et al., 2001; Kim et al., 2001]. The sum of the concentrations of all the other elements analyzed by ICP-OES and ICP-MS are reported as trace elements. The reconstructed PM2.5 mass is computed by summing of the concentrations of OM, EC, sulfate, nitrate, carbonate, ammonium, chloride, crustal material and trace metals. In order to quantify the level of contamination during filter handling, shipping and analysis, four field blanks for PM2.5 and ions, carbonaceous compounds, and elements were also collected and analyzed. Detection limits were estimated on the basis of 3 standard deviations of the blank values for a given species. The average concentrations of the field blanks were less than 5% of the concentrations of the samples. The mean concentration of each species measured on the field blanks was subtracted from the field samples to estimate atmospheric concentration.
3. Results and Discussion
3.1. Aerosol Radiative Properties
 As shown in Table 1, the means and standard deviations of the hourly averaged aerosol light absorption coefficient σap, scattering coefficient σsp, and single-scattering albedo ω for PM2.5 are 6 Mm−1 (11 Mm−1), 158 Mm−1 (193 Mm−1), and 0.95 (0.05), respectively. As previously noted, the upper limit underestimation of σsp due to truncation error is ∼20%. Adjusting σsp by 20% results in a single-scattering albedo of 0.96, rather than 0.95. Hence truncation error has only a minor influence on ω. The mean aerosol scattering coefficient is roughly a factor of 2 lower than the mean value of 353 Mm−1 reported for the Yangtze delta region [Xu et al., 2002] and a factor of 3 lower than the mean value of 488 Mm−1 reported for Beijing [Bergin et al., 2001]. It is approximately 30% greater than the mean value of 120 Mm−1 measured in Atlanta, Georgia, in August 1999 as part of the Atlanta Super Site Experiment [Carrico et al., 2003a]. The mean aerosol absorption coefficient of 6 Mm−1 observed in Yulin during the field study is about a factor of 4 lower than the value of 23 Mm−1 [Xu et al., 2002] measured in the Yangtze delta region, and roughly a factor of 3 lower than the mean value of 16 Mm−1 in Atlanta, Georgia [Carrico et al., 2003a], in August 1999. One possible reason for this relatively low value of σap is that this remote desert region does not contain a large number of diesel vehicles, which generate a significant amount of light-absorbing soot particles. These low σap values result in a relatively high aerosol single-scattering albedo of 0.95, which is within the range of ∼0.94–0.96 reported for desert dust aerosols in Saudi Arabia (1998–2000), the Bahrain-Persian Gulf (1998–2000), and Cape Verde (1993–2000) by Dubovik et al.  and significantly higher than the mean value of 0.81 measured during June 1999 in Beijing [Bergin et al., 2001]. The single-scattering albedo is slightly lower than the in situ aircraft measured value of ∼0.97–0.99 for “clean” dust aloft during ACE-Asia [Anderson et al., 2003; Seinfeld et al., 2004], suggesting that both local pollution and sand storms influence aerosol properties in this region.
Table 1. Means and Standard Deviations of Measured Aerosol Physical and Radiative Properties in Yulin, China (30 March 2001 to 1 May 2001)a
PM2.5, μg m−3
SD, standard deviation; N, number of measurements; ω values are based on hourly averaged σsp and σap measured for PM2.5 at ambient conditions; PM2.5 values are based on the 24-hour filter samples.
 A time series of σap and σsp is shown in Figure 2. There is a great deal of diurnal as well as day-to-day variability in aerosol radiative properties, with σap ranging from <1 to 120 Mm−1, and σsp ranging from 10 to 2510 Mm−1. The data presented in Figure 2 also indicate that almost all of the σap peaks occurred in the early morning around 0800 LT. This is generally a time period of very little atmospheric mixing and relatively high local cooking emissions as well as emissions froMmobile sources used for transportation and agriculture. A large sandstorm occurred on 8 April, as is evidenced by the peak σsp value of ∼2500 Mm−1. The sandstorm-induced peak at ∼2000 LT yielded a single-scattering albedo of 0.99 which agrees with the “clean” dust value previously mentioned. The associated PM2.5 mass concentration on this day is 453 μg m−3, showing that dust storms can dominate the local aerosol mass concentration.
Figure 3a presents values of σap and σsp averaged for each hour of the day during the field study (error bars represent standard errors of mean values). The data of 8 April, which represents an extreme dust event in terms of atmospheric dust loading, are excluded because of the fact that the aerosol loadings during this afternoon were extremely high and biased the mean values from typical diurnal conditions. There is a clear diurnal pattern in both σap and σsp, with peaks occurring in the early morning at around 0800 LT and later in the afternoon between 1800 LT and 2000 LT. The diurnal variability of temperature T and relatively humidity RH are shown in Figure 4. The minimum in T and maximum in RH occur around 0600 LT. The large difference between early morning and afternoon temperatures (∼15°C) suggests that the mixing height dramatically increases during the daytime. The decrease in the surface temperature after sunset creates a temperature inversion and hence a stable atmospheric boundary layer. The presence of the temperature inversion minimizes the vertical mixing of aerosol particles. The values of σap and σsp remains relatively constant during the night from 2200 until 0600 LT, most likely because of the lack of emission sources during the late evening and early morning (i.e., no cooking and minimal mobile sources). Morning cooking emissions begin around 0700 LT. It was observed that the primary fuels used for cooking were coal and biomass. The relatively low mixing height and combustion associated with cooking in the early morning results in morning peaks in σap and σsp. After sunrise, the temperature and therefore mixing height increase. This leads to the dilution of surface air with air aloft and results in a corresponding decrease in σap and σsp. As the atmosphere becomes unstable, the wind speed typically increases, resulting in higher dust emissions from the desert surface. This results in somewhat constant values of σap and σsp in the afternoon from 1300 to 1700 LT. Localized sandstorms frequently occur when the wind speed increases, resulting in a second peak in the late afternoon. If the data of 8 April, which had maximum σsp values occurring at approximately 2000 LT, are included the second peak in σsp observed in the evenings around 2000 LT is about 280 Mm−1 as shown in Figure 3b, even greater than the observed morning peak. Overall, the strong diurnal variability suggests significant aerosol loadings from local combustion sources in the morning and from dust sources in the afternoon, although time-resolved mass and composition data are needed to verify this conclusion.
3.2. PM2.5 Mass Concentration and Aerosol Mass Scattering Efficiencies
 As is shown in Table 1, the mean and standard deviation of the 24-hour-average PM2.5 mass concentration given in Table 1 are 96 μg m−3 and 107 μg m−3. The mean value is ∼30% lower than the mean daily PM2.5 mass concentration value of 136 μg m−3 reported in Beijing during June 1999 [Bergin et al., 2001], and is close to the value of 90 μg m−3 measured in the Yangtze delta region of China [Xu et al., 2002]. The mean PM2.5 mass concentration is about 60% of the average total aerosol mass concentration of ∼160 μg m−3 reported by Zhang et al.  based on ∼200 samples collected from the northwest China desert region (Minqin) during the spring and summer of 1995 and 1996. In addition, the mean PM2.5 mass concentration at Yulin is about a factor of 1.5 greater than the proposed 24-hour-average U.S. National Ambient Air Quality Standards (NAAQS) for PM2.5 of 65 μg m−3, and a factor of 6 greater than the proposed annual average U.S. NAAQS for PM2.5 of 15 μg m−3.
 The ratio of σsp to PM2.5 mass concentration is the aerosol mass scattering efficiency for particles having aerodynamic diameters less than 2.5 μm (Escat_2.5). The aerosol mass scattering efficiency depends on several factors including particle size, chemical composition, morphology, relatively humidity and wavelength of light. A peak in Escat_2.5 occurs when the particle diameter is roughly equal to the wavelength of incident light. The mass scattering efficiency decreases for particles having diameters both larger and smaller than the wavelength of incident light. Since windblown dust is typically composed of relatively large particles, the mass scattering efficiency of mineral aerosols at solar wavelengths is generally lower than that of anthropogenic pollutants such as sulfate [Waggoner et al., 1981; Li et al., 1996; Seinfeld and Pandis, 1998]. The PM2.5 mass scattering efficiency is estimated from the ratio of 24-hour-mean σsp and the corresponding 24-hour-average PM2.5 filter mass concentrations and is shown in Figure 5. Because of the fact that aerosol scattering coefficients were measured at the RH of 35 ± 21% while PM2.5 mass concentrations were measured at the RH of 35 ± 3%, condensed water may lead to a 10–20% overestimation of the mass scattering efficiencies during those high-RH days [Xu et al., 2002]. There are two distinct populations of Escat_2.5, one for an aerosol chemical composition dominated by dust (∼1.0 m2 g−1) and the other for an aerosol chemical composition dominated by local pollutants (∼3.0 m2 g−1). Because of the truncation error of the Radiance Research nephelometer [Molenar, 1997], it is likely that the maximum underestimation of σsp is about 20% on dusty days. Therefore our Escat_2.5 values may be as much as 20% low on these days. On the basis of the values of Escat_2.5, the sampling periods during the field campaign can be divided into two groups: one with aerosols dominated by desert dust and the other with aerosols dominated by local pollutants. The time intervals of 6–12, 15–19, and 29–30 April had relatively low Escat_2.5 values and are identified as sand storm periods. The aerosol mass scattering efficiency of 3.0 m2 g−1 for aerosols dominated by local pollutants is within the range of values (2.9 to 3.2 m2 g−1) reported by Waggoner et al.  for various urban locations within the United States, and is at the low end of the values of 3.3 to 3.6 m2 g−1 estimated for local pollutant and dust mixed aerosols for particles having aerodynamic diameters less than 1.8 μm in Beijing by Bergin et al. . It is about 30% lower than the mean value of 4.0 m2 g−1 reported for the urban aged aerosols in the Yangtze delta region of China [Xu et al., 2002]. The mean value of 1.0 m2 g−1 for large desert dust aerosols is somewhat higher than the value of 0.83 m2 g−1 reported for mineral dust over the North Atlantic [Li et al., 1996] probably because of the fact that the mass scattering efficiencies estimated here are only for PM2.5 and not total suspended particles (TSP). The value is roughly 1/3 of the value for local polluted air because of the fact that large dust particles are generally less efficient at scattering solar radiation. Table 2 shows the means and standard deviations of aerosol physical and radiative properties for the two different aerosol types. Although the mass scattering efficiency of dust-dominated aerosols is much lower than that of aerosols having local sources, the mean σsp of dust is still roughly 1.7 times that of local pollutants because of the fact that the mean PM2.5 mass concentration of dust is about 4 times greater than that of local pollutants.
Table 2. Means and Standard Deviations of Measured Aerosol Physical and Radiative Properties in Yulin, China, for Different Aerosol Typesa
PM2.5, μg m−3
SD, standard deviation; N, number of measurements; ω values are based on hourly averaged σsp and σap measured for PM2.5 at ambient conditions; PM2.5 values for desert-dust-dominated aerosols are based on daily filter samples collected over 13 days and for local-pollutant-dominated aerosols are based on filter samples collected during 18 days.
3.3. Aerosol Chemical Properties
 Daily integrated PM2.5 mass concentrations of major chemical compositions are presented in Figure 6. Negative unidentified values result when PM2.5 mass concentrations are less than the sum of the concentrations of all identified chemical species multiplied by assumed correction factors to account for the oxygen and hydrogen associated with the measured species. Relatively good agreement is found between the reconstructed (sum of the concentrations of OM, EC, sulfate, nitrate, carbonate, ammonium, chloride, crustal material, and trace metals) and measured PM2.5 mass concentrations as shown in Figure 7. Figure 6 indicates that there was a large amount of day-to-day variability in aerosol loadings during our field campaign, with daily mean PM2.5 mass concentrations ranging from 23 to 453 μg m−3. As discussed in the previous section, the aerosol mass scattering efficiencies suggested that there are several periods during our field campaign (6–12, 15–19, and 29–30 April) when large coarse particles dominate PM2.5 mass. This agrees with the aerosol chemical composition data presented in Figure 6, which shows that during these periods the aerosol loadings are much higher than the local pollution periods, and a large proportion of the PM2.5 mass is crustal material. The highest aerosol mass concentration of 453 μg m−3 is found on 8 April. This corresponds to the peak in σsp and is the most significant dust storm observed during the study. The significant day-to-day variability of both aerosol absorption and scattering coefficients as well as PM2.5 mass concentration suggests that both local pollution and sand storms influence aerosol properties in this region.
 On the basis of the aerosol mass scattering efficiencies, the data are divided into two groups, one for aerosols dominated by local emissions and the other for aerosols dominated by dust emitted from the desert surface during dust storm events. The means and standard errors (SE) of PM2.5, as well as the concentrations of major chemical species and their relative contributions to the PM2.5 mass for the two populations of aerosols are shown in Table 3 and Figure 8. Once again, negative unidentified values indicate that PM2.5 mass concentrations are less than the sum of the concentrations of all identified chemical species. The mean PM2.5 mass concentration during dust storm periods is 169 μg m−3, about 4 times greater than the mean value of 44 μg m−3 during local pollution periods. The reconstructed PM2.5 mass is on average about 13% higher than the measured value for local-pollutant-dominated aerosols, and about 20% lower for dust-dominated aerosols. For local-pollutant-dominated aerosols, organic matter (OM) dominates the chemical composition contributing 41% to the PM2.5 mass, followed by crustal material (29%), sulfate (17%), EC (13%), ammonium (6.2%), and nitrate (5.8%). The chemical speciation for aerosols dominated by local pollution in Yulin is similar to that measured for dust influenced urban polluted air in Beijing [He et al., 2001]. The mean of the ratio of organic to elemental carbon is 2.6 with a standard deviation of 1.2. This is similar to the value of 2.7 reported for Beijing in spring 2000 [He et al., 2001] and 2.6 reported for Taiwan for the period of November 1998 to April 1999 [Lin, 2002]. The presence of high concentrations of OM, EC and sulfate and the observed extensive coal usage for heating and cooking suggest that coal combustion may be one of the most important sources of local pollutants in Yulin. A large portion of K+ (about 1.4%) is found during the local pollution periods, suggesting that biomass burning may also contribute to the aerosol loading in this region [Reid et al., 1998]. Overall, the results indicate that local pollution sources make a significant contribution to the aerosol loadings even in this remote desert region.
Table 3. Mean and Standard Error Values of Daily Average Concentrations of PM2.5 and Major Chemical Compositions and Their Percentage Contributions to the PM2.5 for Two Populations of Aerosolsa
Mean, μg m−3
SE, μg m−3
Mean, μg m−3
SE, μg m−3
SE, standard error.
 During the sand storm periods, about 51% of PM2.5 mass is crustal material. Other major chemical species include carbonate (11%), OM (9.5%), sulfate (3.4%), and EC (2.1%). Although the PM2.5 mass concentrations are much higher during sand storm periods, the concentrations of chemical species generally associated with local combustion sources (OM/EC, NH4+, SO42− and NO3−) are noticeably lower than during local pollution periods. This is very likely due to the change in direction where the air mass come from, the increase in the mixing height associated with the high wind speeds (20–40 m s−1) of sand storms, and losses of particle precursors (e.g., HNO3 and SO2) to surfaces by coarse dust particles. During these periods, the ratio of organic to elemental carbon averaged 5.7 with a standard deviation of 4.6, more than double the ratio of 2.6 observed during local pollution periods. This may also have been due to the fact that during sand storm periods, a large portion of the air mass are typically transported from the desert region northwest of Yulin where the EC concentration may be relatively lower. During sand storm periods, high concentrations of carbonate (18.7 μg m−3), which is usually associated with dust sources, were observed. The concentrations of trace metals are also generally about an order of magnitude higher during the sand storms. About 20% of the PM2.5 is unidentified for desert-dust-dominated aerosols. This is similar to the approximately 20% unidentified PM2.5 mass reported for the chemical speciation of dust aerosols in Beijing determined by X-ray fluorescence by Bergin et al.  and He et al. . Possible reasons for this large portion of unexplained mass for dust-dominated aerosols in this study may include the following: (1) The concentrations of the crustal materials are underestimated by summing Al2O3, SiO2, CaO, Fe2O3, TiO2, K2O, and MnO2 because of the presence of other compounds, such as MgO, NaO and H2O in the crust [Kim et al., 2001; Carrico et al., 2003b]. (2) The molar ratio of ammonium to sulfate is ∼0.8 for dust-dominated aerosols (∼0.2 during the peak of the sand storms), compared to about 1.8 for local-pollutant-dominated aerosols. This suggests that during the dust storm events, more than half of the sulfate is not balanced by ammonium and may have been present in some other chemical form such as CaSO4 or sulfuric acid which is highly hygroscopic. Therefore the water uptake by the filters may have significantly increased filter mass even under our clean room conditions. (3) Finally, there are some uncertainties associated the mass concentration measurements. Further work needs to be done to resolve the unidentified mass.
3.4. Enrichment Factors and Sources of Elements
Table 4 lists the mean and standard error values of the elements analyzed by ICP-OES and ICP-MS for both desert-dust- and local-pollutant-dominated aerosols. The mean concentrations of the major crustal elements (Na, Mg, Al, Ca, Fe, Ti, and Mn) [Zhang et al., 1997, 2002; Choi et al., 2001] are about an order of magnitude higher during dust storm events than during local pollution periods. On the other hand, although the PM2.5 mass concentrations are about a factor of 4 higher during sand storm periods, the concentrations of Zn, As, Mo, Ag, Cd, Sb, and Pb are close to or even lower than during local-pollution-dominated periods. This strongly implies that local pollution is the major source of these elements.
Table 4. Mean and Standard Error Values of Daily Average PM2.5 Concentrations and Enrichment Factors of Major Elementsa
SE, standard error; EF, enrichment factor (EFcrust = (X/Al)air/(X/Al)crust). Crustal abundance values are from Lide . Read 3.7E-01 as 3.7 × 10−1.
 Because of the lack of detailed source profiles and limited amount of data, we cannot quantify the contributions of the different specific sources (i.e., coal combustion, biomass burning, cooking, etc.) to the aerosol loadings in this region. In order to better interpret the sources of the aerosol trace elements, the mean and standard error values of the enrichment factors (EF) relative to crustal material for both aerosols dominated by desert dust and local pollutants are calculated and listed in Table 4. The enrichment factors are calculated using the equation:
where EFX is the enrichment factor of element X relative to crustal material, and Ref is a reference element. Crustal material is considered to be the dominant source of element X if EFX approaches unity. High elemental enrichment factors suggest that local pollution sources are responsible for that particular element. Aluminum is used as the reference element in this study because it is a key crustal tracer and usually has no local pollution sources. The crustal compositions are quite source dependent. In fact, the crustal material in our samples may come from several different source regions with quite different crustal profiles. Because of the lack of crustal composition data in this region the crustal abundance values of the elements are adapted from Lide . As shown in Table 4, the EF values of Na, Mg, Al, Ca, Fe, Ti, V, Cr, Mn, Sr, Ba, Ce, and U are close to unity both for desert-dust-dominated and local-pollutant-dominated aerosols. This indicates that the crust is the dominant source for these elements. The EF of potassium is 2.2 during dust storm events and rise to as high as 4.8 during local pollution events. This suggests that both local pollution sources such as biomass burning and windblown desert dust make significant contributions to the potassium concentration in this region [He et al., 2001]. Similar to K, the EF of 1.4 for Cu during dust storm events is much smaller than the value of 5.9 observed during local pollution events. Very large enrichment factors were found for Zn, As, Mo, Ag, Cd, Sb, W, Tl, and Pb during both local pollution and dust storm periods. However, the values during local pollution periods are more than an order of magnitude higher than during dust storm periods. This suggests that these elements are very important pollution tracers in Yulin. As and Sb are two of the major impurities in coal [Zhuang et al., 2001]. High concentrations and enrichment factors imply that coal combustion is a very important local pollution source. The enrichment factor of Pb is approximately 500 during local pollution periods, suggesting that local mobile source emissions make a significant contribution to aerosol loadings in this region [Fang et al., 1999; Zhang et al., 2002]. Overall, the enrichment factors of the elements suggest that both desert dust and local pollution sources, such as coal combustion, mobile source emissions and biomass burning, contributed to aerosol loadings in Yulin during our field campaign.
4. Summary and Conclusions
 Two distinct populations of aerosols are found in Yulin during the ACE-Asia field campaign, one dominated by desert dust aerosols with a mean Escat_2.5 of approximately 1.0 m2 g−1 and the other dominated by local pollution with Escat_2.5 of about 3.0 m2 g−1. The means and standard deviations of the measured aerosol light absorption coefficient σap, scattering coefficient σap, single-scattering albedo ω, and PM2.5 mass concentration in Yulin for aerosols dominated by local emissions are 6 (9) Mm−1, 125 (110) Mm−1, 0.94 (0.05), and 44 (20) μg m−3, and for desert-dust-dominated aerosols are 7 (12) Mm−1, 207 (251) Mm−1, 0.96 (0.04), and 169 (135) μg m−3, respectively. The mean PM2.5 mass concentration of 169 μg m−3 during dust storm periods is about 4 times greater than the mean value of 44 μg m−3 observed during local pollution periods. Organic matter (OM) is the major chemical component measured during local pollution periods comprising approximately 41% of the PM2.5 mass. Other major components include crustal material (29%), sulfate (17%), EC (13%), ammonium (6.2%), and nitrate (5.8%). During sand storm periods, approximately 51% of PM2.5 mass is crustal material followed by CO32− (11%), OM (9.5%), sulfate (3.4%), and EC (2.1%). The data indicate that even during the local-pollution-dominated days, there is still ∼29% crustal material in PM2.5. Local pollutants also exist during the dust-dominated days. Diurnal patterns for both σsp and σap are observed with peaks both in the morning and the afternoon. The morning peak is primarily associated with local coal combustion and a low mixing height resulting from the nighttime temperature inversion. The afternoon peak is likely associated with dust sources. The significant day-to-day variability of σsp, σap, and PM2.5 suggests that both local pollution and sand storms influence aerosol properties in this region.
 The EF values of Na, Mg, Al, Ca, Fe, Ti, V, Cr, Mn, Sr, Ba, Ce, and U are close to unity both for desert-dust- and local-pollutant-dominated aerosol samples, suggesting that desert dust is the primary source of these elements. During dust storm events, the EF of potassium was 2.2, much lower than the value of 4.8 observed during local pollution periods. This suggests that in addition to desert dust, local pollution sources such as biomass burning make a significant contribution to the potassium concentration in this region. Very large enrichment factor values are found for Zn, As, Mo, Ag, Cd, Sb, W, Tl, and Pb during both local-pollution- and dust-storm-dominated periods, suggesting that local emissions are the primary sources for these elements. High enrichment factors of As and Sb indicate that coal combustion is a very important local pollution source. The enrichment factor of over 500 for Pb during local pollution periods suggests that mobile source emissions also contribute to the aerosol loadings in this region. Overall, the results suggest that both desert dust and local pollution sources, such as coal combustion, mobile source emissions and biomass burning, contributed to the aerosol loadings, and hence direct aerosol radiative forcing in Yulin during our field campaign.
 This project was supported by the ACE-Asia NSF grant ATM-0080325. We thank J. Zhang from Yulin Desert Research Institute of China and G. Shi, X. Zhang, and H. Mei from Institute of Earth Environment of Chinese Academy of Science (IEECAS) for their help with our field campaign in Yulin, China.