TRACE: Tennessee Research and Creative TRACE: Tennessee Research and Creative Exchange Exchange Air quality and emissions in the Yangtze River Delta, China Air quality and emissions in the Yangtze River Delta, China

. Regional trans-boundary air pollution has become an important issue in the ﬁeld of air pollution modeling. This paper presents the results of the implementation of the MM5-CMAQ modeling system in the Yangtze River Delta (YRD) for the months of January and July of 2004. The meteorological parameters are obtained by using the MM5 model. A new regional emission inventory with spatial and temporal allocations based on local statistical data has been developed to provide input emissions data to the MM5-CMAQ modeling system. The pollutant concentrations obtained from the MM5-CMAQ modeling system have been compared with observational data from the national air pollution monitoring network. It is found that air quality in winter in the YRD is generally worse than in summer, due mainly to unfavorable meteorological


Introduction
The Yangtze River Delta (YRD), characterized by high population density and well-developed industry, is one of the largest economic regions in China. With rapid economic development in recent years and high energy consumption, air pollutant emissions are increasing steadily and the regional environment is deteriorating. Regional visibility is decreasing, ozone concentrations are increasing, and the ecological environment and health of the people are suffering. Many observational studies in the past (Chameides et al., 1999;Xu et al., 1999;Luo et al., 2000;Cheung and Wang, 2001;Wang et al., 2001a) showed that high ozone concentrations were beginning to appear in Eastern China. A project sponsored by China's National Natural Science Foundation, "The atmospheric physical and chemical process and its influence on ecosystem in the Yangtze River Delta", has conducted field surveys on air quality in the YRD, which indicate that the ozone concentration can be high in this region; the occurrence frequency of ozone concentrations higher than 0.16 mg m −3 reached 20% at some sites .
Published by Copernicus Publications on behalf of the European Geosciences Union.
Fine particles and ozone are considered to be the most serious air pollutants of concern in China and the United States today, as well as in most metropolitan areas around the world (Streets et al., 2007). High ozone concentration at ground level increases the frequency of urban photochemical smog, accelerates the aging process of materials, threatens people's health, and causes serious damage to the ecological environment (Wang et al., 2001b). Thus, high ozone concentration has been a great concern of environmental scientists in China (Xu et al., 1999;Zhu and Xu, 1994). Fine particles not only cause worsening of regional visibility, but also seriously affect people's health. Fine particles and ozone are both formed through a serious of complex chemical reactions among primary pollutants in the air. The lifetime of ozone formed through VOC and NO x reactions is usually several days, which allows further meteorological processes to complicate the ozone formation and destruction processes (Finlayson-Pitts and Pitts, 1993). In addition, NO x , VOC, SO 2 , and NH 3 can generate particulate matter such as (NH 4 ) 2 SO 4 and NH 4 NO 3 through chemical reactions. Therefore, it is difficult to assess their environmental impacts solely by experimental studies and field observations (Tesche, 1983).
To assess air quality in the YRD, the US EPA's "One Atmosphere" model, the Community Multi-scale Air Quality Modeling System (CMAQ) (Dennis et al., 1996;Byun et al., 1998;Byun and Ching, 1999), is used to simulate the chemical transport of air pollutants, driven by a new emission inventory. CMAQ is a chemical dispersion model representative of the third generation of air quality models and includes ozone, aerosols, acid deposition, visibility, and air toxics chemistry. Under a US EPA initiative, CMAQ has already been successfully transferred to the Asian situation and used to assess air quality in Beijing Streets et al., 2007;Fu et al., 2008), the YRD , the Pearl River Delta (PRD) (Wang et al., 2010), and across East Asia (Zhang et al., 2004(Zhang et al., , 2005Fu et al., 2008;Lin et al., 2009). In this paper, the fifth-generation NCAR/Penn State Mesoscale Model (MM5) (Chen and Dudhia, 2001;Chen et al., 2005) is used to produce the necessary meteorological fields in 4-D data assimilation mode; CMAQ version4.4 is used for the numerical simulation of air pollutant transport and transformation in the YRD.

Input data
The methodology used in this paper is to simulate the atmospheric processes over the YRD domain with MM5-CMAQ and a new emission inventory, in order to obtain a dataset with air pollutant concentrations in time and space for the simulation periods, which can then be compared with monitored concentrations. The driving meteorological inputs for CMAQ are provided by MM5, and the meteorology-chemistry interface processor (MCIP) is used to transfer MM5 output into gridded meteorological field data as the input to CMAQ. The inputs for MM5 are NCEP FNL (Final) Operational Global Analysis data, which are available on 1.0 × 1.0 degree grids continuously for every six hours (http://dss.ucar.edu/datasets/ds083.2/). The Carbon Bond-IV chemical mechanism (CB-IV) is used in the CMAQ model, which consists of 36 chemical species, 93 chemical reactions, and 11 photochemical reactions (Lamb, 1982). Geographic Information System (GIS) technology is applied in gridding the YRD regional emission inventory to the model domain.

Model domain and simulation episodes
The CMAQ model domain is based on a Lambert Conformal map projection, using a one-way nested mode with grid resolutions of 36 km (covering the whole of China), 12 km (covering Eastern China), and 4 km (covering the YRD region), as shown in Fig. 1. The MM5 domain is larger than the CMAQ domain, with three grids more than the CMAQ domain on each boundary. The mother domains for both MM5 and CMAQ are centered at (110 • E, 34 • N). The YRD domain for CMAQ has 118 × 136 horizontal grid cells and includes 16 major cities, which are also shown in Fig. 1. Each "city" in the domain is a large administrative area that includes smaller cities, towns, and villages. Thus, the 16 cities together cover the majority of the YRD area. The pollution episodes chosen are 1-31 January 2004, and 1-31 July 2004, which represent winter and summer seasons, respectively. The initial conditions of CMAQ for each seasonal run are prepared by running the model five days ahead of each start date with clean initial conditions (IC). Sensitivity tests show that the influence of IC generally dissipates after about three days. The boundary condition (BC) used for the largest domain of CMAQ is clean air, while the BCs for the nested domains are extracted from the CMAQ Chemical Transport Model (CCTM) concentration files of the larger domain. Both of the two models employ 14 vertical layers of varying thickness with denser layers in the lower atmosphere to better resolve the mixing height.

Regional emission inventory
In this paper, a new emission inventory for the YRD was prepared, which consists of large point sources, industrial sources, mobile sources, residential sources, and biogenic sources. The basis of this inventory is a new compilation from local authorities of sources in the 16 major cities of the YRD domain.
Point sources in this inventory consist of power plants and large industrial combustion and process sources. The point source data are obtained from a national environmental statistical database provided by the Chinese Academy for Environmental Planning. Point sources are located at their exact lat/long coordinates within the domain.  Mobile sources consist of on-road vehicle emissions in the 16 major cities. The vehicle volume data are collected from the statistical yearbooks of Jiangsu and Zhejiang Provinces and the Shanghai Municipality (Jiangsu Statistical Yearbook, 2005;Zhejiang Statistical Yearbook, 2005;Shanghai Statistical Yearbook, 2005).
Areas sources in this paper include fugitive emissions from industrial activity and residential fuel combustion. Residential emissions are calculated based on the consumption of coal, LPG, coal gas, and natural gas in 2004, as obtained from the statistical yearbooks of the 16 cities in the YRD.
For biogenic VOC emissions, this study uses the natural VOC emission inventory of the GEIA Global Emissions Inventory Activity 1990 (http://geiacenter.org). In July, total biogenic VOC emissions in Shanghai were 3,286 tons, taking a share of 23% of the total biogenic emissions in the YRD. The GEIA emissions may be outdated due to rapid development of the YRD region and consequent changes in the natural ecosystem, and we are considering to update these emissions in future work.
Emission factor is an important element that influences the estimation of emission inventory. Supporting data on emission factors and activity data were assembled from related studies. Emission factors in this study are mainly referred to widely accepted studies or calculated based on related parameters. The SO 2 emissions from major pollutant sources are calculated based on fuel consumption and sulphur content. Results have been compared and verified with istribution of point sources in the YRD (top: power plants; bottom: industrial point sources); ovince is shown in blue, Zhejiang Province in pink, and the Shanghai Municipality in yellow.  other environmental statistical data. Emission factors and fuel parameters are mainly derived from national literature (Hu, 1990;MEP, 1996), including "User's Guide on the Production and Emission Factors of Industrial Pollutants" and "Environmental Protection Data Manual". Related data have been widely used in other studies (Streets et al., 2000;Kuang et al., 2001;Di et al., 2005). Emission factors of other pollutants like NO x and CO are obtained through calculation or referred to related studies (Fang et al., 1985;Hu, 1990;Kato et al., 1992;Akimoto et al., 1994;MEP, 1996;Streets et al., 2000Streets et al., , 2001Streets et al., , 2003a. Emission factors of VOC, NH 3 , PM 10 and PM 2.5 are mainly obtained through literature survey (Kato et al., 1992;Akimoto et al., 1994;Streets et al., 2000Streets et al., , 2001Streets et al., , 2003aUSEPA, 2006). Currently, we are lacking measurement data in the YRD to verify those emission factors, which could likely be taken into consideration in future studies.
The newly calculated emissions for the 16 cities in the YRD are then inserted into the regional East Asian emission inventory provided by TRACE-P (Streets et al., 2003a,b;Fu et al., 2008), which provides emissions data for the remaining parts of the domain. The TRACE-P emission inventory has been described and demonstrated to be reliable for China in previous studies (Streets et al., 2003a,b;Carmichael et al., 2003).

Results and discussion
3.1 Regional emission inventory for the YRD

Point sources
There were 378 power plants in the 16 cities of the YRD in 2004. In addition, there were 9793 major industrial sources, among which there were 4381 in Jiangsu Province (44.7%), 3864 in Zhejiang Province (39.5%), and 1548 in Shanghai Municipality (15.8%). Figure 2 shows the distribution of these point sources in the YRD. This is the first time that a comprehensive allocation of emissions to major point sources in the YRD has been developed. In 2004, the emissions of SO 2 , NO x , CO, PM 10 , PM 2.5 , and VOC in the power-plant sector were 1381, 727, 157, 660, 346, and 7 Gg, respectively. Emissions from power plants in Jiangsu Province are the highest, representing shares of 48%, 48%, 53%, 60%, 60%, and 51%, respectively, in the power-plant emissions of the pollutants listed above. Although there are only 19 power plants in Shanghai Municipality, they tend to be large ones, with shares of 18%, 19%, 20%, 7%, 7%, and 20%, respectively.
Emissions of SO 2 , NO x , CO, PM 10 , PM 2.5 , and VOC from fuel combustion in the industrial sector are 311,168,38,341,140,and 2 Gg,respectively V eh icle Po p u latio n , th o u san d U n it H a n g z h o u N i n g b o J i a x i n g H u z h o u S h a o x i n g Z h o u s h a n T a i ' z h o u N a n j i n g W u x i C h a n g z h o u S u z h o u N a n t o n g Y a n g z h o u Z h e n j i a n g T a i z h o u S h a n g h a i  Figure 3 shows the vehicle stocks in each of the major cities of the YRD in 2004. There were 3 286 000 cars in the 16 cities, among which Jiangsu Province contains 1 270 000, Zhejiang Province contains 1 141 000, and Shanghai contains 878 000. Among all the 16 cities, Shanghai, Hangzhou, Suzhou, Ningbo, and Nanjing contain the largest vehicle stocks, with shares of 26%, 12%, 11%, 8%, and 8%, respectively. In addition, there were 9 798 000 motorcycles and 7 266 000 light-duty motorcycles in the major cities of the YRD in 2004. Jiangsu, Zhejiang, and Shanghai contain 59%, 40%, and 1% of the total motorcycles and 12%, 74%, and 14% of the light-duty motorcycles, respectively. The compositions of the vehicle fleets are very different between Shanghai Municipality and the two neighboring provinces. Figure 4 shows the vehicle stocks by category in the two provinces and Shanghai. In Shanghai, light-duty cars represent 30% of total vehicles; while motorcycles (including light-duty motorcycles) are a major vehicle category in Jiangsu and Zhejiang Provinces, with shares of 80% and 73%, respectively.

Mobile sources
Based on these vehicle stock data and local emission factors, the mobile source emissions of CO, VOC, NO x , PM 2.5 , and SO 2 in the YRD in 2004 were estimated to be 2302, 446, 358, 62, and 11 Gg, respectively. Regional shares for each of the species were for Jiangsu Province 49%, 50%, 46%, 50%, and 44%, for Zhejiang Province 28%, 30%, 29%, 29%,   and 29%, and for Shanghai 24%, 20%, 25%, 22%, and 28%. Since Shanghai is a megacity with large numbers of vehicles, it has a larger share of mobile-source emissions than stationary sources. Since the vehicle mileage of heavy-duty trucks in Jiangsu Province is very high, the NO x and PM 10 emissions in the cities of Jiangsu Province are high. In addition, the large stocks of motorcycles and light-duty motorcycles are major sources of CO and VOC emissions in the YRD.

Area sources
Emissions from residential fuel combustion are calculated directly from fuel-use data for each of the cities in the region. Table 1 shows the residential fuel consumption in the cities of the YRD in 2004, as well as the populations of the cities. Though coal and LPG are used extensively in Jiangsu Province and Zhejiang Province, gas -both coal gas and natural gas -is preferred in Shanghai. Fugitive emissions from industrial activity are calculated based on the consumption of coal and fuel oil in the YRD after energy consumption in the power plant, industry, transportation, and residential sectors is eliminated. Results show that in 2004, the emissions of SO 2 , NO x , CO, PM 10 , PM 2.5 , and VOC from residential fuel combustion were 16.4, 8.6, 17.6, 2.9, 1.7, and 1.0 Gg, respectively. Emissions of SO 2 , NO x , CO, PM 10 , PM 2.5 , VOC, and NH 3 from fugitive emissions of industry were 288, 114, 257, 28, 19, 26 and 1 Gg, respectively.

Observational sites used to compare against MM5-CMAQ model results
In this paper we use selected hourly concentration data for SO 2 , NO 2 , PM 10 , and O 3 measured during the periods 11-20 January 2004, and 11-20 July 2004, at each of the 43 national observational sites located in the YRD to assess the performance of the model. The locations of the measurement sites are shown in Fig. 5. re 5 Locations of the national observational sites used in the model performance assessment

MM5 performance
The daily average wind speed and temperature data observed on the site of Baoshan in January and July 2004 in Shanghai are compared with the MM5 results, as shown in Figs. 6 and 7. Comparisons between the figures show that the parameters adopted in MM5 could basically reflect the variations of the wind speed and temperature in the region. MM5 correctly reproduces the major patterns of the observed wind flow and that wind speed is within acceptable error limits.
To further study the MM5 performance, several statistical measures including the Bias Error (B), the Root Mean Square Error (RMSE), and the index of agreement (IOA) are calculated as part of the meteorological model evaluation. These measures are calculated for daily average wind speed and temperature at the surface observational site and the 10-mheight model results.
Model daily avg.
Model daily avg. time period (hourly or daily). It is a good overall measure of model performance, which is calculated by Eq. (2): In Eqs. (5) and (2), Oij is the individual observed quantity at site i and time j , and the summations are over all sites (I) and over time periods (J). P i j is the individual predicted quantity at site i and time j .
The index of agreement is calculated by Eq. (3): where o denotes the average observed concentration and a value of 1 indicates perfect agreement between predicted and observed values.
Statistical results between MM5 model and observation data are shown in Table 2. Table 2 demonstrates that the MM5 performance is reliable.  The finest grid resolution is 4 km × 4 km, with the grid area of 16 km 2 . Some observational sites are quite close and are within one grid cell, as is shown in Fig. 5. Thus, we give the monitoring max, min and average values of the monitoring data at several sites within one grid, to be compared with the model result. The average model result means the data of the specific grid that contains those observational sites. And the vertical layer of the model result is only the surface layer. Figure 8 shows a comparison between the observed and modeled SO 2 concentrations at the Nantong, Shanghai, and Hangzhou observational sites during 11-20 January and 11-20 July 2004. It is to be expected that CMAQ would perform relatively well at predicting SO 2 concentrations, because the Temperature comparison between observed and prediction data for January and July, 2004 emission inventory is most reliable for SO 2 emissions that do not involve large contributions from small, dispersed sources, and because the chemistry and transport of SO 2 has been extensively studied for many years. Results show that CMAQ reproduces the variation trends of this pollutant well at each site. This provides additional confidence in the emission inventory of SO 2 in the YRD in 2004 and confirms that the meteorological fields used and the assumptions made in the CMAQ model reflect the true SO 2 pollution situation in the YRD. Figure 9 shows the comparison between model results and observational data of NO 2 hourly concentrations at Nanjing, Shanghai, and Hangzhou monitoring sites during 11-20 January and 11-20 July 2004. Results show that CMAQ can reproduce the variation trends of the pollutant concentrations. The modeling results for Nanjing, Shanghai, and

Nitrogen dioxide
Hangzhou are good and reflect the actual NO 2 pollution status in these cities. However, The NO x emission sources include not only power plants, industrial boilers and kilns, but also the emissions from mobile sources and residential combustion. Therefore, their magnitude and spatial distribution are more uncertain. In addition, of course, the local chemistry of NO x emissions conversion is more complex than that of primary SO 2 emissions. NO x conversion may be affected by possible underestimation of VOC emissions; some fugitive VOC emissions such as painting and evaporative emissions at gasoline refilling stations are not included in the inventory. In addition, the estimation of NH 3 emissions in the 16 city areas may be lower than in reality, because NH 3 emissions from fertilizer usage and livestock breeding in rural areas can be significant but are not included in the inventory. Thus, the transformation from NO x , VOC, and NH 3 to particles may be underestimated in the model, which causes uncertainty in the NO 2 modeling result.

Particulate matter (PM)
Influenced heavily by meteorological conditions, PM 10 concentrations are quite significant in winter time. Figure 10 shows a comparison between model results and observational data of PM 10 hourly concentrations at Nantong, Ningbo, Jiaxing, and Huzhou monitoring sites during 11-20 January and 11-20 July 2004. Results show that CMAQ can reflect the trends of PM 10 in July, while in January the model tends to underestimate the PM 10 concentrations. There are four possible reasons. First, the PM 10 emission inventory for the region may be underestimated, since fugitive dust emissions from construction sites, road transportation, coal stockpiles, etc., are not included. Second, since NH 3 emissions from rural areas are not considered, the transformation from NO x and NH 3 to particles is weaker than expected. Third, VOC emissions in this study may be underestimated since there are some source types that are not included in the emission inventory; thus, fine particles formed through complex chemical reactions involving NO x , VOC, etc., may be underesti-mated. Last but not least, on the occasions of high pollution episodes, particularly in winter, the dominant wind is from a northerly direction and carries pollutants from north to south and affects the regional air quality in the YRD. In such cases, especially when there are dust storms in Northern China, our local emission inventory and even the regional TRACE-P inventory may not include all PM emissions and the modeled results may be lower than the observed ones.

Ozone
Hourly O 3 concentration data were only collected for Nanjing in January 2004 and for Shanghai in July 2004. Therefore, the observational data that are available for model testing are limited. Figure 11 shows the comparisons between model results and available monitoring data for hourly O 3 concentrations in January and July. Results show that CMAQ can reflect the variation trends of the pollutant. Specifically, the model gives reliable results for the summer case. O 3 concentration. With increase of solar radiation early in the day, the O 3 concentration rises; while in the afternoon, with decrease of radiation, the O 3 concentration gradually declines. The ozone concentration is influenced by emissions of precursors like NO x and VOC, VOC speciation, and the temporal profiles of the emissions. The VOC speciation is compiled based on a literature survey (Streets et al., 2003a), but is known to be based largely on western sources, rather than Chinese sources, most of which have never been measured. The modeled O 3 concentration at midnight is not as low as expected, which is due to the PBL height simulation by MM5. In this paper, the MRF parameters are adopted in MM5 modeling, which usually gives high PBL height in daytime and low PBL at night . This systematic error tends to result in underestimation of ozone concentrations in the daytime and overestimation at nighttime. Although the MRF parameter could not give a satisfactory result in the PBL prediction, it is good in the simulation of other meteorological elements like wind speed and temperature. Thus, it is chosen in this study.

Factor 2 analyses
Factor 2 analysis is one of the methods commonly used to check whether the air quality model results are acceptable. Factor 2 calculates the percentages of the ratios of model value to observational value that lie between 0.5 and 2. The calculation formula is shown in Eq. (4): where R is the percentage of the ratios between 0.5 and 2; N [1/2,2] is the number of the ratios between 0.5 and 2; and N t is the total number of comparison points. The larger the R value, the better the model performs. R = 100% means the model performance is perfect. Figure 12 shows the factor 2 analysis results for SO 2 , NO 2 , and PM 10 . Results show that in January, 59%, 80%, and 51% of the ratios between model and observational data for SO 2 , NO 2 , and PM 10 , respectively, are within the factor 2 ranges. For the month of July, 75%, 81%, and 73% of the ratios of SO 2 , NO 2 , and PM 10 , respectively, are within the factor 2 range. Generally speaking, the model performs quite well in the prediction of SO 2 , NO 2 , and PM 10 . However, for ozone, there are only 32% and 50% of the results in winter and summer, respectively, within the factor 2 range. As shown in the figure, there are many points with modeled data higher than the observed ones, which is possibly due to the low PBL height at nighttime obtained by MM5, as mentioned above.

Statistical analysis of the performance results
To further study the model capability and applicability in the model domain, the following statistical measures were applied: normalized bias, index of agreement, correlation coefficient, and factor of two. The normalized bias is calculated by Eq. (5): where p i represents the predicted data and o i represents the observational data. N means the number of data points. A normalized bias less than ±15% is suggested by EPA as indicative of acceptable ozone model performance. The index of agreement is calculated by Eq. (3) as described above.
The statistical results comparing predicted values and observation data are given in Table 3. Table 3 shows that the correlation coefficients between prediction and observational data are mostly >0.5. The indexes of agreement of SO 2 , NO 2 , and PM 10 are 0. 55,0.79,and 0.45,respectively,in January;and 0.79,0.75,0.53 in July. The CMAQ model overpredicts SO 2 and NO 2 concentrations by about 13-25% and 22-29%, while it underpredicts PM 10 by about 24% in January and overpredicts by about 25% in July. The model performance for summer is better than winter. The reason for overprediction of SO 2 , NO 2 , and underestimation of PM 10 is possibly due to underestimation of NH 3 emissions. The secondary transformation process is underestimated as a result. In winter, the meteorological dispersion conditions in the YRD are not good, generally with low precipitation, and frequent inversions; such complex meteorological conditions are not well simulated by meteorological models. For example, the wind speed obtained by MM5 is often overestimated and then the model results for winter are not as good as for summer. Although there are some biases in the model performance, these biases are generally within acceptable ranges.
4 Regional air quality model results and discussion 4.1 Ozone Figure 13 shows the monthly average (the monthly mean of hourly average of the air pollutants) concentrations of O 3 and NO 2 in the 16 cities of the YRD. In January, the monthly average concentration of O 3 is 0.052 ± 0.011 mg m −3 , and 0.054 ± 0.008 mg m −3 in July. In the clean atmosphere away from its sources, NO represents generally less than one third of NO x during the day, and is rapidly titrated after sunset by ozone. In contrast, in the polluted atmosphere, such as in urban areas, where NO is constantly emitted from the surface, ozone is slowly titrated by NO throughout the night, forming high NO 2 mixing ratios. The correlation between the O 3 and NO 2 concentrations shows that the two pollutants are negatively correlated, with a correlation coefficient of −0.85, as shown in Fig. 14. This result agrees with the observational work of Wang et al. (2001a).
To illustrate the formation and dissipation of ozone, Fig. 15 presents the variation of hourly O 3 concentration on 5 July 2004, when the highest hourly ozone concentration occurs during the modeling period. The spatial distribution of ground-level O 3 , influenced by meteorological conditions, varies with time. After the sun rises, the O 3 concentration gradually increases. As time goes by, the primary air pollutants including NO x and VOCs diffuse from urban areas to the downwind region, where the high O 3 concentration occurs due to photochemical reaction. By 13:00-15:00 LT, the high hourly O 3 concentration covers the cities of Shaoxing, Hangzhou, Huzhou, and Jiaxing. The highest hourly concentration of O 3 on July 5 reached 0.24 mg m −3 (110 ppb) in the grid cell (111, 25) (Zhoushan City) at 13:00 LT. In Shanghai city, the figure shows obvious differences of O 3 concentrations at different locations. The ozone concentration in the urban area is lower than that in the rural area, because the emission characteristics in different regions of the city vary significantly and the high O 3 concentration is titrated by NO in the urban area with strong NO x emissions. This result agrees with the work of Geng et al. (2008), who found that  the O 3 concentration is highest in the rural area that has low emissions of O 3 precursors. In contrast, the O 3 concentration in the urban area is low due to the O 3 suppression processes caused by urban emissions.

Visibility
Visibility in the YRD is often poor, as can be readily observed by residents and visitors. This paper applies the concept of deciview (dcv) to reflect the visibility status. This indicator is calculated based on the extinction coefficient shown in Eq. (6).
where dV means deciview, and β represents the extinction coefficient, which is the sum of the scattering and absorption coefficients.
In the United States, dcv is usually divided into five classes to reflect the subjective visibility levels, as shown in Table 4. Figure 16 shows the monthly average dcv values for the 16 major YRD cities obtained from the MM5-CMAQ model.  visibility in the YRD in winter is quite bad, while it is relatively better in summer. This is mainly because the dominant wind in summer is southeast, which is from the East China Sea. Clean marine air makes the air quality good in summer. However, in winter, the meteorological conditions are not good for air pollutants to disperse. The precipitation is rare, and frequency of inversion is high, which causes the air pollutants to accumulate readily. Thus the visibility in winter is relatively bad. Related studies show that visibility is strongly negatively correlated with PM 2.5 , with a correlation coefficient of −0.68 (Huang et al., 2009). It is clear that PM 2.5 concentrations in the YRD cities are worse in the winter than in the summer, leading to poor visibility in winter.
This result agrees with the work of Wang et al. (2006) and Ye et al. (2003), who found that the seasonal variation of ion and PM 2.5 concentrations was significant, with the highest concentrations observed in winter and spring and the lowest in summer and autumn. Figure 17 shows the daily average dcv values for 10 January and 10 July 2004. Comparing the two images, we can see quite clearly that the visibility in winter is much worse than summer. In winter, the visibility in the center of the YRD -Hangzhou, Huzhou, Nanjing, Wuxi, and Changzhou cities -can be quite bad. While in the more eastern and southern areas, like Shanghai and Zhoushan, it is relatively better. 40.00 C h a n g z h o u H a n g z h o u H u z h o u J i a x i n g N a n ji n g N a n to n g N i n g b o S h a n g h a i S h a o x in g S u z h o u T a iz h o u T a iz h o u ' W u x i Y a n g z h o u Z h e n ji a n g Z h o u s h a n monthly visibility, dcv January July Figure 16 Visibility, measured as deciview (dcv), in the 16 cities of the YRD Figure 17 Modeled visibility in the YRD Fig. 16. Visibility, measured as deciview (dcv), in the 16 cities of the YRD. Figure 18 shows the monthly average concentrations of SO 2 in the 16 cities of the YRD. In January, the monthly average concentration of SO 2 is 0.035 ± 0.015 mg m −3 , and 0.019 ± 0.010 mg m −3 in July. As CMAQ overpredicts SO 2 concentrations by about 25% in January and 13% in July, we may anticipate that the real monthly average concentrations of SO 2 in the YRD in January and July are 0.026 ± 0.011 mg m −3 , and 0.017 ± 0.009 mg m −3 , respec-tively. Due to the air pollution transport from Northern China to the YRD, together with the high local emissions, the regional SO 2 pollution in winter is worse than in summer. The pollution centers include Shanghai, Suzhou, Jiaxing, and Ningbo. In summer, due to inflow of clean air from the sea to the southeast, the SO 2 concentrations in cities are lower. The SO 2 concentrations in southern cities are obviously better than in the northern ones. Figure 19 shows the monthly average concentrations of NO 2 in the 16 cities of the YRD. In January, the monthly average concentration of NO 2 is 0.029 ± 0.012 mg m −3 , and 0.018 ± 0.010 mg m −3 in July. As CMAQ overpredicts NO 2 concentrations by about 29% in January and 22% in July, we may anticipate that the real monthly average concentrations of NO 2 in the YRD in January and July are 0.021 ± 0.009 mg m −3 , and 0.014 ± 0.008 mg m −3 , respectively. Similar to SO 2 , Shanghai is also the high pollution center of NO 2 . Comparisons between winter and summer show that, although there is clean air inflow from the sea in summer, the difference between seasons are not so obvious as for SO 2 , which is due to the vehicle emissions in cities. In addition, in summer, due to use of air conditioners in the cars, the emissions are relatively higher than in winter. This result is consistent with the work of Uno et al. (2007), who found that the seasonal variation of NO 2 over W u x i Y a n g z h o u Z h e n j i a n g Z h o u s h a n January July ed as deciview (dcv), in the 16 cities of the YRD 17 Modeled visibility in the YRD H a n g z h o u H u z h o u N a n ji n g C h a n g z h o u N a n to n g N i n g b o J i a x i n g S h a n g h a i S u z h o u W u x i Z h e n ji a n g S h a o x in g T a iz h o u ' Y a n g z h o u T a iz h o u Z h o u s h a n SO2 monthly average conc., mg m

39
H a n g H u N a C h a n g N a n N i J i S h a n S u W Z h e n S h a o T a iz Y a n g T a i Z h o u s January July Figure 18 Monthly average concentrations of SO2 in the 16 cities of the YRD 0.000 0.020 0.040 0.060 N a n to n g H a n g z h o u H u z h o u N a n ji n g C h a n g z h o u N i n g b o J i a x i n g S h a n g h a i S u z h o u W u x i Z h e n ji a n g S h a o x in g T a iz h o u ' Z h o u s h a n Y a n g z h o u T a iz h o u NO2 monthly average conc, mg m -3 January July Figure 19 Monthly average concentrations of NO2 in the 16 cities of the YRD Fig. 19. Monthly average concentrations of NO 2 in the 16 cities of the YRD.
East Asia shows a summer (July-August) minimum and a winter (December) maximum. The result is also similar to that of Wang et al. (2009), who presented a seasonal variation in NO 2 columns in the YRD region that showed higher NO 2 in winter due to the longer lifetime of NO 2 and greater NO x emissions in the region. Figure 20 shows the monthly average concentrations of PM 10 in the 16 cities of the YRD. In January, the monthly average concentration of PM 10 is 0.064 ± 0.023 mg m −3 , and 0.034 ± 0.021 mg m −3 in July. As CMAQ underpredicts PM 10 concentrations by about 24% in January and overpredicts about 26% in July, we may anticipate that the real monthly average concentrations of PM 10 in the YRD in January and July are 0.080 ± 0.028 mg m −3 , and 0.025 ± 0.015 mg m −3 , respectively. The regional PM 10 pollution in winter is much worse than in summer. The pollution centers in winter include Shanghai, Jiaxing, and Shaoxing.

Conclusions
This paper applies the MM5-CMAQ modeling system to the study of regional air pollution in the YRD. Model performance studies show that the system can reliably reproduce the air pollution situation in the region. Model performance results indicate that the correlation coefficients between prediction and observational data are greater than 0.5. The indexes of agreement for SO 2 , NO 2 , and PM 10 are 0.55-0.79, 0.75-0.79, and 0.45-0.53, respectively. The CMAQ model tends to overpredict SO 2 and NO 2 concentrations by about 13-25% and 22-29%, while it underpredicts PM 10 by about 24% in January and overpredicts by about 25% in July. The model performance for summer is better than winter. The model performance assessments show that the model results are acceptable for this application.
Observational data show that the ozone concentrations in summer are quite high, while the visibility in winter is bad. Model results show that the highest hourly concentration of O 3 on 5 July reached 110 ppb, which appeared at 13:00 LT in Zhoushan City, slightly lower than the observed value. The results also show that the deciview values H a n g z h o u J ia x i n g S h a n g h a i S h a o x in g H u z h o u W u x i N a n ji n g S u z h o u Z h e n ji a n g T a iz h o u ' Y a n g z h o u Z h o u s h a n PM 10 monthly average conc., mg m -3 January July for various cities in winter are larger than 20, with an average value of 26.4 ± 2.95 dcv, while the average visibility in summer is 17.6 ± 3.3 dcv. The monthly average concentration of SO 2 in January is 0.026 ± 0.011 mg m −3 , and 0.017 ± 0.009 mg m −3 in July. The monthly average concentrations of NO 2 in the YRD in January and July are 0.021 ± 0.009 mg m −3 , and 0.014 ± 0.008 mg m −3 , respectively. Monthly average concentrations of PM 10 in the YRD in January and July are 0.080 ± 0.028 mg m −3 , and 0.025 ± 0.015 mg m −3 , respectively. Due to pollutant transport from Northern China to the YRD, the unfavorable meteorological dispersion conditions, and the high local emissions, the regional SO 2 pollution in winter is worse than in summer. Clearly, elevated O 3 episodes are occurring in the YRD, and regional haze is becoming more obvious. These results confirm the conclusions reached by Chan and Yao (2008), who reviewed the current state of understanding of the air pollution problems in Shanghai. However, details of these episodes, such as the frequency, the weather conditions, and the temporal and spatial variations of O 3 and haze, can vary significantly with geographical location within the region. Currently, ozone and haze have become extremely important issues in regional air quality. The YRD is presently undergoing tremendous economic growth, and the threat of high regional pollutant emissions and high pollution is very real. In this point, we recommend that integrated measures be taken together in all the cities to improve the regional air pollution situation in the YRD.