CC BY-NC-ND
0Contents lists available at ScienceDirect
Atmospheric Environment
journal homepage: www.elsevier.com/locate/atmosenv
Process analysis of regional aerosol pollution during spring in the Pearl River Delta region, China
Qi Fan a, Jing Lan b'c, Yiming Liu a, Xuemei Wang a' *, Pakwai Chan d, Yingying Honga Yerong Feng e, Yexin Liu f, Yanjun Zeng f, Guixiong Liang f
a School of Environmental Science and Engineering, Sun Yat-sen University, Guangzhou 510275, China b Guangzhou Climate and Agrometeorology Center, Guangzhou 511430, China c Guangdong Ecological Meteorology Center, Guangzhou 510507, China d Hong Kong Observatory, Hong Kong 999077, China
e Guangdong Key Laboratory of Regional Numerical Forecast, Guangzhou 510080, China f Guangzhou Environmental Monitoring, Guangzhou 510030, China
HIGHLIGHTS
• The contributions of various physical and chemical processes to PM2 5 were quantified.
• The aerosol characteristics and their formation mechanisms during spring were explored.
• The contribution of local emissions and foreign sources from different cities to PM2 5 was analyzed.
CrossMark
ARTICLE INFO
Article history:
Received 19 January 2015
Received in revised form
2 September 2015
Accepted 2 September 2015
Available online 8 September 2015
Keywords:
WRF/SMOKE/CMAQ model system PM2.5
Process analysis
ABSTRACT
A numerical simulation analysis was performed for three air pollution episodes in the Pearl River Delta (PRD) region during March 2012 using the third-generation air quality modeling system Models-3/ CMAQ. The results demonstrated that particulate matter was the primary pollutant for all three pollution episodes and was accompanied by relatively low visibility in the first two episodes. Weather maps indicate that the first two episodes occurred under the influence of warm, wet southerly air flow systems that led to high humidity throughout the region. The liquid phase reaction of gaseous pollutants resulted in the generation of fine secondary particles, which were identified as the primary source of pollution in the first two episodes. The third pollution episode occurred during a warming period following a cold front. Relative humidity was lower during this episode, and coarse particles were the major pollution contributor. Results of process analysis indicated that emissions sources, horizontal transport and vertical transport were the primary factors affecting pollutant concentrations within the near-surface layer during all three episodes, while aerosol processes, cloud processes, horizontal transport and vertical transport had greater influence at approximately 900 m above ground. Cloud processes had a greater impact during the first two pollution episodes because of the higher relative humidity. In addition, by comparing pollution processes from different cities (Guangzhou and Zhongshan), the study revealed that the first two pollution episodes were the result of local emissions within the PRD region and transport between surrounding cities, while the third episode exhibited prominent regional pollution characteristics and was the result of regional pollutant transport.
© 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
The rapid urbanization and industrialization of China in recent
* Corresponding author. E-mail address: eeswxm@mail.sysu.edu.cn (X. Wang).
years has caused complex regional atmospheric pollution episodes to occur in various megalopolises (Chan and Yao, 2008; Liao et al., 2014). For example, widespread fogs and hazes were reported in the central and eastern regions of China in mid-January 2013, and these resulted in severe air pollution in North China, Central China and the Pearl River Delta (PRD) region. The development of
http://dx.doi.org/10.1016/j.atmosenv.2015.09.013
1352-2310/© 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
pollution has attracted global attention because of the extent of the affected area, the duration of the events and the high concentration of pollutants involved. However, meteorological fields, emissions characteristics and the spatial distribution of emissions sources affect pollution processes in various degrees, resulting in the generation of distinctive pollution profiles for different megalopolises. Because the PRD region is the most prominent city group in South China, some studies have focused on regional aerosol pollution within the PRD region. The China Meteorological Administration (CMA) has set up a monitoring station in Panyu, Guangzhou to observe atmospheric composition; and both Wu et al. and Deng et al. conducted a series of studies based on these observations (Wu et al., 2005, 2007; Deng et al., 2008). Low visibility is always accompanied by aerosol pollution, which has a serious impact on traffic and affects people's daily lives. Research indicates that the reduction in visibility across the PRD region since 1980 is the result of an increase in fine particulate matter concentrations stemming from human activities and photochemical reactions (Deng et al., 2008). PM2.5 (particles with an aerodynamic diameter of 2.5 mm or less) account for 58%—77% of PM10 (particles with an aerodynamic diameter of 10 mm or less) that occur during low visibility events, which is a significantly higher ratio than was measured 15 years ago. China's Program 973, "The Program of Regional Integrated Experiments on Air Quality over the Pearl River Delta (PRIDE-PRD)" (2002CB410801 and 2002CB211605), performed tests in 2004 and 2006 to improve out-field observations in the PRD region (Zhang et al., 2008). These tests included research into the characteristics of the atmospheric boundary layer (Fan et al., 2008, 2011), an aircraft aerial survey (Wang et al., 2008), and improvements in ground observations and aerosol radiative properties observations (Garland et al., 2008; Liu et al., 2008). The aforementioned research has contributed significantly to our understanding of the characteristics of aerosol pollution in the PRD region.
Air quality numerical models are useful tools for studying the mechanisms of regional aerosol formation. Much research has been devoted to the localized application of the Models-3/CMAQ (Community Multi-scale Air Quality) model in the PRD region. For example, Feng et al. (2007) used the CMAQ model to study a severe air pollution process in the PRD region under the influence of Typhoon Melor on November 1—3, 2003. The results indicated that downdrafts at the outer edges of the typhoon brought dry, warm air to the region, increased the stability of the lower atmosphere and caused a reduction in the boundary layer height, thus resulting in the accumulation of pollutants. Quan et al. (2008a), Quan and Zhang, (2008) used the CMAQ model to investigate the impact of ammonia and sulfate emissions on sulfur transformation and deposition in China. Meanwhile, Chen et al. (2009) used the CMAQ model to explore the influence of local sea-land breeze circulation on the regional haze process. However, each of these studies has focused on pollution events occurring in summer and autumn; and there has been relatively little research focused on the occurrence of spring events, even though spring is when the climate characteristics in South China shift from winter monsoon to summer monsoon conditions. Therefore, March 2012 was selected as the subject of this paper, and a detailed simulation study was performed to identify the pollution processes occurring in the PRD region during this transformational season.
The process analysis function within the CMAQ model is a widely employed quantitative method used to analyze the contributions of aerosols from different physical and chemical processes to pollution events (Jiang et al., 2003; Xu et al., 2008; Zhang et al., 2009a,b). Liu and Zhang (2011) used the CMAQ model and its process analysis function to analyze a regional fine particulate matter pollution event in the U.S. The results showed that aerosol
processes and emissions sources were major positive contributors to PM25 concentrations and its secondary components, while horizontal and vertical transport and dry deposition were negative contributors. Cloud processes favored to the formation of PM2.5 and sulfate but played a clearing role for nitrate and ammonium. Liu et al. (2010) used the CMAQ model to simulate ozone and particulate matter concentrations in China during January, April, July and October. Their process analysis indicated that the main factors governing PM10 concentrations were emissions sources and aerosol processes, while horizontal transport governed the clearing process. The study also highlighted the chemical complexity of ozone and particulate matter and suggested that distinct emissions control strategies should be developed for different areas based on the specific and seasonal characteristics of each particular area. Huang et al. (2006) used a CMAQ model called PATH (Pollutants in the Atmosphere and their Transport over Hong Kong) to simulate an ozone process over Hong Kong. The results showed that 30% of the ozone was generated during local chemical processes while the other 70% was generated by transport; ground-level ozone levels were affected by advection, vertical transportation, photochemical reactions and deposition. In summary, the aforementioned studies illustrate that the characteristics of air pollution and its major contributors vary by region and season. The quantitative study of how various processes contribute to aerosol concentrations supports the analysis of the impacts of local emissions and foreign sources on air quality, thereby providing useful input to further the development of local emissions control strategies.
The main purposes of this paper are to explore aerosol characteristics and their formation mechanisms during spring in the PRD region; to quantify the contributions of various physical and chemical processes using process analysis; and to analyze the contribution of local emissions and foreign sources from different cities to pollution. Section 2 introduces the model's settings and process analysis; Section 3 describes the March 2012 cases; Section 4 assesses the simulation's outputs; Section 5 shows the results of the process analysis and Section 6 provides the conclusions.
2. Model settings and process analysis
2.1. Model settings
This paper used the WRF/SMOKE/CMAQ model system. The modeling domains were configured using the Lambert projection, with a triple-nested grid (Fig. 1a) centered at 23°N 113°E. The domains had 36/12/4 km horizontal resolutions and 24 vertical layers (the altitudes of the lower 15 layers are 25 m, 65 m, 120 m, 200 m, 280 m, 400 m, 560 m, 730 m, 900 m, 1065 m, 1240 m, 1415 m, 1590 m, 1820 m and 2100 m, respectively). The results of the third domain (4 km resolution) were used in this study. The simulation was run for the period from February 25 to March 28, 2012. The outer solid lines in Fig. 1a denote the WRF grid domain, while the inner dashed lines denote the CMAQ grid domain; the third-layer grid covered the majority of the 9 cities located in the PRD region (Guangzhou, Foshan, Dongguan, Zhaoqing, Zhongshan, Jiangmen, Huizhou, Shenzhen and Zhuhai). Terrain data used in the model was taken from the topographical data revised from the 2004 MODIS satellite data.
The physical parameter settings used in the WRF model were: the WSM6 scheme for microphysics, the Kain-Fritsch scheme for cumulus parameters, the YSU scheme for the planetary boundary layer, the RRTM scheme for longwave radiation and the Goddard scheme for shortwave radiation. An Urban Canopy Model was added to the simulation, and a Noah Land Surface model was used as the corresponding land surface parameter.
Anthropogenic and biogenic emissions were considered in this
Fig. 1. (a) Triple-nested modeling domains for WRF (solid square frame) and CMAQ (dashed square frame) models. (b) The 4 km grid WRF domain pointed with 9 cities in Pearl River Delta region (The shaded is the terrain height, unit is m). (c) Spatial distribution of total PM25 emissions (ton) in March generated using the SMOKE model. The black circles indicated the locations of stations used to evaluate the chemical species: 1. Huadu station; 2. Zhenlong station; 3. Gongyuanqian (GYQ) station; 4.86 Middle School station; 5. Panyu Middle School station. The black triangles indicated the locations of stations used for process analysis: 1. Guangzhou station; 2. Zhongshan station.
study. The anthropogenic emissions include point, area and mobile emissions and were processed using the Sparse Matrix Operator Kernel Emissions (SMOKE) model. This emissions inventory was performed by Feng (2006), and the data were processed using a top-down method. The point emissions include point sources from Guangdong province and its surrounding provinces. Area emissions included emissions from cooking, schools, residents, etc. Mobile emissions were calculated using the vehicle miles travelled (VMT) method. The inventories contain the major pollutants: CO, NOx, VOC, SO2, NH3, PM10 and PM2.5. Biogenic emissions were generated from the Model of Emissions of Gases and Aerosols from Nature (MEGAN). The annual emissions in the inventory file were processed using the SMOKE model to identify hourly emissions, a speciation matrix and a gridding matrix. The merge step then combined the point, area, mobile and biogenic results to create model-ready emissions data. The calculated surface PM2.5 emissions in Domain 1-3 were 1638 Tg/year, 420 Tg/year and 189 Tg/ year, respectively. Compared with the Intercontinental Chemical Transport Experiment-Phase B (INTEX-B, Zhang et al., 2009a,b) in 2006 and Multi-resolution Emission Inventory for China (MEIC, He, 2012) in 2010, PM25 emissions in this inventory was slightly higher, but it may actually be in reasonable agreement since it was reported that INTEX-B under-estimated OC and BC emissions due to the larger uncertainties of small industries, residential combustion and transportations (Zhang et al., 2009a,b). This emissions inventory has been used in many studies (Chen et al., 2009; Fan et al., 2013; Feng et al., 2007) and has proved to be reliable for modeling. Fig. 1c shows the spatial distribution of PM2 5 emissions generated using the SMOKE model. Based on this figure, the main emissions sources appear to be situated in Guangzhou, Foshan and Shenzhen.
2.2. Process analysis
Process analysis (PA) is a diagnostic tool provided in the CMAQ model, and it includes IPR (Integrated Process Rate) and IRR
(Integrated Reaction Rate) analysis. The IPR analysis accounts for changes in different pollutants, while the IRR analysis accounts for different precursor sources of secondary pollutants. This paper used the IPR analysis, and its schematic equation is as follows (Byun and Ching, 1999):
VC VC 9C, H u—1 + v—L + w—1 -Vt \dx Vy Vz
V (kJC) + ^
+è№
+ (va) + (va) + m
\Vt ) aero \Vt ) chem
+ m + + ADJC
\Vt J drydep \Vt J emis
The IPR analysis isolates changes in pollutant concentrations into inputs from seven different types of physical and chemical processes and includes a mass conservation adjustment. The seven types of physical and chemical processes are: horizontal transport (HORI), vertical transport (VERT), aerosol process (AERO), chemical reaction process (CHEM), cloud process (CLDS), dry deposition (DDEP) and emissions source (EMIS). It should be noted that HORI is defined as the sum of horizontal advection and horizontal diffusion, while VERT is defined as the sum of vertical advection and vertical diffusion.
2.3. Simulation case
Fig. 2 shows the time series of daily average visibility, PM10 and PM2.5 concentrations, and the ratio of PM2.5/PM10 recorded at Gongyuanqian (GYQ) monitoring station in Guangzhou during March 2012. GYQ is an urban station in Guangzhou city located at 23.13°N, 113.26°E (showed in Fig. 1c). According to the aerosol concentrations recorded, three aerosol pollution episodes occurred in March of 2012: March 1—5, 13—17 and 25—26, respectively.
Fig. 2. Daily average visibility (km), PM2.5 and PMi0 concentrations (mg/m3), and PM2.5/ PM10 (%) at GYQ station in Guangzhou during March 2012. The shaded vertical areas denote three pollution episodes.
During the first two episodes, the PRD region was under the influence of southeasterly warm and wet air flows; temperatures rose and water vapors were abundant in the air, and thus relative humidity exceeded 80% (Fig. 3). The third pollution episode occurred during a warming period following a cold front, in which a northerly continental air mass moved in after the cold front; relative humidity was below 50%. High relative humidity creates favorable conditions for a liquid phase reaction to occur and leads to the transformation of more gaseous pollutants into particulate pollutants. In addition, particulates expand by absorbing moisture under high relative humidity, resulting in an increase in the light extinction coefficient and thereby reducing atmospheric visibility. Mists accompanied by southerly air flows further reduce visibility; therefore, visibility during the first two air pollution episodes was lower compared to the third episode. Fig. 2 also shows that PM10 increased during the three pollution episodes, with peak values of 216 mg/m3,191 mg/m3 and 165 mg/m3, respectively. However, PM2.5 levels increased more during the first two episodes, with peak values of 195 mg/m3 and 114 mg/m3 and PM2.5/PMi0 ratios that reached 90% and 60%, respectively. These data indicate that the first two pollution episodes were composed mainly of fine particles. Although the third episode had a high PM10 value, its peak PM2.5 and PM2.5/PM10 ratio only reached 55 mg/m3 and 39%, respectively;
thus, pollution during this episode consisted mainly of coarse particles.
3. Model assessments
3.1. Assessment of meteorological fields
Meteorological observation data used in the WRF model assessment came from ground observations taken at automatic monitoring stations and provided by Guangdong Provincial Meteorological Station. In this study, 9 cities in the PRD region were selected for study. The simulation results were compared to observation data and averages (M, O), the correlation coefficient (r), average deviation (MB), root mean square error (RMSE), normalized mean bias (NMB), normalized mean error (NME) and index of agreement (IOA) were calculated.
Fig. 3 shows the comparison between the model simulation and March 2012 measured data from GYQ monitoring station. The temperature series showed that there were three warming periods and two cooling periods in the PRD region during March 2012. The three pollutant episodes occurred during these three warming periods, respectively. The temperature pattern from the WRF simulation was consistent with measured data. However, simulated temperatures were slightly higher than measured values, especially from March 10-17, when simulated values were an average 17.5% higher. The relative humidity series showed that before March 23, the PRD region was under the influence of a southeasterly warm, wet air flow, which maintained relatively high humidity levels. There was a brief drop in humidity due to a cold front on March 10, but as wind direction changed in the PRD region (from northerly winds to southerly winds) relative humidity rapidly rose back up to 90%. With relation to wind field, the WRF model generally reflected changes in wind direction in the PRD region throughout the month of March. However, there was a positive bias associated with simulated wind speed, as the average measured wind speed was 2.49 m/s and the average simulated wind speed was 4.09 m/s. This bias is likely attributed to the underestimation of urban areas' impact on wind speed in the model. Despite that, the WRF model generally described the changes in wind speeds throughout the
Fig. 3. Simulated and measured weather data at GYQ station during March 2012.
Table 1
Statistical comparison between the average simulated and measured data for meteorological stations in 9 cities in PRD region during March 2012.
Variable Measured average Simulated average r MB RMSE NMB NME
Temperature (°C) 18.19 19.59 0.94 1.40 2.02 0.08 0.09
Relative humidity (%) 81.12 76.03 0.84 -5.10 10.88 -0.06 0.11
Wind speed (m/s) 2.49 4.09 0.77 1.59 1.89 0.71 0.73
month of March, and it was able to reflect the increase in wind speed after a cold front passed. Table 1 shows the statistical comparisons between WRF simulated data and observations for various meteorological variables in 9 cities during March 2012. The simulation best modeled temperature, where correlation coefficient was 0.94, and simulated average was 1.4 ° C higher than actual average, along with NME of 8% and RMSE of 2.02. It also achieved a high correlation coefficient of 0.84 for relative humidity and its simulated average was 6% smaller than measured average, along with NME of 11% but a higher RMSE at 10.88. Meanwhile, although correlation coefficient for wind speed was higher at 0.77, simulated average was larger than measured average by 1.59 m/s. Therefore, the model was able to reflect actual weather conditions and provided relatively reliable weather field inputs for the CMAQ model.
3.2. Assessment of concentration fields
Pollutant concentrations were assessed at 5 monitoring stations. The locations of these stations were shown in Fig. 1c. The data were provided by the Guangzhou monitoring station. The assessed chemical pollutants included PM2.5, SO2, NO2 and O3. SO4- and NO-concentrations are also monitored by GYQ station, and data on these compounds was used to evaluate the model in this study. Fig. 4 shows the comparisons between simulated aerosol concentrations and measured data measured from the GYQenvironmental monitoring station. Model simulations yielded good results for PM2.5 and SO24- and generally reflected the fluctuations in measured data during March 2012. The CMAQ model over-predicted NO- concentrations prior to March 14, while it under-
predicted NO3- concentrations thereafter. Compared to SO24- and PM2.5, the bias observed for NO3- in the simulation was relatively large. Table 2 lists the statistical comparisons between the CMAQ simulations and observations for each pollutant taken from the 5 monitoring stations during March 2012. In general, IOA coefficients showed that simulated values exhibited a large degree of agreement with measured data, as IOAs were all above 0.75. However, the CMAQ model had a negative bias when simulating pollutant concentrations, as indicated by negative MB values for PM2 5, SO2 and NO2. On the other hand, O3 concentrations were over-predicted, with an MB of 2.8 mg/m3. The underestimation of pollutant concentrations resulted partly from the uncertainty of emissions values generated by the SMOKE model. On the other hand, the overestimation of wind speed in the WRF model may have helped the diffusion of pollutants and resulted in lower predicted pollutant concentrations in the CMAQ model. As for the underestimation of NO2, this may be related to the reaction rate between NOx and O3.
The PRD Regional Air Quality Index (RAQI) was developed by the Environmental Protection Bureau of Guangdong Province and the Environmental Protection Department of the Hong Kong Special Administrative Region. The RAQI uses five grades to identify air quality in the PRD region. The lower the grade, the better the air quality. This grade is calculated according to the measured air concentrations of major pollutants, including PM10, SO2, NO2 and O3. Fig. 5 shows the distribution of RAQI values and simulated PM2 5 values during the three pollution episodes in March 2012. The RAQI distribution graphs indicate that the first two episodes involved primarily localized pollution that affected a limited area; in these
SO42-(^g/m3)
NO^g/m3)
40 30 20 10
3/1 3/4 3/7 3/10 3/13 3/16 3/19 3/22 3/25
■ obs Bsim
20 15 10 5
3/1 3/4 3/7 3/10 3/13 3/16 3/19 3/22 3/25
■ obs ■ sim
PM25(^g/m3)
300 200 100 0
JjjJri
Ihhilill
3/1 3/4 3/7 3/10 3/13 3/16 3/19 3/22 3/25
■ obs Bsim
Fig. 4. Simulated and measured data of sulfate, nitrate and PM2.5 concentrations at GYQ station during March 2012.
Table 2
Statistical comparison between the average simulated and measured data for 5 environmental monitoring stations in PRD region during March 2012.
Pollutant Measured average Simulated average 1OA MB RMSE NMB NME
PM2.5 (mg/m3) 55.0 47.5 0.85 -7.5 44.4 -0.13 0.59
SO2 (mg/m3) 24.5 20.1 0.76 -4.4 25.3 -0.16 0.72
NO2 (mg/m3) 55.3 49.9 0.89 -5.4 37.3 -0.06 0.53
O3 (mg/m3) 32.2 35.0 0.75 2.8 40.1 -0.18 0.98
episodes pollution was concentrated in cities with numerous emissions sources (Guangzhou and Foshan) and the downwind region (Zhaoqing). In contrast, the third episode affected a much larger area. In that scenario, pollutants were transported and diffused throughout a greater range, resulting in level III RAQI values that covered most of the PRD region. As seen in the simulated PM2.5 distribution graph, the first two pollution episodes were influenced by southeasterly winds, with higher particulate matter concentrations occurring in the northwest part of the region. On March 4, the highest PM2 5 values were located in Guangzhou and Foshan. A more northerly convergence zone on March 15 shifted the highest PM2.5 values northwestward so that they were located in Zhaoqing and Foshan. In summary, the distribution of particulate matter during the first two pollution episodes was more concentrated and localized. During the third episode, northeasterly winds controlled dispersion, and thus higher PM2.5 values were recorded in the western PRD region. Heavy pollution covered a much larger area in the third episode and included Foshan, Zhongshan, Zhuhai, Jiangmen and other cities. In all, the CMAQ model generally simulated the changes in pollutant concentrations well and provided a reliable concentration field.
4. Process analysis
Given that PM2 5 is the major factor affecting visibility, and the model produced good simulation results for PM2.5, the researchers chose to perform a process analysis on simulated PM2.5 values. From the time—height profile at GYQ station (Fig. 6a), simulated PM2.5 was mainly concentrated within the lower atmospheric boundary layer (below the seventh layer (560 m) of the model). PM2.5 concentrations generally decreased as height increased and decreased rapidly above the seventh layer. An analysis of the evolution of PM2.5 values over time demonstrated that concentrations were greatest during the three pollution episodes. Fig. 6b shows the evaluation of the backscattering signal of Lidar measured at the Hong Kong site (obtained from Hong Kong University of Science and Technology's website). It was also shown that aerosol concentrations were higher in the planetary boundary layer and declined with increasing height. The backscattering signal was also significant during the three pollution episodes. The CMAQ model simulated a reasonable vertical distribution of PM2.5 concentrations.
Fig. 7 is a time series chart showing the breakdown of PM2 5 contributions from various physical and chemical processes in the near-surface layer and at approximately 900 m above the Guangzhou station during March 2012. As shown in Fig. 7a, near surface emissions sources were the primary contributors to air pollution throughout the month, consistently contributing 38.8% of total PM25 levels. Aerosol processes, primarily homogeneous nucleation and condensation, also contributed an average 4 mg/m3/h to total PM2 5 levels. Meanwhile, vertical transport, dry deposition and cloud processes reduced total PM2.5 concentrations throughout the month. To elaborate, particulate matter concentrations decreased with increasing height. Vertical transport carried lower-layer pollutants upward and therefore had a negative effect on pollution concentrations at the surface layer. Dry deposition is always a
negative contributor, but its magnitude depends on particulate concentration and the rate of dry deposition. The clearing mechanism of cloud processes reduces pollutant concentrations and, on average, resulted in a 4 mg/m3/h reduction in pollution levels. On the other hand, horizontal transport can either be a positive or negative contributor, depending on the flow field during a given day and PM2.5 concentrations in the surrounding areas. Overall, emissions sources and vertical transport were the two major contributors that impacted near-surface-layer PM2.5, while horizontal transport, aerosol processes, cloud processes and dry deposition had smaller impacts and contributed a combined 22.1% to pollution levels. Compared with other processes, gaseous chemical processes had a negligible effect on PM2.5 concentrations.
Pollution characteristics at high altitude were drastically different from near-surface layers. PM2.5 concentrations decreased with increasing height, and thus the absolute values of process contributions were smaller. As a result, the contribution of emissions sources at high altitude was negligible. Fig. 7b shows that PM25 concentrations at approximately 900 m were primarily related to cloud processes, aerosol processes, horizontal transport and vertical transport, with contribution rates of 29.9%, 25.2%, 24.7% and 20.2%, respectively. Cloud processes were a major positive contributor, especially during the first two pollution episodes because high water vapor content under warm southerly flow conditions is conducive to an aqueous chemical reaction that results in increasing secondary aerosol production. Meanwhile, aerosol processes were a major negative contributor, likely because atmospheric NaCl was neutralized by substantial amounts of HNO3, such that Cl- originally present in the particulates was converted to gaseous HCl. This resulted in a reduction in particulate levels. Another major negative contributor was horizontal transport, which diffused particulate matter concentrations to surrounding areas through the upper wind field after they were carried to high altitude from the lower layers. Vertical transport had an inverse relationship with horizontal transport in terms of pollution levels at high altitudes. Horizontal convergence (divergence) corresponded to diffusion (supplement) in the vertical direction; thus, horizontal and vertical transport had opposite impacts on changes in upper-layer pollutant concentrations.
Given that weather conditions during the first two episodes were similar, the results of the second and third episodes are provided for comparison in the following discussion. Vertical PA profiles from the second and third pollution episodes at Guangzhou station (an urban site; 113.25°E, 23.12°N, shown in Fig. 1c) clearly show that contributions from emissions sources were concentrated in the first three layers above the ground and decreased significantly with increasing height. Consequently, vertical transport was a negative contributor in the first three layers and became positive within the boundary layer (fourth layer) and above (Fig. 8a and b). Horizontal transport had a significantly greater impact during the third pollution episode than in the first two, causing more severe pollution in downstream areas from cities with significant emissions (Guangzhou and Foshan) during the third episode. Dry deposition only had an impact on pollution levels in the first layer, with a contribution of over -7 mg/m3/h. This was because the model treated dry deposition as a bivariate variable and integrated
Fig. 5. Spatial distribution of Regional Air Quality Index (RAQI) values and simulated PM2.5 concentrations (Unit: mg/m3) during pollution processes in March 2012. (a, c, e denotes PRD RAQI distribution on March 4, March 15 and March 25, respectively; b, d, f denotes CMAQ simulated PM2.5 distribution on March 4, March 15 and March 25, respectively).
it over the entire atmospheric column; hence, values were only generated in the first layer. Cloud processes contributed more to pollution levels during the second pollution episode. In addition, because vapor content was abundant above 200 m, aqueous oxidation reactions caused an increase in PM25; therefore the contribution of cloud processes to pollution levels was greater at
upper levels than lower levels and was positive. Meanwhile, cloud processes were had a negligible effect during the third pollution episode because of the low vapor content present at during the event.
To further understand the different causes and characteristics of pollution processes, the PA results for Zhongshan station (a
n 1.60000 0.80000 0.40000 0.20000 0.10000 0.05000 0.02500 0.00001 Backscattering signal
Fig. 6. (a) Simulated PM25 concentration time-height profile at GYQ station (Unit: mg/m3). The black line refers to the corresponding planetary boundary layer height. (b) Evaluation of the backscattering signal of Lidar measured at the Hong Kong site.
80 60 40
20 > 0
- -20 -40 -60 -80
1 3 5 7 9
11 13 15 17 19 21 23 25 27 Day
b II II llllljllllll III
II ll-lllll-lllllll 1"
9 11 13 15 17 19 21 23 25 27 Day
Fig. 7. Time series of the breakdown of PM25 contributions from various physical and chemical processes in the near-surface layer (a) and at approximately 900 m (b) above the Guangzhou station during March 2012 (Unit: mg/m3/h).
suburban site; 113.25° N, 22.71 °N, shown in Fig. 1c) were compared with those of Guangzhou (Fig. 8c and d). During the second pollution episode, southerly warm air flows resulted in southerly winds throughout the PRD region. In both cities, emissions sources positively affected pollution by contributing approximately 23 mg/ m3/h of pollutants, while dry deposition and vertical transport reduced pollution levels by approximately 10 mg/m3/h in the near-surface layer. Because relative humidity was high, cloud processes had a greater effect, contributing large amounts of pollutants through the creation of secondary aerosols during liquid phase reactions. On March 15, cleaner air from South China Sea diluted PM2.5 concentrations in Zhongshan; as a result, horizontal transport reduced pollution levels in the near-surface layer in Zhong-shan. Guangzhou is located downwind from Zhongshan, and particulates from Zhongshan, Zhuhai, Shenzhen and other upwind cities were transported to Guangzhou through southerly winds; therefore, horizontal transport increased pollution levels in the near-surface layer at Guangzhou. However, by comparing these values to contributions from local emissions, the analysis revealed
that horizontal transport had a relatively lower impact. Fig. 6a shows the temporal variation in the planetary boundary layer height (PBLH) at GYQ station. The PBLH during the first two episodes (less than 1200 m) was lower than the PBLH during the third episode (greater than 1800 m). A thinner boundary layer made the layer available for the accumulation of local emitted particles and resulted in increased PM2.5 concentrations. In summary, the second pollution episode was mainly comprised of local pollutants within the PRD region.
In contrast, the third pollution episode occurred during a warming period following a cold front. Northeasterly winds dominated the PRD region with lower relative humidity, and the PRD region exhibited extensive regional pollution features. The vertical profiles of Guangzhou (Fig. 8b) and Zhongshan (Fig. 8d) showed that horizontal transport in both cities positively contributed to pollution levels, which mean PM2.5 values were mainly influenced by sources from the northeastern PRD region and north of Guangdong Province. Because humidity was relatively low, cloud processes had a low impact in both cities, and PM2.5 concentrations were mainly governed by emissions sources, transport, dry deposition and aerosol processes.
Total contribution across layers on both March 15 and March 26 at Guangzhou station (Fig. 9a) and Zhongshan station (Fig. 9b) show that emissions sources were the most significant positive contributors to pollution levels. In Fig. 9a, there were negligible differences between emissions sources on both days, and the daily emission total was approximately 29 mg/m3/h on both days. Par-ticulate matter was released into the atmosphere and fell back to the surface through the dry deposition process, and the rates of dry deposition during these two pollution episodes were -8 mg/m3/h and -7 mg/m3/h, respectively. At the same time, atmospheric par-ticulate matter was also transported and diffused through vertical and horizontal transport processes. Total contributions from vertical transport were -31 mg/m3/h and -4 mg/m3/h, respectively; the variation in these figures stems from the fact that the boundary layer was thinner during the second pollution episode and partic-ulates were concentrated in near-surface layers. Thus vertical transport had a greater clearing effect for PM2.5 in that episode. On the other hand, the boundary layer was thicker during the third episode, and vertical mixing of PM2.5 was more uniform. Therefore, vertical transport had a smaller impact in that episode. In addition, given that the wind speed was faster during the third episode, the contribution from horizontal transport was greater and further reduced the contribution rate from vertical transport compared to the second episode. Horizontal transport contributed -6 mg/m3/h and -17 mg/m3/h during these two episodes, respectively, and mainly provided dilution and clearing effects. In the case of high
Fig. 8. Vertical profile of PM2.5 contributions from various physical and chemical processes at the Guangzhou (a, b) and Zhongshan (c, d) stations on March 15 (a, c) and March 26, 2012 (b, d).
Fig. 9. Total contributions to PM2.5 from various physical and chemical processes at the Guangzhou (a) and Zhongshan (b) stations on March 15 and March 26,2012 (Unit: mg/m3/h).
relative humidity, the positive contribution of cloud process became more prominent. Under this scenario, increasing secondary aerosols from liquid phase reactions resulted in increased partic-ulate matter levels. The total contribution from cloud processes during the second pollution episode was 24 mg/m3/h, which was only slightly less than the contribution from emissions sources.
Fig. 9b shows that emissions sources were also a primary PM2.5 contributor. With high relative humidity, cloud processes had a greater contribution during both days and contributed more than 17 mg/m3/h. In the case of low humidity, cloud processes contributed a negligible amount of pollution and could be ignored. Because Guangzhou has numerous emissions sources that produce a large quantity of local emissions, particulate matter concentrations were higher in general there; therefore, both horizontal and vertical transport had negative impacts on pollution levels during both pollution episodes and served to clearing PM2.5 in that city. On the other hand, in Zhongshan, a city with relatively clearer air, horizontal transport served as a PM2.5 clearing agent when there was light pollution in upwind areas. However, when pollution was severe in upwind areas, horizontal transport instead increased PM2.5 concentrations.
5. Conclusions and discussion
This paper used a third-generation air quality modeling system Models-3/CMAQ to perform a numerical simulation analysis of air
pollution processes in the PRD region during March 2012. The process analysis function was used to quantify various physical and chemical contributions to air pollution. Major conclusions are as follows:
(1) There were three pollution episodes during March 2012 and their formation mechanism varied. During the first two episodes, the PRD region was under the influence of southerly warm, wet air flows; and relative humidity was high. Thus, more secondary aerosols were generated through liquid phase reaction, resulting in lower visibility and increased fine particle concentrations during the first two episodes. Meanwhile, the third episode occurred during a warming period following a cold front; and humidity was relatively low. In the third episode, coarse particulates dominated, and there was higher visibility.
(2) The WRF/SMOKE/CMAQ model system performed fairly well in simulating pollution processes during March 2012. The WRF model was capable of capturing the variations in meteorological elements in the PRD region, including three warming periods and two cold fronts; however, wind speeds generated by the model had a positive bias. The CMAQ model performed better simulating PM25, SO2 and NO2 concentrations and was able to reproduce their variation patterns.
(3) The process analysis showed that particulate matter concentrations were affected by different processes at the
surface and high altitudes. Emissions sources, vertical transport and horizontal transport were the primary contributors to pollutant concentrations in the lower layers, while vertical and horizontal transport, aerosol processes and cloud processes were the major contributors at a height approximately 900 m above the ground. Cloud processes had a greater positive contribution to PM2.5 when relative humidity was high.
(4) By comparing process analyses for different cities, this paper demonstrated that during the second pollution episode, the weather was dominated by southerly warm air flows and significant pollution concentrations were located in Guangzhou, Foshan and Zhaoqing areas. This pollution stemmed primarily from local emissions. Meanwhile in Zhongshan, horizontal transport via warm southerly air flows significantly reduced pollution. During the third pollution episode, the PRD region was dominated by northeasterly winds, such that emissions sources in northeastern area and outside of Guangdong Province had a significant influence on pollution distribution and significantly expanded the extent of areas with severe pollution throughout the PRD region. Zhongshan was downwind of Guangzhou during the third episode, and horizontal transport contributed significantly to pollution levels.
Acknowledgments
This research was supported by funds from the National Natural Science Foundation (41275100), the China Special Fund for Meteorological Research in the Public Interest (GYHY201306042 and GYHY201406031), the National Natural Science Foundation of Guangdong Province as key project (S2012020011044) and the European Union FP7 project PANDA (3206429). This work was also partly supported by the Jiangsu Collaborative Innovation Center for Climate Change and the high-performance grid-computing platform of Sun Yat-sen University.
References
Byun, D.W., Ching, J.K.S., 1999. Science Algorithms of the EPA Models-3 Community Multi-scale Air Quality (CMAQ) Modeling System. U.S. Environmental Protection Agency, U.S. Government Printing Office, Washington, DC. Chan, C.K., Yao, X., 2008. Air pollution in mega cities in China. Atmos. Environ. 42 (1), 1-42.
Chen, X.L., Feng, Y.R., Li, J.N., Lin, W.S., Fan, S.J., Wang, A.Y., Fong, S.K., Lin, H., 2009. Numerical simulation on the effect of sea-land breezes on atmospheric haze over the Pearl River Delta region. Environ. Model Assess. 14, 351-363. Deng, X.J., Tie, X.X., Wu, D., 2008. Long-term trend of visibility and its characterizations in the Pearl River Delta Region (PRD), China. Atmos. Environ. 42, 1424-1435.
Fan, S.J., Wang, B.M., Tesche, M., Engelmann, R., Althausen, A., Liu, J., Zhu, W., Fan, Q., Li, M.H., Ta, N., Song, L.L., Leong, K.C., 2008. Meteorological conditions and structures of atmospheric boundary layer in October 2004 over Pearl River Delta area. Atmos. Environ. 42, 6174-6186. Fan, S.J., Fan, Q., Yu, W., Luo, X.Y., Wang, B.M., Song, L.L., 2011. Atmospheric boundary layer characteristics over the Pearl River Delta, China, during the summer of 2006: measurement and model results. Atmos. Chem. Phys. 11,
6297-6310.
Fan, Q., Yu, W., Fan, S.J., Wang, X.M., Lan, J., Zou, D.L., Yu, W., Feng, Y.R., Chan, P.W., 2013. Process analysis of a regional air pollution episode over Pearl River Delta Region, China, using the MM5-CMAQ model. J. Air & Waste Manag. Assoc. 64, 406-418.
Feng, Y.R., 2006. Mechanical analyses and numerical simulations on the aerosol pollution over the Pearl River Delta. Sun Yat-sen University, Guangzhou, China. PhD thesis.
Feng, Y.R., Wang, A.Y., Wu, D., Xu, X.D., 2007. The influence of tropical cyclone Melor on PM10 concentrations during an aerosol episode over the Pearl River Delta region of China: numerical modeling versus observational analysis. Atmos. Environ. 41, 4349-4365.
Garland, R.M., Yang, H., Schmid, O., Rose, D., Nowak, A., Achtert, P., Wiedensohler, A., Takegawa, N., Kita, K., Miyazaki, Y., Kondo, Y., Hu, M., Shao, M., Zeng, L.M., Zhang, Y.H., Andrea, M.O., 2008. Aerosol optical properties in a rural environment near the mega-city Guangzhou, China: implications for regional air pollution, radiative forcing and remote sensing. Atmos. Chem. Phys. 8, 5161-5186.
He, K.B., 2012. Multi-resolution Emission Inventory for China (MEIC): Model Framework and 1990-2010 Anthropogenic Emissions. International Global Atmospheric Chemistry Conference, 17-21 September, Beijing, China.
Huang, J.P., Fung, J.C.H., Lau, A.K.H., 2006. Integrated processes analysis and systematic meteorological classification of ozone episodes in Hong Kong. J. Geophys. Res. 111, D20309.
Jiang, G., Lamb, B., Westberg, H., 2003. Using back trajectories and process analysis to investigate photochemical ozone production in the Puget Sound region. Atmos. Environ. 37, 1489-1502.
Liao, J., Wang, T., Wang, X., Xie, M., Jiang, Z., Huang, X., Zhu, J., 2014. Impacts of different urban canopy schemes in WRF/Chem on regional climate and air quality in Yangtze River Delta, China. Atmos. Res. 145-146, 226-243.
Liu, P., Zhang, Y., 2011. Use of a process analysis tool for diagnostic study on fine particulate matter predictions in the U.S. Part II: analyses and sensitivity simulations. Atmos. Poll. Res. 2, 61-71.
Liu, X.G., Cheng, F., Zhang, Y.H., Jung, J.S., Sugimoto, N., Chang, S.Y., Kim, Y., Fan, S.J., Zeng, L.M., 2008. Influences of relative humidity and particle chemical composition on aerosol scattering properties during the 2006 PRD campaign. Atmos. Environ. 42, 1525-1536.
Liu, X.H., Zhang, Y., Xing, J., Zhang, Q., Wang, K., Streets, D.G., Jang, C., Wang, W.X., Hao, J.M., 2010. Understanding of regional air pollution over China using CMAQ. Part II: process analysis and sensitivity of ozone and particulate matter to precursor emissions. Atmos. Environ. 44, 3719-3727.
Quan, J., Zhang, X., 2008. Assessing the role of ammonia in sulfur transformation and deposition in China. Atmos. Res. 88 (1), 78-88.
Quan, J., Zhang, X., Zhang, Q., Guo, J., Vogt, R.D., 2008. Importance of sulfate emission to sulfur deposition at urban and rural sites in China. Atmos. Res. 89 (3), 283-288.
Wang, W., Ren, L.H., Zhang, Y.H., Chen, J.H., Liu, H.J., Bao, L.F., Fan, S.J., Tang, D.G., 2008. Aircraft measurements of gaseous pollutants and particulate matter over Pearl River Delta in China. Atmos. Environ. 42, 6187-6202.
Wu, D., Tie, X., Li, C., Ying, Z., Lau, A.K.H., Huang, J., Deng, X., Bi, X., 2005. An extremely low visibility event over the Guangzhou region: a case study. Atmos. Environ. 39, 6568-657 .
Wu, D., Bi, X., Deng, X., Li, F., Tan, H., Liao, G., Huang, J., 2007. Effect of atmospheric haze on the deterioration of visibility over the Pearl River Delta. Acta Meteor. Sin. 21, 215-223.
Xu, J., Zhang, Y., Fu, J.S., Zheng, S., Wang, W., 2008. Process analysis of typical summertime ozone episodes over the Beijing area. Sci. Total Environ. 399 (1-3), 147-157.
Zhang, Y.H., Hu, M., Liu, S.C., Wiedensohler, A., 2008. The special issue on PRIDE-PRD2004 Campaign. Atmos. Environ. 42, 6155-6156.
Zhang, Q., Streets, D.G., Carmichael, G.R., He, K.B., Huo, H., Kannari, A., Klimont, Z., Park, I.S., Reddy, S., Fu, J.S., Chen, D., Duan, L., Lei, Y., Wang, L.T., Yao, Z.L., 2009a. Asian emissions in 2006 for the NASA INTEX-B mission. Atmos. Chem. Phys. 9, 5131-5153.
Zhang, Y., Wen, X.Y., Wang, K., Vijayaraghavan, K., Jacobson, M.Z., 2009b. Probing into regional O3 and PM pollution in the U.S. Part I: an examination offormation mechanisms through a process analysis technique and sensitivity study. J. Geophys. Res. 114, D22304.
