Impact of Meteorological Parameters and Gaseous Pollutants on PM2.5 and PM10 Mass Concentrations during 2010 in Xi'an, China

ABSTRACTMass concentrations of PM2.5 and PM10 from the six urban/rural sampling sites of Xi’an were obtained during two weeks of every month corresponding to January, April, July and October during 2010, together with the six meteorological parameters and the data of two precursors. The result showed that the average annual mass concentrations of PM2.5 and PM10 were 140.9 ± 108.9 µg m–3 and 257.8 ± 194.7 µg m–3, respectively. Basin terrain constrains the diffusion of PM2.5 and PM10 concentration spatially. High concentrations in wintertime and low concentrations in summertime are due to seasonal variations of meteorological parameters and cyclic changes of precursors (SO2 and NO2). Stepwise Multiple Linear Regression (MLR) analysis indicates that relative humidity is the main factor influencing on meteorological parameter. Entry MLR analysis suggests that SO2 from local coal-burning power plants is still the primary pollutant. Trajectory cluster results of PM2.5 at BRR indicate that the entrained urban pollutants carried by the westerly or winter monsoon forms the dominant regional pollution sources in winter and spring. Ultraviolet (UV) aerosol index verified the source and pathway of dust storm in spring.


INTRODUCTION
Atmospheric particulate PM 2.5 (particles with aerodynamic equivalent diameter ≤ 2.5 µm), also known as the lungs of particulate matter, is considered as an important indicator of air quality because of its effects on human health (Haywood and Boucher, 2000;IPCC, 2007, Tie et al., 2009).The extent of PM 2.5 concentration depends on both direct emissions and the quantity of gaseous precursors, such as SO 2 and NO 2 , etc. as well as meteorological conditions.
The major source of SO 2 is exclusively from coal-burning power plants and the NO 2 in Xi'an is exclusively from vehicles.In the last decade, SO 2 was the primary gaseous pollutant in China, but recent studies from ground measurement (Zhang et al., 2012;Cao et al., 2013), satellite wind direction (WD), wind speed (WS), temperature (T), relative humidity (RH), surface pressure (PS) and planetary boundary layer height (PBLH) etc. are expected to have important effects on PM 2.5 variation to some extent.For example, WS can alter the dispersion state of the atmosphere, while WD provides information on the path of pollutants (Luvsan et al., 2012;Tie et al., 2015).Low PBLH, a strong temperature inversion, and descending air motion under weak surface WS can increase the frequency of haze phenomenon in China (Jones et al., 2010;Hu et al., 2012), particularly in some cities located in basin regions during winter, such as Beijing and Xi'an (Chen et al., 2012;Lin et al., 2012).However, there is no clear linear relationship between meteorological parameters (i.e., wind speed, temperature or relative humidity) and the concentration of PM 2.5 due to effects of chemical reactions and transformations (Chan and Yao, 2008).For example, sulfate concentration is expected to increase with temperature rising due to the faster SO 2 oxidation, but on the other hand, semi-volatile components, such as nitrate and organics, are expected to decrease because the transformation from the particle phase into the gas phase is easily formed at low temperature.An increase of cloud can cause an increase of sulfate and nitrate due to the enhancement in sulfate heterogeneous reactions in water (Zhang et al., 2013).However, an increase in precipitation causes a decrease in PM 2.5 concentrations through scavenging, so it is necessary to have a comprehensive understanding of the uncertain sensitivity between meteorological parameters and PM 2.5 and PM 10 concentrations.
Xi'an (34°16′N, 108°54′E), the capital city of Shaanxi province, is located in the Guanzhong Plain (Fig. 1).Xi'an has a temperate, semi-arid climate with northeast prevailing wind.An annual precipitation is 550 millimeters and an annual average of temperature is 13.7°C.Dust storm often occurs during March and April.Geographically, Xi'an is a basin surrounded by Qinling Mountainin in south and the Loess Plateau in north and west.Moreover, only one outlet in the northeast of Xi'an determines that the pollutants in Xi'an cannot be easily dispersed out.Furthermore, regional source is easily transported from adjacent district throughout the outlet.For example, coal-dominant energy consumption in northeast part of Shaanxi province produces large amount pollutants, which can be transported through the outlet to the Guanzhong Plain (Zhao et al., 2015).Previous findings show that the average PM 2.5 concentration in Xi'an reached its the highest level during winter (375.2 µg m -3 ) and summer (130.8 µg m -3 ) during 2003 (Cao et al., 2012a).The air quality in Xi'an has been degrading for years and especially in winter, and the poor air quality posts adverse effects on local residents (Cao et al., 2007;Huang et al., 2012).
The objective of this study is to discuss the variation of PM 2.5 and PM 10 mass concentrations in Xi'an driven by the meteorological parameters and the emissions of their precursors.The discussion includes: (1) PM 2.5 and PM 10 characteristics, (2) the key meteorological factors, which affect PM mass concentration and the conversion of gaseous SO 2 and NO 2 to PM 2.5 and PM 10 , (3) effects of potentially regional sources on PM 2.5 .This study will provide necessary reference information for the local government to develop atmospheric environment pollution control strategies and an emergency plan.

The Monitoring Locations
The six urban/rural sampling sites are shown in Fig. 1, illustrating their topography, key polluting enterprises and traffic route.Four urban sites include Institute of earth environment, Chinese academy of science (IEE) in the south, Micro motor factory (MMF) in the west, Municipal government hall (MGH) in the Downtown zone, and Chan-ba ecological district (CBE) in the east.Two rural sites include Gaoling county (GLC) in the northeast and a reference station of Black river reservoir (BRR) in the southwest.The reason for choosing BRR station as a reference to investigate potentially regional sources is that it is located on the edge of Qingling Mountains.

PM 2.5 and PM 10 Sampling Analysis
Daily PM 2.5 and PM 10 samples were collected during 4 periods, including 1) 12 to 25 January; 2) 14 to 27 April; 3) 12 to 25 July; and 4) 12 to 25 October, 2010, corresponding to winter, summer, spring and fall, respectively.Minivals (Airmetrics Corp., Springfield, OR, USA) were installed at each station, with 10 m above ground level.These samplers used the Whatman quartz microfiber filters (QM/A), with 47 mm in diameter and operated at 5 L min -1 of inlets.Filter blanks were collected from each site during the sampling time.Prior to sampling, all the samplers were carefully checked and calibrated.A total 56 pairs of PM 2.5 and PM 10 filters were collected at the six sampling sites.
PM 2.5 and PM 10 mass concentrations of the sample filters were analyzed by using an electronic microbalance, with 1 µg sensitivity (MC5; Sartorius, Goettingen, Germany) in a controlled environment (35-45% RH at 20-23°C).

Data of Gaseous Pollutants and Meteorological Parameters
Daily average SO 2 and NO 2 concentrations (Fig. 2(d)) at eleven automatic monitor stations were obtained from the website of Xi'an environment monitoring center (http://www.xianemc.gov.cn).
the ground level are calculated at 10:00 LST (Local Standard Time) per day during the observation period.The individual trajectories were grouped into 2-3 clusters (Fig. 5 and Table 3) by using the clustering tool in TrajStat a GIS Trajectory Analysis Tool (MeteoInfo) (http://www.ready.noaa.gov/HYSPLIT.php).
Giovanni is a web-based application developed by the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC).It provides a simple and easy way to explore, visualize, analyze, and access a vast amount of Earth science remote sensing and model data (http://disc.sci.gsfc.nasa.gov/giovanni/overview/index.html).Ozone Monitoring Instrument (OMI) Ultraviolet (UV) Aerosol Index (AI) of Aura satellite is most sensitive to absorbing aerosols above the planetary boundary layer, i.e., smoke plumes and dust (de Graaf et al., 2005) (Fig. 6).

PM 2.5 and PM 10 Spatial and Seasonal Variations
Fig. 2 shows six daily average meteorological parameters and, SO 2 , NO 2 , PM 2.5 , PM 10 mass concentrations, the ratio of PM 2.5 /PM 10 during the 4 periods.Table 1 summary the seasonal average concentrations (mean concentration ± standard deviation) of the meteorological parameters, SO 2 , NO 2 , PM 2.5 and PM 10 during observed date in 2010.The average concentration of PM 2.5 is 140.9 ± 108.9 µg m -3 with an arrange from 23.1 to 562.6 µg m -3 .This is about 13 times higher than that of the World Health Organization (WHO) standard (10 µg m -3 ), 8 times higher than U.S. standard (15 µg m -3 ), 5 times higher than EU PM 2.5 (25 µg m -3 ), and 3 times higher than that of the latest average annual standard (GB 3095-2012) issued by the Chinese Ministry of Environmental Protection (35 µg m -3 ) (http://www.envir.gov.cn/law/airql.htm).PM 10 is 257.8 ± 194.7 µg m -3 (range value: 46.1-926.8µg m -3 ), and is twice of the average annual standard of China (100 µg m -3 ).The PM general variation is similar to that in previous studies in Xi'an (Han et al., 2008;Shen et al., 2008Shen et al., , 2009Shen et al., , 2011)).
Fig. 3 presents the daily variations of PM 2.5 and PM 10 mass concentrations at six sampling sites spatially.The peaks of PM 10 in six sites found during a dust storm time (Fig. 3(b) with yellow maker) indicates that the course particles loading in PM 10 is caused by a constant trend and a low PM 2.5 /PM 10   ratio at six sites.While, a "sawtooth" variation of fine particles PM 2.5 indicates that local sources could be the major contributors, especially in wintertime (Fig. 3(a)).A "pear-shaped" geography of Xi'an surrounded by three mountains with an outlet to northeast determines that the pollutants cannot disperse easily (Fig. 1), which could support this argument above (Zhu et al., 2010;Wang et al., 2015).
Seasonal variations are shown in Fig. 2(e) and Table 1, which indicate that PM 2.5 /PM 10 mass concentrations are high in wintertime and low in summertime.This is attributed to the seasonal variations of meteorological parameters and cyclic changes of precursors from source emission.The highest concentrations of PM 2.5 and PM 10 can be found in wintertime (except for during a dust storm time) for following reasons.Firstly, the strong temperature inversion (average concentration ± standard deviation: -0.3 ± 2.9°C) and descending in air motions in PBLH (386.1 ± 106.9 m) allows pollutants to accumulate in a shallow layer (Table 1).However, on the other hand, the low surface wind speed (2.5 ± 0.5 m s -1 ) is not favor to diffuse SO 2 and NO 2 concentrations (62.1 ± 12.4 µg m -3 and 40.5 ± 11.6 µg m -3 , respectively) (Table 1).In addition, in winter heating period, PM 2.5 and PM 10 concentrations at the three sampling sites (MMF, MGH and IEE) in urban area are obviously higher than that of two sites in suburban area (GLC and BRR) (Table 1).This suggests that central heating has a serious impact on PM 2.5 and PM 10 in central urban area.This conclusion is also verified in precious studies via the PMF analysis, which show that coal combustion contribution of percentage of 21% for PM 1 during 2008 (the contribution of coal combustion is 21% for PM 1 ) (Shen et al., 2010), 52.2% (the sum of coal combustion and secondary aerosols) for PM 2.5 during 2009 (Cao et al., 2012b) and 18.5% for PM 2.5 during 2010 (Wang et al., 2015) in Xi'an.According to Fig. 4(a), the spatial distributions of SO 2 from coal-burning energy consumption can also verify that the precursors of PM 2.5 are mainly concentrated around the central urban area.The variation of the peak points of PM 2.5 concentration in wintertime is influenced by the interaction of driving forces of meteorological factor (low temperature, low wind speed and low PBLH) and high emissions of SO 2 .
In springtime, compared to PM 2.5 (except at reference site BRR) , the peaks of PM 10 mass concentrations in the six sampling sites are almost equal (Figs.3(a) and 3(b)), which indicates that a dust storm time (25, 26, 27/04, 2010) influences on PM 10 with northwest prevailing wind at mean speed of 4.4 m s -1 (Fig. 2(a)).During a dust storm time, the lowest PM 2.5 /PM 10 ratio (0.3) is observed in 25/04, which is also the lowest ratio over the whole observed time, then followed by 0.4 in 26/04 and 0.5 in 27/04 (Fig. 2(e)).Those facts indicate that the coarse mode particles are the dominant PM type during the dust storm.
The lowest concentrations of PM 2.5 and PM 10 over the entire observed time appeared in summertime and whose trend remains stable (Figs.2(e) and 3).This is caused by the accelerated homogeneous and heterogeneous reactions between SO 2 and radicals, such as OH, H 2 O 2 , etc., with a high average temperature (20.9 ± 2.0°C) (Table 1).But the high temperature is also expected to decrease PM 2.5 species compositions, for instance, semi-volatile components: nitrate and organics, because their dynamic equilibrium of the transformation changes from the particle phase into the gas phase (Cao et al., 2007).For example, ammonium nitrate volatilizes partially at a temperature over 20°C, meanwhile it forms gaseous nitric acid.When the temperature rises over 25°C, the volatilization completes (Calvo et al., 2013).Moreover, based on the statistical analysis from 1970 to 2009 in Xi'an weather station (34.3°N, 109.0°E,399 m above sea level), the average annual precipitation mainly occurs in July and September (http://www.tianqi.com/xian/index.html).Therefore, the low concentrations of PM 2.5 and PM 10 can be explained by that the combining result of the rainout effect and suppression effect of RH exceeds the accelerating effect of RH.Thus, the low PM 2.5 /PM 10 values in summertime may be caused by the rainout and the decrease of the semivolatile components in PM at a high temperature.
The moderate concentrations of PM 2.5 and PM 10 (Fig. 3(e)) in autumn could be explained by the low average PBLH (310.2 ± 51.5 m) and the air stagnant caused by a low wind speed (2.2 ± 0.8 m s -1 ) (Table 1).
In short, basin terrain constrains PM 2.5 /PM 10 spatial diffusion.Seasonal variations of meteorological parameters and cyclic changes of precursors from source emission are the main causes for the PM 2.5 /PM 10 seasonal distribution.

Factors Mainly Influencing to PM 2.5 and PM 10 Concentrations, Respectively
Meteorological parameters have complex effects on the total PM concentration due to no clear linear relationship between each other (Chan and Yao, 2008).However, based on the knowledge of PM and five meteorological factors mentioned above, the stepwise MLR can be applied to extract the major influencing factors (Stehr et al., 2000).All Variance Inflation Factors (VIF) of independent variables are less than 10, indicating that it does not have significant collinearity effects before model diagnosis (Tong et al., 2005).Table 2 presents stepwise MLR results, together with parameter estimation (T-test) of independence variable and model formula estimation (F-test and R).The result shows that the main contributors to mass concentration of PM 2.5 are the precursor of SO 2 , temperature and RH.In contrast, the main contributors to mass concentration of PM 10 are NO 2 , wind speed, surface pressure and RH.RH, as a coexistent influencing factor for mass concentrations of PM 2.5 and PM 10 can indicate the major liquid-phase/heterogeneous reaction in the process of PM formation.However, there is an interesting result showed by MLR analysis between RH and concentrations of PM 2.5 and PM 10 .The result indicates a negative correlation coefficient between RH and PM, which is not common in other cases.Thus, more in-depth study should be conducted.In addition, sulfate formation through SO 2 oxidation dominates mass concentrations of PM 2.5, and nitrate formation through NO 2 oxidation is the dominant substance in PM 10 .Those findings are consistent with previous study (Yao et al., 2002).
Recent studies show that SO 2 emissions started decreasing in China after 2006 (Lu et al., 2010).Nitrogen oxides would be the major pollutant to be concerned about in the present and future.Satellite observation and model simulation also detected and predicted a strong increase of NO 2 in Eastern China (Li et al., 2010a).In particular, Shanghai had a significantly linear increase of NO 2 column concentrations with about 20% increase rate annually during the period of 1996-2005 (He et al., 2007).However, SO 2 from local coal-burning power plants is the major primary pollutants, and the majority of it is produced by its raw material-based economic development and west-east electricity transmission project in Xi'an.

Contribution Rates of SO 2 and NO 2 to PM 2.5 and PM 10
It is a common understanding that point sources (mainly coal-burning) emit high level of SO 2 , and relatively low level of NO 2 .However, mobile sources (mainly fossil fuelburning) emit high level of NO 2 .Based on the observed data from eleven automatic states of Xi'an environment monitoring center, the relative contributions from point sources (stationary) and mobile sources to PM 2.5 and PM 10 are investigated through entry MLR is applied (Tong et al., 2005;van der A et al., 2006).The mathematical expression of the model is where α 1 , α 2 , β 1 and β 2 are the linear coefficients between [PM 2.5 ], [PM 10 ] and [SO 2 ] and [NO 2 ], respectively, and δ 1 , δ 2 are the intercept, respectively.Before the parameterizations of α 1 and β 1 , and α 2 and β 2 , are calculated, the collinearity effects of SO 2 and NO 2 on the estimates of regression coefficient are diagnosed.The VIF is 1.55 (< 4) suggesting that the collinearity is not significant.Results show that standardized parameters of α 1 , β 1 , α 2 and β 2 are 0.486, 0.175, 0.395 and 0.007 respectively, which suggest that the emission from coal-burning makes a major contribution to PM 2.5 and PM 10 , with up to 48.6% and 39.5%, respectively The contribution of vehicle emission are 17.5% and 0.7%, and the contribution of other sources are 33.9% and 59.8%.The contribution from SO 2 to PM 2.5 and PM 10 outweighing that from NO 2 proves that the dominant polluting gas is SO 2 in Xi'an again.Previous study shows that about 73% of SO 2 emitted from power plant contributes 11% of the total PM 10 and 12% of the total PM 2.5 .31% of NO 2 from  traffic sectors (fossil fuel burning) is accountable for 7% of the total PM 10 and 10% of the total PM 2.5 in the year of 2003 in Huabei region, China (Stehr et al., 2000).The contributions of SO 2 and NO 2 to PM 2.5 and PM 10 in 2010 in Xi'an are quite different from those in 2003 in Huabei region.The major uncertainties depend on the two independent variables of regression model and co-emission of SO 2 and NO 2 by power plant and transformation sources.

Effects of Potentially Regional Source on PM 2.5
The contributions of local source to PM 2.5 /PM 10 via emitting precursors of SO 2 and of NO 2 have been mentioned above.The effects of regional source on PM 2.5 studied by trajectory clusters are elaborated as follow.Trajectory analyses provide an insight into the impact of long-range air transport on PM variation at receptor site (Li et al., 2010b;He et al., 2012).Taking PM 2.5 of BRR at reference sampling site as an example, Fig. 5 shows the mean trajectory of the 2-3 clusters and their percentages to the total number of trajectories.PM 2.5 concentrations are associated with the trajectories and grouped according to selected pollution trajectory criteria of seasonal average values for PM 2.5 in BRR (Table 2).Then each group of data is summarized for statistical analysis (Table 3).It is generally that the long distance air mass responds to regional source and the short air mass responds to local source.Uncertainty of trajectory clusters could be caused by the selected threshold of the polluting trajectory from different seasons.This determines that which air mass travelling at receptor site is considered as a polluting trajectory.It can be seen from Table 3 that polluting trajectory numbers of winter, spring, summer and autumn corresponding to seasonally observed data (1 × 14 day) are 6, 3, 7 and 4, respectively.The mean value of polluting trajectory numbers selected in descending order is spring (333.7 ± 203.4 µg m -3 ) > winter (146.7 ± 24.2 µg m -3 ) > autumn (99.0 ± 20 µg m -3 ) > summer (52.5 ± 21.1 µg m -3 ).As it shows in Fig. 5, the long distance air masses are dominant in winter (including cluster 1 of 83.3%) and spring (66.7%), and the short distance air mass is the major type in summer (71.4%) and autumn (75%).The entrained urban pollutants carried by the westerly or winter monsoon forms the dominant regional pollution sources in winter and spring (e.g., Lanzhou of Gansu and Yinchuan of Ningxia).However, in spring (Fig. 5(b)), high levels of external polluting trajectory (333.7 µg m -3 ) may be a reasonable explanation due to the dilution of dust particulates in the dust storm time, which is consistent with results of impact of Gobi desert dust on aerosol chemistry of Xi'an (Cao et al., 2005;Wang et al., 2009).
The results of UV aerosol index analysis turn out to be consistent with trajectory analysis of air mass in Fig. 5.The data and results showed in Fig. 6, which also illustrates the   source (Taklimakan Desert in Xinjiang and Gobi Desert in Inner Mongolia) and pathway (Hexi corridor) during a dust storm time.In short, effect of seasonally regional source on PM 2.5 in BRR reference site based on percentage of polluting trajectory to total trajectory is possible.

CONCLUSION
The average concentrations of PM 2.5 and PM 10 in 2010 are 140.9 ± 108.9 µg m -3 , 257.8 ± 194.7 µg m -3 , respectively.Wind speed in calm and basin terrain constrains PM spatial diffusion.Seasonal variations of meteorological parameters and cyclic changes of precursors from source emission are the main causes of PM 2.5 /PM 10 seasonal distribution.
Factors that influence on PM 2.5 from stepwise MLR analysis application are SO 2 , temperature and RH.While NO 2 , wind speed, surface pressure and RH are the major ones for PM 10 .RH as the co-existing factor is the main meteorological parameter, indicating the major liquid-phase/ heterogeneous reaction in the process of PM formation.
Entry MLR analysis demonstrates that coal-burning (mostly SO 2 ) and vehicle emission (mostly NO 2 ) contribute 48.6% and 33.9%, and 17.5% and 0.7% of the total PM 2.5 and PM 10 , respectively.This indicates that SO 2 from local coal-burning power plants is still the primarily polluting gas in Xi'an.
Trajectory cluster results of PM 2.5 at BRR indicate that the dominant regional sources in winter and spring could be contributed by the Westerly or winter monsoon invasion entrained urban pollutants.And the UV aerosol index verifies the source and pathway of a dust storm in spring.
b T = 2 meter temperature.c RH = 2 meter relative humidity.d PS = Surface pressure.e PBLH = Planetary boundary layer height.f Sum = Average concentration ± standard deviation of six sapling sites.g Av. ± std.= Seasonal average concentration ± standard deviation.

Fig. 3 .
Fig. 3.The diurnal variations of average concentrations of PM 2.5 and PM 10 from observed data of six sampling sites during 2010 in Xi'an.

Fig. 4 .
Fig. 4. Spatial distribution of annual average SO 2 (a) and NO 2 (b) via inverse distance interpolation of ArcGIS at urban area during 2010 in the main city area of Xi'an (grid with 50 × 50 m).

Fig. 5 .
Fig. 5. Clusters of 72-hours air mass backward trajectories arriving at 500 m above ground level at10:00 LST for (a) 12 to 25 January, (b) 14 to 27 April, (c) 12 to 25 July and (d) 12 to 25 October during 2010 at BRR reference sampling site in Xi'an during 2010. 1 (5/9, 83.3%) mean number of trajectory cluster (number of polluting trajectory/number of trajectory, percentage of polluting trajectory in cluster to total polluting trajectory).

Table 1 .
Seasonal average concentrations (mean concentration ± standard deviation) of five meteorological parameters, SO 2 , NO 2 , PM 2.5 and PM 10 from GDAS of NOAA, eleven automatic monitor stations and six sampling sites, respectively, during observed date in 2010.

Table 2 .
Stepwise multiple linear regression among PM 2.5 , PM 10 , SO 2 , NO 2 and six meteorological parameters, respectively, during the observed period in 2010.

Table 3 .
Cluster statistics result of 72-hours air mass backward trajectories arriving at 500m above ground level at18:00 LST (Local Standard Time) at BRR reference sampling site in Xi'an for (a) 12 to 25 January, (b) 14 to 27 April, (c) 12 to 25 July and (d) 12 to 25 October over 2010.Av. ± std.= Average concentration ± standard deviation.Cluster method: the Euclidean distance.Select pollution trajectory criteria: respectively seasonal average PM 2.5 concentration in BRR reference sampling site. a