Ozone source apportionment (OSAT) to differentiate local regional and super-regional source contributions in the Pearl River Delta region, China

Authors

  • Y. Li,

    1. Division of Environment, Hong Kong University of Science and Technology, Hong Kong
    Search for more papers by this author
  • A. K.-H. Lau,

    Corresponding author
    1. Division of Environment, Hong Kong University of Science and Technology, Hong Kong
    2. Atmospheric Research Center, Fok Ying Tung Graduate School, Hong Kong University of Science and Technology, Hong Kong
    3. Pearl River Delta Atmospheric Environmental Research Joint Laboratory, Guangzhou, China
    • Corresponding author: A. K.-H. Lau, Atmospheric Research Center, Fok Ying Tung Graduate School, Hong Kong University of Science and Technology, Rm. 302B, Nansha IT Park, Nansha, Guangzhou 511458, China. (alau@ust.hk)

    Search for more papers by this author
  • J. C.-H. Fung,

    1. Division of Environment, Hong Kong University of Science and Technology, Hong Kong
    2. Atmospheric Research Center, Fok Ying Tung Graduate School, Hong Kong University of Science and Technology, Hong Kong
    3. Pearl River Delta Atmospheric Environmental Research Joint Laboratory, Guangzhou, China
    Search for more papers by this author
  • J. Y. Zheng,

    1. Pearl River Delta Atmospheric Environmental Research Joint Laboratory, Guangzhou, China
    2. College of Environmental Science and Engineering, South China University of Technology, Guangzhou, China
    Search for more papers by this author
  • L. J. Zhong,

    1. Atmospheric Research Center, Fok Ying Tung Graduate School, Hong Kong University of Science and Technology, Hong Kong
    2. Guangdong Provincial Environmental Monitoring Center, Guangzhou, China
    Search for more papers by this author
  • P. K. K. Louie

    1. Hong Kong Environmental Protection Department, Hong Kong
    Search for more papers by this author

Abstract

[1] It is well-known that ground-level ozone is not just a local or regional air quality problem; emission sources from super-regional (sources outside the PRD region) scales are known to contribute significantly to local ozone concentrations. However, source apportionment studies differentiating the relative contributions of local, regional, and super-regional ozone precursors are still limited. In this paper, using the Pearl River Delta (PRD) as an example, we have conducted a detailed apportionment (by source categories and by source regions) study of surface ozone using photochemical model source apportionment tools. Our results show that, while the super-regional contribution is dominant under mean ozone conditions, elevated local and regional sources are the causative factor for the formation of high ozone episodes. In particular, the local and PRD regional contributions increase from about 30% during non-episode days to about 50% during high ozone episode days in the autumn (November 2006) and even up to about 70% during high ozone episodes in the summer (July 2006). These results suggest that local and regional controls of ozone precursors are still very important for ozone reduction, particularly for episodic events. Furthermore, our results show that mobile emission is by far the highest contributing source category to ozone levels in the PRD for episodic ozone events. Moreover, we find substantial seasonal variations in the way ozone precursors from neighboring areas affect ozone levels in any particular city, suggesting that regional collaborations are important for developing effective long-term strategies to reduce ozone over the PRD region.

1. Introduction

[2] In recent decades, the rapid economic development and urbanization in the Pearl River Delta region (PRD), located in southern China, have already led to serious air quality deteriorations [Wang et al., 1998, 2001a, 2003; Zhang et al., 2008; Zheng et al., 2010]. The high ozone (O3) episodes in this region are widely recognized as one of its major air pollution issues in recent years [Wang et al., 1998, 2001a, 2003; Zhang et al., 2008; Zheng et al., 2010]. Ozone is formed by a series of chemical reactions involving nitrogen oxides (NOx) and volatile organic compounds (VOCs), in the presence of sunlight. These chemical reactions involving ozone formation and depletion may take a few hours, and cross over large spatial scales from tens to hundreds of kilometers with meteorological forcing. Therefore ozone pollution typically exhibits a regional characteristic, and the ozone levels are influenced by both local and regional emission sources [Li, 2011; Kleinman et al., 1997; So and Wang, 2003; Wang et al., 1998; Zhang et al., 2008; Zhang and Ying, 2011], even influenced by ozone and precursors from long-range transport at intercontinental scales [Fiore et al., 2002, 2009; Nagashima et al., 2010; Vingarzan, 2004; Wang et al., 2011; Wild and Akimoto, 2001; Wild et al., 2004]. For example, Tong et al. [2009]used air quality modeling to study the source-receptor relationship between NOxemission and ozone exposures over the United States (U.S.) and found that in-state emissions are responsible for less than 15% of O3exposures in 90% of the U.S. They suggested that the government of U.S. needs a regional strategy to effectively reduce ozone pollution. Identification of local (sources in the city areas), regional (other city sources within the region) or super-regional (all sources outside the study region) contributions is fundamental for designing effective ozone control strategies and has been carried out for different regions [Kemball-Cook et al., 2009; Lei et al., 2008; Mauzerall et al., 2005; Nagashima et al., 2010; Tong et al., 2009; Wang et al., 2009; Zhang and Ying, 2011; Wang et al., 2011]. The Hemispheric Transport of Air Pollution (HTAP, 2010, www.htap.org) community reported that the background ozone, including ozone from emissions outside the receptor regions and from emissions of natural origin, such as stratospheric and from biogenic precursors, has been shown to make significant contributions to surface ozone for mean ozone level in China [Wang et al., 2011] and many other regions [Fiore et al., 2002, 2009; Nagashima et al., 2010; Vingarzan, 2004; Wild and Akimoto, 2001; Wild et al., 2004]. We note that the background ozone used in these studies is a similar concept as the super-regional contributions in this paper, while the background ozone generally includes both emissions outside a large area (i.e., country or continent) and the natural sources inside, but super-regional referred here emphasizes the contributions from sources (both anthropogenic and biogenic) outside a polluted region (i.e., the PRD region).

[3] In Hong Kong (HK) and the PRD region, ozone concentration has an obvious seasonal variation. Seasonal averaged ozone is low both in winter and in summer, while it is relatively high in autumn, and there is also a second peak in spring. The averaged ozone of summer is generally low mainly associated with the dilution of fresh oceanic air mass brought by the southwest Asian monsoon. However, the highest ozone episode can also be observed in the summer which is associated with the travel of tropical cyclones at synoptic scale [Huang et al., 2006]. Therefore, in HK and the PRD region, high ozone episodes generally occur in summer and autumn seasons. Two major regimes with different synoptic patterns associated with ozone episodes were summarized by Huang et al. [2005, 2006]. The one regime is that tropical cyclones locate over the northwestern Pacific or the South China Sea to the east or southeast of the PRD region. These coming tropical cyclones can substantially worsen the air quality in southern China through subsidence, enhance stability and reduced dispersion ability of the atmosphere. This regime is found to be the most conducive weather pattern to the occurrence of high concentration ozone episodes in the region [Chan and Chan, 2000; Huang et al., 2005, 2006; Lee et al., 2002]. The other regime is the continental anticyclone presenting over northwestern or northeastern China to the north of the PRD region, which leads to clear sunny skies at the periphery of the high-pressure system over the PRD region. During these regimes, the surface wind is dominated by northerlies or northeasterlies and will transport pollutants from the Northeast of the PRD region to the Southwest of the PRD region. Ozone episodes associated with these major regimes will be studied with source apportionment method in the paper.

[4] Most studies of source contributions conducted in HK and the PRD region have examined the local and non-local source contributions for a specific location or area [Wang et al., 2001b, 2003; Zhang et al., 1999, 2008]. For example, previous studies for Hong Kong [Huang, 2005; Huang et al., 2006] identified the relative contributions of regional transport and local chemical production for those O3episodes in Hong Kong, and showed that, in the lower atmosphere boundary layer, about 70% episode ozone of Hong Kong is contributed by interregional transport from southern China. However, with fast economic development and urbanization, HK and the PRD region become one of the big city clusters controlled by the same air-shed [Chan and Yao, 2008; Tang, 2004]. In recent year, some studies reported that the ozone problem in the PRD exhibits both regional and super-regional properties based on analysis of observation data [Wang et al., 2009; Zheng et al., 2010]. There is an increasing trend of about 0.5 ppbv per year at a background site in Hong Kong during 1994–2007 [Wang et al., 2009], while the ground level ozone concentrations over the PRD region also have increasing trends in the past few years [Zhang et al., 2008]. The Hong Kong government identifies ozone as a regional air pollution problem, and always resorts to regional efforts for its reduction [Environmental Protection Department, 2010]. However, the source contributions at different temporal and spatial scales have not been fully quantified. Therefore, comprehensive ozone source apportionment analysis is needed to obtain a better understanding of ozone formation and variation both spatially and temporally over southern China, which is the base for making effective ozone control strategies and policies.

[5] Recently, two achievements provided necessary conditions to better support the ozone source apportionment studies over the whole PRD region. First, the latest local emission inventory and air quality observational data of the PRD have been made available for air quality modeling simulation and evaluation; in the previous air quality modeling studies, typically only the Hong Kong observational data was available. Due to the lack of observations and local emission inventories, little model evaluation in the PRD was carried out in previous air quality model studies [Cheng et al., 2010; Huang, 2005; Huang et al., 2006; Kwok et al., 2010]. Most of simulations were done without model evaluations for the PRD, and it was hard to further apply the models in the regulatory analysis. Second, we've applied the Comprehensive Air quality Model with extensions (CAMx) with the Ozone Source Apportionment Technology (OSAT) extensions in the PRD region [Yarwood et al., 1996, 1997, 2003]. This model provides the detailed surface ozone source apportionment for multiple targets by source categories and source regions, and even by emission species (NOxand VOC), in one single simulation run for the PRD region, instead of multiple runs using Brute Force method by the chemical-transport models (CTM). This enables us to quantify the contribution of transported ozone and its precursors by different source regions and source categories without prohibitive costs of computing time.

[6] In this paper, we use OSAT in CAMx to study on local, regional, and super-regional source contributions to surface ozone at different time scales (monthly averaged ozone and hourly averaged ozone during episodes) and different spatial scales (ozone at a single site, city-averaged ozone, and regional-averaged ozone) for HK and the PRD. We incorporate the latest PRD emission inventory (2006-based) in our modeling system and validate CAMx modeling results using the recent available observations from PRD monitoring sites network [Guangdong Provincial Environmental Protection Department, 2011]. The rest of this paper is organized as follows. Section 2 describes the data, methodology and model configurations. Section 3presents model evaluation results and assesses the model performance. The local, regional, and super-regional effects on surface ozone being investigated by the OSAT tool are given insection 4. Finally, the implications, findings and recommendations are presented in section 5.

2. Data and Methodology

2.1. Model and Data Description

[7] The detailed description of the model setup and configuration can be found in Li [2011]. Here only a brief summary is presented. In our modeling system, the Fifth-Generation NCAR/Penn State Mesoscale Model Version 3.7 (MM5) [Grell et al., 1994] is adopted as the meteorological model to drive the emission processing model and the chemical transport model. The emission processing model is the Sparse Matrix Operating Kernel Emissions model (SMOKE version 2.4; http://www.smoke-model.org/). We adopt Comprehensive Air quality Model with extensions (CAMx version 5.10; http://www.camx.com/) as the CTM component.

[8] Three nested meshes with grid resolutions of 27 km, 9 km and 3 km are used for the modeling domains (Figure 1), which are denoted as D1, D2, and D3 respectively. The coarsest outer D1 (27 km) includes almost all of China, Japan, Korea as well as southeast Asia, covering major meteorological systems that can affect the PRD and Hong Kong areas, such as Asia monsoon, western Pacific high pressure, typhoon, and others, so that it can adequately provide reasonable background concentrations to the subsequent CTM grids by the model simulation of large scale air pollution transport. The D2 (9 km) covers most parts of South China and the northern part of the South China Sea. The fine domain D3 (3 km) includes most of Guangdong Province, Hong Kong and Macau, which covers the target area for this model study. The initial and boundary conditions for the outmost domain are extracted from GEOS-Chem (http://www-as.harvard.edu/chemistry/trop/geos/) driven CMAQ outputs [Fu et al., 2009].

Figure 1.

Nested domains with grid resolution of 27 km, 9 km and 3 km denoted as D1, D2, and D3 respectively.

[9] Emissions were processed differently for outer (D1 and D2) and fine domains. An Asian emission inventory named INTEX-B 2006, developed byZhang et al.[2009] was used in this study for D1 and D2, supplemented with biogenic emissions from the Model of Emissions of Gases and Aerosols from Nature (MEGAN) [Guenther et al., 2006]. The temporal, spatial and speciation allocations of emissions for outer domains can be found in Kwok et al. [2010] and Zhang et al. [2009]. The emission data for fine domain D3 is based upon 2006-based PRD regional emission inventory [Zheng et al., 2009a, 2009b] and 2006-based Hong Kong emission inventory developed by Hong Kong Environmental Protection Department (HKEPD).Zheng et al. [2009a] estimated that NOx and VOC emissions in the PRD region for the year 2006 are 891.9 kt, and 1180.1 kt, respectively. The spatial distribution of the anthropogenic NOx emission of D3 and D1 are shown in Figure 2 to show the anthropogenic emission pattern within and surrounding the PRD area. As shown in Figure 2b, the PRD region is a relatively isolated but highly polluted region in China. Within the PRD region (Figure 2a), there are several hot spots with high NOx emission, which mainly caused by high density of the traffic and population, and point sources [Zheng et al., 2009a].

Figure 2.

Spatial distribution of NO2 emission over model grids at 14:00 8 November 2006 (a) D3 and (b) D1.

2.2. Model Evaluation

[10] Prior to using the model system to analyze the air quality problems, the model performance evaluation, through the comparison between observations and simulations, is necessary to assess the ability of the model system to reproduce the ambient pollutant concentrations from given emissions and meteorological conditions. In this study, we use both Hong Kong observational data from the HKEPD and the PRD observation from the Pearl River Delta Regional Air Quality Monitoring Network operated since the end of 2005. The Network consists of 16 automatic air quality monitoring stations across the PRD region. Two one-month numerical simulations (July 2006 and November 2006) are performed to evaluate the model performance. The reasons we selected this time period to do the model evaluation are: first, the base year of the new emission inventory used in this study is 2006; second, we have relatively abundant observation for this month in the PRD region; third, there were ozone episodes happened during this time period caused by the two major typical weather patterns [Huang, 2005; Huang et al., 2006].

[11] We selected two typical stations as representatives (Luhu site in Guangzhou and Tung Chung site in Hong Kong), to show the time series comparison between model simulation and observation (see Figure 2a). The surface O3 comparison of model simulation and observation at these two sites is shown in Figure 3. In general, the O3 diurnal variation is well reproduced (high in daytimes and low at night) at both stations. The simulated daily trend of O3 matches well with that of observation for both July and November.

Figure 3.

Ozone time series comparison between observation and simulation of 2006 at (a) Luhu site of Guangzhou in November, (b) Luhu site of Guangzhou in July, (c) TungChung site of Hong Kong in November, and (d) TungChung site of Hong Kong in July.

[12] The averaged statistical metrics for hourly observation and model simulation of O3, and NO2 pollutants at twenty available monitoring stations in the PRD and Hong Kong are shown in Table 1. On average, NO2 is moderately underestimated, while ozone is overestimated in certain sites. The Normalized Mean Bias (NMB) for NO2 ranges from −0.08 to −0.26; while as the NMB for O3 ranges from +0.08 and +0.72. The relative large NMB and Root Mean Square Error (RMSE) for O3 in July may be associated with the over estimation of O3during the very low ozone concentration of non-episodic days. For other metrics, the RMSE ranges from 11.53 to 20.44; the Index of Agreement (IOA) ranges from 0.56 to 0.76; the Correlation Coefficient (R2) ranges from 0.46 to 0.74 respectively. The overall model performance of NO2 and O3 are comparable to other model simulations in this region [Fung et al., 2005; Huang et al., 2005, 2006; Li, 2011; Kwok et al., 2010; Lo et al., 2006a, 2006b; Yim et al., 2010; Wang et al., 2010]. The agreement between the observation and simulation shows that the model has a reasonable skill to reproduce the temporal and spatial patterns of O3, as well as the reliability of the latest PRD local emission inventory integrated in the modeling system.

Table 1. Averaged Model Performance of NO2 and O3 in July and November 2006a
MonthPollutantNMBRMSE (ppb)IOAR2
  • a

    NMB, RMSE, IOA and R2 are the short name for Normalized Mean Bias, Root Mean Square Error, Index of Agreement, and correlation coefficient respectively.

Jul. 2006NO2−0.0815.210.600.46
O30.7220.440.760.74
Nov. 2006NO2−0.2614.530.560.52
O30.0811.550.750.72

2.3. Ozone Source Apportionment Technology

2.3.1. OSAT

[13] Ozone is not a directly emitted pollutant, but a secondary pollutant from the chemical reaction of its precursors (mainly NOx and VOC). The precursors may, or may not, form ozone immediately after the release, depending on their life time, the relative abundance of NOx and VOC, solar radiation, etc. Hence, the location of ozone formation may not be the same as the location where the precursors are emitted. The amount of ozone accumulated in a particular region is a mixture of ozone formed at different locations and time by NOx and VOC from different source region/category under either NOx-limited or VOC-limited conditions. An ozone source apportionment method has the capability to separate out how much ozone at a specific receptor is due to each source region, as well as each source category. It can also estimate that how much of ozone formation is limited by VOC and how much is limited by NOx. Basically, the OSAT is a mass balance analysis technique for helping identification of source contributions to elevated surface ozone. The OSAT is capable of estimating the contributions of multiple source areas, source categories, and pollutant types to ozone formation in a single model run and identifying which pollutants, source categories and source regions should be controlled to attain ozone control objectives.

2.3.2. Source Groupings and Receptors Setting of OSAT

[14] OSAT in CAMx 5.10 attributes the ozone and its precursors based on source grouping, which is defined in terms of geographic areas (source regions) and/or source categories. In this study, we divide the whole D3 domain into 12 source areas, and divide the emission inventory into 7 source categories. The 12 source areas mainly classified by the administrative boundary of cities, which include Foshan (FS), Jiangmen (JM), Zhongshan (ZS), Dongguan (DG), Huizhou (HZ), Shenzhen (SZ), Guangzhou (GZ), Macau and Zhuhai (M&ZH), PRD water area (PRDw), Hong Kong (HK), and Zhaoqing (ZQ); the remaining areas are called the non-PRD area (NPRD) (Figure 4 and Table 2). The 7 source categories include anthropogenic source outside the PRD economic zone (INTEXB), biogenic source (biogenic), Hong Kong mobile sources (hkmv), PRD mobile sources (prdmv), marine shipping sources (marine), point sources such as power plants and large industrial point sources (point), and area sources such as small industrial sources, domestic and commercial fuel combustion (area). In addition, the model boundaries (East (ESTBC), West (WSTBC), South (STHBC), North (NTHBC), and Top (TOPBC)) and the initial concentration are also tracked as separate source groupings. Thus there are a total of 90 source groupings in the model runs. We analyze the ozone source apportionment with respect to the above source regions and source categories for both city or region averaged ozone and ozone for certain locations. In this study, for each source receptor, we define the source contribution from its city as a local (city) effect, the other city contributions within the PRD region as the (PRD) regional effect, while the rest are considered to be the super-regional effect (NTHBC + STHBC + ESTBC + WSTBC + TOPBC + NonPRD, which contains all the boundary inputs plus the non-PRD area within the D3 domain).

Figure 4.

Sub-division of D3 into 12 source regions for geographic source apportionment. The white color region represents the non-PRD (NPRD) source region. The numbers 1–4 label the locations of additional small target receptors for OSAT analysis: 1, GZurb (GZ_urban); 2, JMurb (JM_urban); 3, HKurb (HK_urban); and 4, HKTC (HK_TC).

Table 2. Source Grouping Definition by Source Regions, Source Categories, and Boundary Conditions
Source Region (SR)SR Short Name
GuangzhouGZ
ShenzhenSZ
HuizhouHZ
DongguanDG
ZhongshanZS
JiangmenJM
FoshanFS
Hong KongHK
PRD waterPRDw
Macau and ZhuhaiM&ZH
ZhaoqingZQ
Non-PRDNPRD
Source Category (SC)SC Short Name
PointPoint
AreaArea
MarineMarine
BiogenicBiogenic
PRD mobilePrdmv
Hong Kong mobileHkmv
Anthropogenic emis. outside the PRDINTEXB
Other Source GroupingsShort Names
North BoundaryNTHBC
South BoundarySTHBC
West BoundaryWTHBC
East BoundaryETHBC
Top BoundaryTOPBC
Initial ConditionIC

3. Results and Discussion

[15] In this study, the OSAT analysis is performed both for episodic events associated with the two major weather pattern regimes introduced in section 1, and the whole month simulation both for: November of 2006 (which is a representative of the autumn season), and July of 2006 (which is a representative of the summer season). Ozone episodes that occurred from 7 to 10 November 2006 and from 21 to 24 July 2006 were identified as associated with the above two typical regimes respectively (here we set 80 ppbv as a threshold for an ozone episode which can be referred to Grade I level of Chinese National Ambient Air Quality Standards).

3.1. Ozone Source Apportionment in the Fall Season (November 2006)

[16] The first six days of November 2006 are skipped as the spinning-up period for the CTM modeling, then we select the period from 7 to 30 November 2006 for analyzing ozone source apportionment results. As the prevailing wind is northeasterly during the study period (Figure 5a), Guangzhou (GZ) is selected as an upwind area, Jiangmen (JM) is selected as a downwind area, and Hong Kong (HK) is chosen as a representative of international port city, which is very close to the east boundary. We also select some specific grids (center area with high population density) in the above cities as receptors to represent the urban or suburban situation (Figure 4). These receptors are denoted as GZ_urban, JM_urban, HK_urban, and HK_TC (Tung Chung is a suburban area of HK; Higher ozone levels are frequently observed in the TC area, compared to other areas in HK) respectively. In addition, the ozone source apportionment results for locations in each city with maximum daytime averaged ozone concentration (GZ_maxp, JM_maxp, and HK_maxp, see Figure 5b) during episode days (9–10 November 2006) are also examined and summarized in Table 3. In the following sections, the ozone source apportionment result at these receptors will be discussed.

Figure 5.

(a) Hourly averaged wind pattern during 7–10 November 2006. (b) Locations with maximum ozone concentration in each city with maximum daytime averaged ozone concentration during episode days (11:00 to 16:00 from 9 to 10 November 2006) overlaying on the averaged ozone distribution for the same time period. The numbers 1–11 indicate the locations: (1) HK_maxp, (2) PRDw_maxp, (3) ZH_maxp, (4) GZ_maxp, (5) SZ_maxp, (6) HZ_maxp, (7) DG_maxp, (8) ZS_maxp, (9) JM_maxp, (10) FS_maxp, (11) ZQ_maxp.

Table 3. Local, Regional, and Super-Regional Source Contributions for Daytime Averaged Ozone From 11:00 to 16:00 During 9–10 November 2006 at Different Receptors
ReceptorsaLocal (%)Regional (%)Super-Regional (%)Averaged Ozone (ppbv)
  • a

    Receptors: GZ, Guangzhou; JM, Jiangmen; HK, Hong Kong (Figure 4); GZ_urban, JM_urban, and HK_urban are small urban area in GZ, JM, and HK respectively; HK_TC is a suburban area in HK (Figure 4); GZ_maxp, JM_maxp and HK_maxp are the maximum ozone location in GZ, JM and HK respectively during the selected period (Figure 5b).

GZ29215083.9
JM16374788.1
HK5276877.4
GZ_urban34283887.7
JM_urban46234112.9
HK_urban8296354.0
HK_TC10276397.1
GZ_maxp27423199.6
JM_maxp16930116.9
HK_maxp17524102.5

3.1.1. Ozone Source Apportionment by Regions in the Fall Season (November 2006)

[17] The source apportionment by different regions during the episode days from 7 to 10 November 2006 is shown in Figure 6. The upwind city GZ is mainly affected by super-regional sources, local sources and its upwind city HZ (Figure 6a), while the downwind city JM has higher ozone concentration and is affected by more complex source regions in the PRD (namely HK, GZ, SZ, HZ, DG, ZS, JM, FS, ZQ, PRDw) besides the super-regional and local contributions (Figure 6b). The HK is largely affected by the super-regional sources, because its upwind regions are mostly outside the PRD under the northeast wind condition. Nevertheless HK is still notably affected by HZ, SZ, PRDw, DG, and its local sources on the last day of the episode (Figure 6c). These results indicate that the high ozone episode in the PRD is attributed by different source regions, especially for the downwind area where the ozone concentration is higher. Our results show that the ozone control within polluted regions affected by the same air-shed is of limited effectiveness when conducted by single city emission reduction. In contrast, it is a strong indication for implementation of the regional joint prevention and control mechanisms to strengthen the coordinated control of ozone pollution.

Figure 6.

Ozone source apportionments by different source regions during 7–10 November 2006 in (a) Guangzhou, (b) Jiangmen, and (c) Hong Kong.

[18] The quantified impacts of the local, regional and super-regional sources in the episode days of all receptors with different scales are shown inTable 3. It can be seen that GZ has a relatively larger local contribution (29%) since GZ is a major source region; JM has a relatively larger PRD regional contribution (37%) because it is located at the downwind area under the northeast prevailing wind pattern. Compared to the GZ and JM regions, the ozone of the HK area is contributed more by the super-regional sources (68%). This may be caused by two phenomena. (i) As shown inFigure 5a, the whole PRD region is under the control of northeast prevailing wind, while the sea surface is affected by the northeasterly prevailing wind along the seashore area. HK is an upwind area under this wind pattern and is also close to the east boundary of the D3 domain. Therefore, the ozone in the Hong Kong area is affected more by the transportation from the boundary which is classified as the super-regional effect (68%). (ii) Because Hong Kong has a relatively small geographic size compared to GZ and JM receptors, the HK emission will contribute to its downwind area under the northeast prevailing wind. For example, the local contribution is about 5% for the averaged ozone of HK area, while the value is doubled for the ozone of HK_TC area, which is a western downwind area under the northeasterly prevailing wind pattern in HK. In consequence, we can observe a notable HK contribution to the ozone of the JM area inFigure 6b. In summary, during the episode days, the major ozone source regions for different receptor cities largely depends on geographic location of the receptor which is associated with the emission spatial distribution, and affected by different meteorological pattern.

[19] The above results are for city-scale receptors, which cover the entire city including both urban and rural regions of that city and represent spatial average results. If we examine the selected urban areas, the local and PRD regional contributions increase to 62%, 66%, and 37% for GZ_urban, JM_urban, and HK_urban compared to 50%, 53% and 32% for the city averaged results respectively. The local and PRD regional sources contribute even more in the locations with maximum ozone concentration, 69%, 70%, and 76% for GZ_maxp, JM_maxp and HK_maxp respectively. The local and PRD regional contributions dominate the ozone formation for the maximum ozone locations, even in HK, where the super-regional contribution is dominant for city averaged ozone. These results show the important roles of the emission sources within the PRD region to the formation of high ozone concentration in the PRD during the episode days.

[20] The source apportionments for GZ_maxp, JM_maxp and HK_maxp also reveal that the upwind cities with high emission sources can make a significant contribution to cause the maximum ozone at downwind areas. In upwind city GZ with high emissions, the local city emission contributes 27% to the ozone concentrations at GZ_maxp. However, for downwind city JM, the regional contribution to JM_maxp is 69% and the local city contribution is only about 1%. Hence, reduction of ozone precursors in upwind high emission areas is very important for ozone reduction in the downwind areas.

[21] Figure 7shows the ozone source contributions from different source regions for GZ, JM and HK for the whole month of November 2006. On average, the super-regional ozone transported into the PRD areas is about 40 ppbv. The hourly maximum contribution from super-regional transport can be up to 60 ppbv in the GZ area, and can even be up to 80 ppbv in the HK area. The transported ozone from super-regional sources is the dominant factor for the measured ozone in non-episode days. The ozone problem of this city cluster is not only a regional problem, but also significantly affected by super-regional sources. Comparing episode days (7–10 November 2006) and non-episode days (the remaining days in November), we can see that the super-regional contribution does not increase but it actually slightly decreases during the episode days. In contrast, the local and PRD regional contributions have significant variations between episode days and non-episode days and these sources can be the dominant contributions during the episode days. Therefore, the super-regional contribution can be roughly taken as a background source to the ozone concentration and the increased local and regional contribution (emission sources within the PRD region) is the major cause of the formation of high ozone episode within the PRD region.

Figure 7.

Ozone source apportionments by different source regions during November 2006 in (a) Guangzhou, (b) Jiangmen, and (c) Hong Kong.

[22] Table 4 shows the quantitative daytime average apportionment results by regions for GZ, JM and HK respectively. Compared with Table 3, we can further quantify the important roles of the emission sources within the PRD region in the formation of high ozone episodes. The local and regional contributions to the ozone concentration increase to 50%, 53%, and 32% for GZ, JM, and HK during the episode days (Table 3) compared with only 27%, 37% and 19% for the monthly averaged ozone (Table 4).

Table 4. Local, Regional, and Super-Regional Source Contribution to City Averaged O3a
 Local (%)Regional (%)Local+Regional (%)Super-Regional (%)Averaged O3 (ppbv)
  • a

    Average of the daytime hours (11:00–16:00) during 7–30 November 2006.

GZ1215277338.7
JM829376339.8
HK217198137.2
Ave.820287138.6

3.1.2. Ozone Source Apportionment by Categories in the Fall Season (November 2006)

[23] In addition to source regions contribution to ozone levels, the source apportionment with respect to source categories is also important, since in practice we usually apply control measures based on source categories or sectors. In this paper, we divide the emission inventory into 7 categories as shown in Table 2. The source apportionment results with respect to these categories are shown in Figure 8 and Table 5for the three representative cities GZ, JM and HK during the episode from 7 to 10 November 2006. When considering source categories, we only focus on the sources within the PRD region in this study. Therefore, the details of super-regional contributions will not be discussed here.

Figure 8.

Ozone source apportionments by different source categories during 7–10 November 2006 in (a) Guangzhou, (b) Jiangmen, and (c) Hong Kong. In the colorbar, the label hkmv is short for Hong Kong mobile source, while prdmv is for PRD mobile source.

Table 5. Ozone Source Contribution in Percentage (%) by Source Categories for Daytime Averaged Ozone From 11:00 to 16:00 During 7–10 November 2006 at Different Receptorsa
Source CategoryGZGZ_ urbanGZ_ maxpJMJM_ urbanJM_ maxpHKHK_ urbanHK_ maxpAverage
  • a

    Receptors: GZ, Guangzhou; JM, Jiangmen; HK, Hong Kong (Figure 4); GZ_urban, JM_urban, and HK_urban are small urban area in GZ, JM, and HK respectively; HK_TC is a suburban area in HK (Figure 4); GZ_maxp, JM_maxp and HK_maxp are the maximum ozone location in GZ, JM and HK respectively during the selected period (Figure 5b).

Mobile29352725323811153628
Area91416172315542514
Point8767621175
Marine<1<1<12436412
Biogenic4684382375

[24] For the sources inside the PRD, the mobile source is the dominant category to form ozone for all the three cities (Figure 8). We notice that even in HK, the PRD mobile source (prdmv) can also be the dominant source contribution instead of HK local mobile source (hkmv). From Figure 6c, we can conclude that HZ and SZ are the first and second PRD source category contributing to HK ozone in terms of source regions. In the upwind city GZ and the downwind city JM, the second and third important source categories are the area source and the point source. In HK, an international shipping port, the second and third important sources are the area source and the marine source. The notable marine source contribution is associated with the international shipping business in the Pearl River Estuary region. It was reported that Hong Kong and Shenzhen's port together handled 11.7% of the world's container throughput [Civic Exchange, 2007].

[25] The quantitative attribution results in Table 5 for the daytime averaged ozone over the episode days also reveal that the mobile source is the largest contributing source in the formation of ozone, for all the three studied cities, also for the urban areas and the locations with maximum ozone concentration. For downwind city JM, the second largest contributing source is the area source for different receptors. For upwind city GZ with high local emissions, the area source and the point source have approximately the same contribution for city averaged ozone concentration; but for selected urban locations or locations with maximum ozone concentration, the area source can contribute much more than the point source and can be the second largest contributing source category. HK as a major port city in the world, both the area source and the marine source have similar contributions for averaged ozone of HK, but the area source can contribute much more than the marine source for HK_maxp. Based on these results, control of mobile source is the maximum effective control option. In addition, the area source, point source, and marine source also accounts for different source contribution which is dependent on location.

3.2. Ozone Source Apportionment During Episode Days in the Summer Season (July 2006)

[26] To further understand the ozone source apportionment of the PRD and Hong Kong under different meteorological conditions, another high ozone case, which occurred in the summer season from 21 to 24 July 2006, is selected to analyze the ozone source apportionment. Figure 9 shows the ozone source contributions from different source regions for GZ, JM and HK respectively for the summer case. Comparing with the autumn case in Figure 6, the super-regional contribution is less important, especially in HK, which can be associated with the different meteorological conditions in the summer season. The wind pattern is considerably different from that of the autumn case (seeFigure 5a): in the autumn case, most of the PRD region is influenced by a prevailing northeasterly wind and the water surface of the southern PRD is influenced by an easterly wind; while in the summer case, most of the PRD landmass is controlled by a northwesterly wind, and the wind speed is relatively low, and the water surface of the southern PRD is controlled by a westerly wind (see Figure 10a). Therefore, in the autumn case, HK is an upwind area; but in the summer case, HK becomes a downwind area with a relatively higher averaged ozone concentration on 24 July 2006 (Figure 9). The wind effect can be demonstrated by the ozone plume direction in Figure 10b. In the following paragraphs, we will discuss the ozone source apportionment results in the summer case compared to the autumn case both for episode days and non-episode days.

Figure 9.

Ozone source apportionments by different source regions from 21 to 24 July 2006 in (a) Guangzhou, (b) Jiangmen, and (c) Hong Kong.

Figure 10.

(a) Hourly averaged wind pattern on 24 July 2006 of D3. (b) The ozone distribution pattern at 15:00 on 24 July 2006 of D3 (unit: ppb).

[27] Table 6shows the quantitative source apportionment for the daytime averaged ozone during 24–25 July 2006. We can observe that higher local city contribution and lower super-regional contribution to the formation of ozone is found as compared with the autumn case. For this summer episode, the super-regional contribution is even less, especially for HK. The super-regional contribution is only about 27% and 29% in HK and in HK_urban; while in the autumn case, the super-regional contribution is up to 68% and 63% in HK and in HK_urban. In contrast, the ozone episode is predominantly affected by the local and PRD regional emission sources in the summer case (∼70%). From the results of the locations with maximum ozone concentration in each city, we can see that the integrated contribution of PRD regional sources is the major cause for the maximum ozone within the PRD region. Therefore, both in the summer case and in the autumn case, the local and PRD regional emissions are the major sources to form the high ozone episode within the PRD region.

Table 6. Local, Regional, and Super-Regional Source Contributions for Daytime Averaged Ozone From 11:00 to 16:00 During 24–25 July 2006 at Different Receptors
ReceptorsaLocal (%)Regional (%)Super-Regional (%)Averaged Ozone (ppbv)
  • a

    Receptors: GZ, Guangzhou; JM, Jiangmen; HK, Hong Kong (Figure 4); GZ_urban, JM_urban, and HK_urban are small urban area in GZ, JM, and HK respectively; HK_TC is a suburban area in HK (Figure 4); GZ_maxp, JM_maxp and HK_maxp are the maximum ozone location in GZ, JM and HK respectively during the selected period (Figure 5b).

GZ35402590.3
JM40223860.0
HK8652795.1
GZ_urban354916103.1
JM_urban30452582.7
HK_urban10612951.9
HK_TC4613594.5
GZ_maxp254729156.1
JM_maxp47224113.6
HK_maxp16534105.8

[28] The source apportionment results for the summer case with respect to the source categories (see Table 2) are shown in Figure 11 and Table 7 for the three representative cities GZ, JM and HK. The important roles of the mobile emission sources in the formation of high ozone episode can be identified. On average, the contribution of the mobile source accounts for 30% ozone formation, while the area, point, marine, and biogenic source account for 12%, 16%, 3%, and 8% respectively. In addition, the point source shows a larger partition in summer case (16%) than that in autumn case (5%). For the anthropogenic source within the PRD region, our results show that controlling the mobile source is still the most effective control option for all these receptors.

Figure 11.

Ozone source apportionments by different source categories from 21 to 24 July 2006 in (a) Guangzhou, (b) Jiangmen, and (c) Hong Kong. In the colorbar, the label hkmv is short for Hong Kong mobile source, while prdmv for PRD mobile source.

Table 7. Ozone Source Contribution in Percentage (%) by Source Categories for Daytime Averaged Ozone From 11:00 to 16:00 During 24–25 July 2006 at Different Receptorsa
Source CategoryGZGZ_urbanGZ_maxpJMJM_urbanJM_maxpHKHK_urbanHK_maxpAverage
  • a

    Receptors: GZ, Guangzhou; JM, Jiangmen; HK, Hong Kong (Figure 4); GZ_urban, JM_urban, and HK_urban are small urban area in GZ, JM, and HK respectively; HK_TC is a suburban area in HK (Figure 4); GZ_maxp, JM_maxp and HK_maxp are the maximum ozone location in GZ, JM and HK respectively during the selected period (Figure 5b).

Mobile33434018283224243030
Area7812871820201012
Point8101813241213122716
Marine1<12<1<1442<13
Biogenic7686313131548

3.3. Ozone Source Apportionment for the Whole PRD

[29] The results of source apportionment by considering the whole PRD as a receptor for both the episode days and month average are shown in Table 8. The super-regional contribution in the summer case (28∼29 ppbv) is generally lower than that in the autumn case (41∼42 ppbv). This is due to the northerly prevailing wind with relatively higher wind speed in the autumn, which may cause more ozone transportation from outside the PRD region in the autumn case, while ozone of non-episode days in summer is more influenced by the oceanic fresh air mass brought by summer monsoon. Comparing to global scale background ozone studies [McDonald-Buller et al., 2011; Nagashima et al., 2010; Wang et al., 2011], the super-regional contribution is the combination of (a) the ozone contribution from anthropogenic sources within China/East Asia but outside the PRD, (b) the ozone contribution from non-China/non-East Asia anthropogenic sources, and (c) the global natural ozone background. In the U.S., the concept of “policy relevant background (PRB)” is used in ozone policy discussion [McDonald-Buller et al., 2011]. The PRB ozone is defined by the U.S. Environmental Protection Agency (EPA) as the ozone levels that would occur in the U.S. in the absence of anthropogenic emissions in continental North America [McDonald-Buller et al., 2011]. The PRB of China/East Asia is hence equivalent to (b) and (c) in the above discussion. Hence, our super-regional contribution is larger than the PRB concept in the China/East Asia perspective. The super-regional contributions to the whole PRD (in ppbv) in the summer and in the autumn are very similar to the results of Total Background Ozone (TBO) in South China (SC) region as reported byWang et al. [2011] from the global model study. Wang et al. [2011]showed that TBO of South China is about 35 ppbv in November and about 25 ppbv in July. Our super-regional contribution to the PRD ozone is about 41∼42 ppbv in November and 28∼29 ppbv in July. If we ignore the various model uncertainties (which will be further discussed in thesection 3.4), the difference between the TBO and super-regional ozone could be explained by (a) the biogenic source contribution (b) contribution from emission sources outside PRD but within China. The biogenic sources contribution of the PRD estimated fromTable 3 and Tables 57is about 5 ppbv in November and 7 ppbv in July. Regardless of the difference that South China is larger than the PRD, we can roughly estimate that the ozone contribution to the PRD from anthropogenic sources within China but outside the PRD region is about 11∼12 ppbv in November and 10∼11 ppbv in July. Of course, this estimation includes various model uncertainties, such as, possible overestimation of super-regional contribution in the regional model study; and discrepancies between the regional 3D CTM and global 3D CTM, such as emission inventory difference. In percentage, the super-regional contribution (55%) in the summer is generally much lower than that (70%) in the autumn. Similar results were also reported byWang et al. [2011]for TBO contributes to South China. In summary, the super-regional contribution from this study is consistent with seasonal variation pattern as well as magnitude of TBO in global scale study. We also like to reiterate that during episodic conditions, local and regional PRD sources are the main cause of ozone episodes in the PRD region.

Table 8. Summary of Contributions to the PRD Averaged Ozone From Emissions Within the PRD and Outside the PRD During Daytime (11:00–16:00)
 Local+Regional (%)Super-Regional (%)Super-Regional (ppbv)
Autumn month mean297141.24
Summer month mean455528.31
Autumn episode505042.37
Summer episode703029.51

3.4. Uncertainties

[30] The ozone source apportionment results are limited by several uncertainties that are common to most air quality modeling studies. First, the OSAT approach is based on the CAMx model and driven by MM5 model, which have numerical and parameterization errors in their model components as well as uncertainties in its emissions inventory. Nevertheless, the agreement shown in section 2.2 between the observed and simulated ozone concentrations suggests that our modeling system does have reasonable skill in ozone simulation. Second, an overestimation of the wind speed [Kwok et al., 2010; Lo et al., 2007] in urban areas may lead to an underestimation of local source impacts and hence an overestimation of upwind contributions. Since ozone episode days in HK/PRD are often associated with accumulation under weak wind situation [Huang et al., 2006], the upstream BC contribution may have been overestimated also during episodic events. Third, because of the limitation of the OSAT tagging capabilities, there may also be an overestimation of BC contribution when ozone formed from precursors in the HK/PRD region flow out of and then re-circulate back into domain D3.

4. Conclusion and Discussion

[31] By looking into the ozone source apportionment results at different spatial scales (PRD regional averages, city averages and selected ozone maxima in different cities) and at different time periods (over a month or during ozone episodes in different seasons), we have developed a detailed understanding of the “by region and by category” precursor source contribution of surface ozone in the HK/PRD areas.

[32] Temporally, both the summer case and autumn case show that the source partitions from local, regional, and super-regional contributions can be very different between high ozone episode days and the monthly average result. On average, the super-regional ozone transported into PRD receptors is about 41 ppbv (70%) in the autumn season, and is about 28 ppbv (55%) in the summer season. By comparing the ozone source apportionment results of episode days with the results of the non-episode days, we show that the super-regional contribution amount (in ppbv) does not increase, and even slightly decrease during episode days; while the local and regional contributions become larger in the episode days both in terms of total amount of concentration and the percentage to the averaged ozone. In autumn, the local and regional contributes about 30% of ozone during the non-episode days, but their contribution can go up to about 50% during ozone episode days in the autumn (Table 8). Moreover, in summer, the contribution from local and regional PRD sources can contribute even up to about 70% during episode days (Table 8). This difference implies that the increased local and regional contribution (emission sources within the PRD region) is the major cause of the high ozone episodes of the PRD region. The ozone from super-region apportioned a fundamental contribution, which is one of the dominant factors for the ozone in non-episode days. In general, the super-regional sources can be treated as a background level, which are dominant contributions for the long-term averaged ozone levels. The substantial increased in local and PRD regional contributions during episodic situations show the importance of local and regional source control of ozone precursors inside the PRD.

[33] The results indicate that there is lots of potential for local governments to reduce the high ozone episode days in the PRD by controlling the emission sources within the PRD region. In particular, ozone is a pollutant with well-known acute but less chronic health impact [Tang, 2004], so if the control measures can be enforced to make sure that the peak ozone values stay below acceptable levels, a significant health gain for the population can be expected. Looking at the current national Grade II standard of 100 ppbv hourly ozone concentration, for example, we note that during ozone episodes in the autumn or summer, local and PRD sources contribute a significant amount, at least 50 or 75 ppbv (50% or 75%) for the whole city averaged results in GZ, and the amount will even be higher for in urban areas (GZ_urban). Control measures applied in the PRD region can be very effective to reduce this portion of ozone concentration, and can make it less likely for the total ozone concentrations to go above the health based national air quality standards. We also note that, for some other pollutants with chronic health impact the control potential of super-regional source contribution to the averaged level of the pollutant may become significant (i.e., particulate matter).

[34] Spatially, the results at locations of maximum ozone concentration in downwind city show that except for GZ which has a relatively higher local city contribution (27% for the autumn case and 30% for the summer case), this local city contribution is very small in JM_maxp (1%) and HK_maxp (1%). The very small local city contribution at JM_maxp and HK_maxp indicate that the maximum ozone concentration in downwind city is generally caused by the well-mixed polluted plume composited by air masses transported from multiple upwind source regions. The PRD regional source accounts for 69% and 75% in JM_maxp and HK_maxp respectively in the autumn case, and accounts for 72% and 65% in the summer case. These results show that the integrated contribution of PRD regional sources is the causative factor for the maximum ozone within the PRD region.

[35] The above understandings can help us to develop a better ozone control strategies. Our results indicate that the ozone control within a polluted region influenced by the same air-shed is of limited effectiveness if control measures are only applied in single city. In contrast, the control will be very effective by regional joint control of ozone precursors, since the high ozone episode in the PRD and Hong Kong region is contributed to by multiple source regions.

[36] The ozone source apportionment results demonstrate that the mobile source accounts for 28% of ozone formation, while the area, point, marine, and biogenic source account for 14%, 5%, 2%, and 5% respectively in the autumn case. However, the mobile source accounts for 30% ozone formation, while the area, point, marine, and biogenic source account for 12%, 16%, 3%, and 8% respectively in the summer case. Therefore, when we consider applying the ozone control strategy to the anthropogenic source within the PRD region, the mobile source is the source category to be controlled, which can have maximum ozone reduction effectiveness both in the autumn case and in the summer case.

[37] The difference between the source apportionment partition of the autumn case and of summer case is mainly caused by variations in meteorological conditions. The local city contribution accounts for higher partition in the summer case than in the autumn case. The relative higher local city contribution and lower super-regional contribution in summer case is possibly associated with the weaker wind speed in the summer case. Despite the quantitative difference, the major conclusion of the autumn case is also similar to that in the summer case. We found substantial seasonal variations in the way ozone precursors from neighboring areas affect ozone levels in any particular city, suggesting that regional collaborations are important for developing effective long-term strategies to reduce ozone over the PRD region.

[38] As July and November are generally the time periods with the highest ozone pollution levels under different typical meteorological patterns, our two cases studies are representative of ozone seasons in the PRD region. The results are limited to uncertainties from meteorological model simulations and emission inventory. For example, the possible over-estimation of surface wind speed may lead to over-estimation of super-regional contribution. The uncertainties in the emission inventory may lead to uncertainty in the quantified source contribution from different source regions. But these uncertainties will not change our major conclusion, the local and PRD regional sources are the major source contributions to the formation of ozone during the high ozone episode days.

Acknowledgments

[39] We are grateful to the Associate Editor and three anonymous reviewers for their constructive comments. We appreciate the assistance of Hong Kong Environmental Protection Department (HKEPD), which provided the air quality monitoring data, the assistance of the Hong Kong Observatory (HKO), which provided the meteorological data. This work was supported by NSFC/RGC grant N_HKUST631/05, NSFC-FD grant U1033001, the Fu Tak Iam Foundation (FTIFL08/09.EG01) and the Fok Ying Tung Graduate School (NRC06/07.SC01). The content of this paper does not necessarily reflect the views and policies of the Government of the Hong Kong Special Administrative Region (HKSAR), nor does mention of trade names or commercial products constitute an endorsement or recommendation of their use.

Ancillary