Estimating Air Quality Impacts of Elevated Point Source Emissions in Chongqing , China

In this study, the CALPUFF/MM5 modeling system was applied to estimate the air quality impacts of elevated point sources in 2004 in Chongqing. An intercomparison of the performance of CALPUFF against the observed data is discussed and an examination of scatter plots and QQ plots is provided. Results show that in 2004 the high emission contribution induced a relatively high contribution to average ambient concentration and significant impact on the urban area (higher than 10 μg/m of 24-hour averaged SO2 concentration, maximum of 650 μg/m). The highest 24-hour averaged SO2 concentration exceeds the Grade NAAQS by 10.6%, 153.3%, 60% and 333.0% for January, April, July and October, respectively. The concentration distributions demonstrate the heterogeneity patterns in spatial and temporal scales due to significant topographic diversity and weather variations over short distances. The source of the SO2 in the Chongqing area is local air pollution, which results from the lower effective stack height, low wind velocity in the area, basin topography, and the use of coal with high sulfur content. Parametric sensitivity analyses are still needed to determine the magnitude of uncertainty associated with CALPUFF.


INTRODUCTION
Air pollution in Chongqing has obviously been reduced since 2000, but it still has a long way to go and big challenges in air quality improvement (Guttikunda et al., 2003). *Corresponding author. Tel.: +86-10-62842631; Fax: +86-10-62911045 E-mail address: yangduoxing@yahoo.com A coal-dominated energy structure is one of the major causes of air pollution in Chongqing and elevated point sources take about two-thirds of the total coal consumption. Elevated point sources are significant emitters of sulfur dioxide (SO 2 ), which is harmful at high concentrations and contributes to the formation of atmospheric fine particulates. It is estimated that Chongqing's elevated point sources emitted 75.6% of the total SO 2 emission (Chongqing EPA, 2004). Evaluating the air quality impact of these elevated point sources is a necessary step in the design of a comprehensive cost effective air pollution control strategy in Chongqing.
Although several studies have estimated the air quality impacts for elevated point sources (e.g., power plants) emissions in China (Zhou et al., 2003(Zhou et al., , 2006Hao et al., 2007), they could well differ in Chongqing settings, because of differences in stack characteristics, topography and meteorology. There are several relevant pieces of research into this area (Lei et al., 1987;Gao et al., 2000). In the study of (Gao et al., 2000), a Eulerian sulfur deposition model was used to study the distributions of sulfur pollutants in East Asia, and it was revealed that a big SO 2 value area lies in Sichuan basin with the biggest level of 47.3 g/m 3 near Chongqing. But the spatial resolution of this model was low and the SO 2 emission inventory applied wasn't full, especially for Chongqing. In addition, this study couldn't take into account single contribution from elevated point sources to local SO 2 levels in Chongqing. One study (Lei et al., 1987) modeled SO 2 air pollution from all anthropogenic sources, but only the urban area, 12% of whole Chongqing area, was covered. On the other hand, this study applied its model with meteorological data from one station, which couldn't capture the relatively real meteorological variations over short distances. Currently available air quality estimates from Chongqing used simplified models (e.g., Gaussian plume model), which may not capture the full impact of a pollution source. With the increment of older fossil-fuel elevated point sources, the impact of this large emission sector on Chongqing's air quality needs further assessment.
In this study we focus on the air quality impact of Chongqing local elevated point sources to the Chongqing area. Emissions of SO 2 in 2004 were estimated using monitored emission factors.
The CALPUFF/MM5 modeling system was applied to model the concentration increments of SO 2 . The aim of this study has been to evaluate current status of the air pollution from elevated point sources in Chongqing, and gain information on the necessity for further control.

Modeling structure
The CALPUFF/MM5 modeling system, which is commonly used in air quality impact studies dealing with routine industrial releases, was deployed to simulate the meteorology as well as pollutant dispersion and transport in Chongqing. There are several reasons for choosing this model. Firstly, it has the ability to generate and handle complex three-dimensional wind fields and has the complex terrain algorithm (Scire et al., 1999a), which is necessary when the modeling domain is expanded to whole Chongqing area. Secondly, it can treat calm wind conditions (< 1.0 m/s), the frequency of which in Chongqing is about 47% (Tao et al., 2006). This allows the modeling of both primary and secondary pollutants and is not too complex to apply to evaluate a single sector, as CAMx or CMAQ, is also attractive.
CALMET is a diagnostic wind field model that produces hourly ranges of meteorological parameters, including wind, temperature, mixing height and dispersion properties on a three-dimensional gridded modeling domain by interpolating routine surface and upper-air meteorological data. The data interpolation accounts for the blocking effects of terrain and slope flows. It also ensures the preservation of air-mass continuity (Scire et al., 1999a). Since diagnostic models interpolate observational data, they are critically dependent on the quality and density of the observations. The CALMET model can ingest prognostic model outputs, such as the Fifth-Generation NCAR/Penn State Mesoscale Model (MM5), and has displayed improved predictions of certain variables (Chandrasekar et al., 2003).
Hence, to develop meteorological data for CALMET, the MM5 version 3.7 was used.
The CALPUFF model is an advanced Lagrangian-Gaussian non-steady state airquality model, promulgated by the U.S. Environmental Protection Agency (U.S. EPA) as the regulatory model for assessing the long-range transport of pollutants and their impacts, on a case-by-case basis, for certain near and far-field applications involving complex meteorological conditions (Scire et al., 2000). It simulates continuous puffs of pollutants released from a source into the ambient wind flow. As the wind flow changes from hour to hour and from place to place, the path of the puff changes according to the new wind flow direction. Puff diffusion is Gaussian, and concentrations are based on the contribution of each puff as it passes over or near a receptor point (Scire et al., 1999b).

Domain coverage
For this study, we evaluated the aggregate impacts of elevated point sources in Chongqing. The southwest corner of the domain is located at longitude 107.7E, latitude 31.4N; the northeast corner is located at longitude 108.3E, latitude 32.0N. There are 1000 receptor points in both the X and Y directions, with a grid spacing of 0.05 km (Fig. 1). The entire domain, 50 km × 50 km, covers most of Chongqing's area and all its heavily populated regions with Lambert projection.

Source characteristics and emissions
In this study we evaluate the aggregated air quality impacts of all the elevated point sources in Chongqing in 2004. There are 29 elevated point sources (including three public power plants, the total generating capacity is 2920 MW) at present. The power plants provided about 48% of the total electricity consumption in Chongqing in 2004. All point sources are coal burning  (older fossil-fueled). Fig. 1 shows the locations of these elevated point sources.
Most of them are in urban area. It not only induces air pollution in the most populous area, but also influences the cityscape of the political, cultural and commercial center of Chongqing. Source characteristics and emissions of these elevated point sources are summarized in m near the ground to 800 m near the top of the domain. This allowed CALMET to interpolate from a higher-to a lowerresolution grid (since CALMET uses ten vertical layers). One-way nesting was used to generate ambient wind fields at multiple grid-cell resolutions (27-, 9-, 3-km)  The meteorological data input for initialization included the NCEP global surface observations data, as well as the NCEP global upper air observation data.
The medium-range forecast (MRF) boundary-layer scheme and a cloudradiation scheme that accounts for long-and short-wave radiative transfers were employed. Deep convection was represented on 27-km and 9-km resolution domains using the Kain-Fritsch scheme. All three domains used an explicit microphysical scheme, with a simple ice phase to represent resolved-scale saturated processes. The NCEP Final Analysis (FNL) data archived at NCAR is for every six hours at a spatial resolution of 1 o × 1 o at standard pressure levels under 100 hPa. The data include twodimensional variables, including sea surface temperature and sea level pressure, and three-dimensional variables of temperature, geopotential height, U and V components, and relative humidity.

CALMET
The CALMET domain was the same as the final CALPUFF domain as described in Section 2.3 and the Lambert Conformal map projection was used to account for the earth's curvature. The two standard parallel latitudes were set at 20N and 60N. The reference longitude and latitude were set at 100E and 25N. These quantities were used to adjust observed and prognostic winds to fit the Lambert Conformal mapping. CALMET uses fewer vertical layers than MM5, in part because air pollution modeling does not require detailed information on the upper atmosphere. Ten vertical layers with cell face heights at 0, 20, 40, 80, 160, 320, 640, 1000, 1500, 2200 and 3000 m were used in the CALMET runs. Due to lack of available meteorological observation data, the MM5 was used in this study to develop high-resolution, three-dimensional meteorological fields (i.e., wind, temperature, pressure, etc.). MM5-generated meteorological fields are used to drive the CALMET model which can use important information contained in MM5 data to resolve terrain features, such as terrain channeling and gravity-driven slope flows (Yang et al., 2006). CALMET automatically interpolated from the MM5 model grid system to the CALMET grid. In our case, the gridded geophysical data, including terrain height and land-use category, were generated from USGS terrain and land-use database (www.usgs.gov).

CALPUFF
CALPUFF is a Lagrangian, non-steadystate, gridded puff model for modeling the transport, dispersion, chemical reactions and deposition of air pollutants in the atmosphere. In this study, SO 2 was modeled. MESOPUFF II chemical mechanism was chosen and hourly ozone monitoring data were input into the model to estimate the daytime oxidation of SO 2 . Under this chemical transformation method, daytime SO 2 oxidation is an hourly varying function of background ozone concentration, solar radiation, atmospheric stability, and relative humidity, with a night-time rate of 0.2 percent/h (Zhou et al., 2003). The plume rise, stack tip downwash and vertical wind shear above the stack top were taken into account in our study. Dry deposition of SO 2 was calculated by using the model default values, such as diffusivity, reactivity, resistance and solubility. For wet deposition, CALPUFF uses an empirically based scavenging coefficient method. We used CALPUFF default scavenging coefficients for SO 2 , with liquid precipitation coefficients of 3 × 10 -5 (Zhou et al., 2003;Hao et al., 2007).

Length of simulation
The year 2004 was chosen for the analyses. We calculated SO 2 concentration for one period in each of the four seasons, which were 1-31 January, 1-30 April, 1-31 July, and 1-31 October, respectively.

Model validation
The modeling system was evaluated comprehensively to ensure reasonable estimates of ambient SO 2 . We present a time series of the SO 2 concentrations at the monitor denoted by the star as shown in Fig.   5. For example, Fig. 5a compares hourly values (mg/m 3 ) (from January 1-31, 2004) for the monitor against the modeled estimates for the grid cell in which that monitor is located. At first glance, the SO 2 prediction seemed to be in relatively good agreement with the observed data. Fig. 5b shows the scatter plots of the modeled concentrations versus the observed values. The straight line in Fig. 5a Fig. 3 (a-d) depicts the patterns and magnitudes of the maximum 24-hour averaged sulfur dioxide concentration increments for each season. As anticipated, the concentration increments for sulfur dioxide are heterogeneous with concentrations peaking further from the source and diminishing more slowly with distance from the source. It's seen from Fig.  Fig. 4. (a-d) Monthly averaged SO 2 concentration increments (µg/m 3 ) by elevated point source emissions in Chongqing in 2004. Gray background represents the terrain elevation.

SO 2 concentrations
3 that the CALPUFF simulation predicts concentrations mostly in the northwest region of Chongqing in amounts between 10 and 160 g/m 3 during February. From April 1 to 30, sulfur dioxide concentration mainly distributed in the west, with values ranging from 20 to 380 g/m 3 . The highest SO 2 concentration is 380 g/m 3 , which already exceeds the Grade NAAQS (National Atmospheric Ambient Quality Standard) (150 g/m 3 ). The high concentration area shrinks greatly. As for July, the sulfur dioxide concentration varied from 10 to 240 g/m 3 and the influenced area gradually expanded. The highest SO 2 concentration reached 240 g/m 3 , equaling the 160% SO 2 standard. During October the SO 2 concentration ranged between 10 and 650 g/m 3 . The highest SO 2 concentration was 650 g/m 3 , equaling to 433% of the SO 2 standard. The influenced area almost covered the whole study region. Given the complex terrain and wind field variation affecting air pollution distribution, the concentration distributions show heterogeneity patterns in spatial and temporal scales. The sulfur dioxide concentration was higher in the valley than on the mountain, because of meteorological parameters impacting both the deposition processes and the transport phenomena, which govern the evolution of these distributions. Another cause is that SO 2 is heavier than air, it tends to accumulate in low-lying areas.

Discussion
As shown in Figs. 3 and 4, the averaged SO 2 concentrations vary even in a range of two orders in magnitude. The higher concentration appears near the power plants and the most polluted areas are northwest of urban area. The central urban area is the second most polluted region. Many factors may contribute to this difference, such as meteorological conditions, proximity of heavily polluted point sources to receptors, and the geographic domains deployed in each study. The three power plants have a significant influence on SO 2 concentration in the west and central urban area due to their intensive urban locations. Of the total SO 2 emitted from 29 stacks 92.38% resulted from the three power plants. Pollution from the power plants cannot be neglected for this area if it is to reach the air quality standard. As seen from Figs. 3 and 4, the less polluted areas are southeast of the urban area due to the fact that this area is far away from the three power plants, and the local meteorological conditions don't support the long distance transport of the SO 2 emissions.
Unfortunately, given the lack of available observed data, a comparison between these discrepancies can't be performed. Efforts to improve the observed data should be undertaken in future studies of this sort.
It is inferred from the calculated results that sulfur dioxide concentration distribution is characterized by significant topographic diversity and climate variations over short distances.
Strongly stratified local meteorological circulation can result in very complex surface and boundary layer meteorology (Yang et al., 2007) (Yang et al., 2006).
Difficulties associated with the lack of information on certain modeling input parameters should be reiterated, given that it was the largest estimated source of uncertainty associated with modeling. For example, there are some of the primary elements of parametric uncertainty within our CALPUFF application. This includes uncertainties that can be quantified (e.g., the incorporation of wet and dry deposition, the choice of chemical conversion mechanism, background pollution concentrations, and the size of the receptor region) and uncertainties in the meteorological data (Levy et al., 2002). For deposition, we would expect substantial uncertainty in the plume depletion terms that produce wet and dry deposition losses. Past researchers have found that uncertainties of at least an order of magnitude exist for dry deposition of small particles (Levy et al., 2002) and that dry deposition velocity and scavenging coefficients range by two to three orders of magnitude across studies (McMahon et al., 1979). Wet deposition would be expected to be just as uncertain, especially related to the uncertainties involved with setting scavenging coefficients. Thus, even ignoring the fact that a deposition-based impact model should include indirect exposure pathways and environmental degradation associated with acid precipitation, our baseline model using CALPUFF-default deposition parameters could underestimate total impacts if deposition is overstated.
As As to mitigating SO 2 , control measures such as fuel substitution and flue gas desulfurization will greatly mitigate the SO 2 , especially alleviating the pressure on the urban area to reach the National Atmospheric Ambient Quality Standard (NAAQS). Desulfurization measures taken before 2010 will reduce SO 2 emission by 94%, which will greatly mitigate the SO 2 pollution. Therefore, the highest 24-hour averaged SO 2 concentration decreases to 39 g/m 3 . SO 2 pollution will be mitigated with 333% decrease in concentration for reaching the 24-hour averaged SO 2 concentration standard. It should be worthwhile assessing the individual contributions of the 29 point sources to the surface SO 2 concentrations.
The sensitivity of source reduction to the improvement of SO 2 concentrations should be demonstrated in detail in future studies of this sort.
Nevertheless, this is the first time the elevated point source induced plumes were simulated using CALPUFF in Chongqing.
The study points to some critical auxiliary information needed for satisfactory modeling of sulfur dioxide concentration, which needed to be estimated in the present study due to lack of availability. Future studies can be directed to fill in such knowledge gaps.

CONCLUSIONS
The CALPUFF dispersion model was applied to estimate the air quality impacts associated with current SO 2 emissions from a set of 29 elevated point sources in the Chongqing area in 2004. Results show that these point sources provide maximum 24hour averaged SO 2 concentration increments of up to 10 g/m 3 (maximum increment of 650 g/m 3 ), with monthly averaged values of up to 5 g/m 3 (maximum increment of 180 g/m 3 ). In 2004, the emission contribution of the point sources had a large impact on Chongqing's average ambient concentration. In some local areas, especially the west and most populous middle urban, they produced notable SO 2 pollution. Concentration distributions show heterogeneity patterns in spatial and temporal scales due to significant topographic diversity and climate variations over short distances. The magnitude of these concentration increments is potentially significant and illustrates that accurate longrange dispersion modeling can provide meaningful and policy-relevant information for the regulatory community. It points out that further investigation is needed to determine the magnitude of uncertainty associated with CALPUFF (e.g., deposition parameters, chemical conversion mechanism) by using parametric sensitivity analyses.

ACKNOWLEDGMENTS
This project research was sponsored by Institute of Crustal Dynamics, CEA (project number: ZDJ2008-08) and CSTC (Project NO. 2005AC7097). The authors greatly appreciate Dr. Chuen-Jinn Tsai, as well as the anonymous reviewers for their comments which have much improved this manuscript. We also thank Dr. Gangcai Chen for providing the emission inventory and the observed data. GLOSSARY CALPUFF: An advanced non-steady-state meteorological and air quality modeling system developed by ASG scientists. The main components of the modeling system are CALMET (a diagnostic 3dimensional meteorological model), CALPUFF (an air quality dispersion model), and CALPOST (a postprocessing package).

CALMET:
One of the main components of the CALPUFF modeling system, a diagnostic 3-dimensional meteorological model.