Atmospheric PM2.5 Characteristics and Long-Term Trends in Tainan City, Southern Taiwan

In order to analyze the characteristics and long-term trend levels of PM2.5 in Tainan City, Taiwan, Mesoscale Modeling System Generation 5 (MM5) and Community Multiscale Air Quality Model (MM5-CMAQ) modelling as well as box plots and time series analysis, were utilized in this study. The long-term trend analysis shows that the levels of PM2.5 (averaged at 38.3 μg m and ranged between 33.1 and 41.9 μg m) in Tainan City for ten years (2005–2014) were above the yearly average standards of 15 μg m, showing non-attainment status. Overall, the results show a decreasing trend (from 41.9 μg m in 2005 to 35.0 μg m in 2014) in the levels of PM2.5 in Tainan atmosphere in the ten year period. The results of the MM5-CMAQ air quality modeling, indicate that the highest contribution on PM2.5 in Tainan City was from trans-boundary pollution from neighbouring cities (34.2%), while long-range transport and local emissions from Tainan each contributed a fraction of approximately 32.9%. In terms of local sources, the highest influence is from area sources (18.6%), followed by line sources (7.7%) and point sources (6.6%). Thus, to control PM2.5 in Tainan City, the focus should be on construction, road dust, and residential activities.


INTRODUCTION
Particulate matter is basically aerosols existing as suspensions of solid or liquid particle in a gas.Currently, two most important kinds of particulate matter (PM) are the PM 2.5 and PM 10 .The aerodynamic diameters of PM 2.5 are from ~0 to 2.5 µm, while those of PM 10 ranges from ~0 to 10 µm (Chow et al., 2015).Both PM 2.5 and PM 10 are made up of other subclasses of pollutants with the major ones being water soluble ions such as SO 4 2-, NH 4 + , and NO 3 -and metal elements as minor constituents (Tsai et al., 2011;Xu et al., 2012;Chang et al., 2013) as well as carbonaceous species such as the organic carbon (OC) and elemental carbon (EC) (Chang et al., 2013;Ho et al., 2006) and volatile organics (Cheruiyot et al., 2015).Particulate matter is emitted into the atmosphere from primary sources which can be either natural or anthropogenic (Kong et al., 2014;Alghamdi et al., 2015) and also from secondary formation in the atmosphere.Specific sources include dust from roads, emission from vehicles, wood and open burning, sea salt, agriculture and construction industries (Fu et al., 2009).The distribution of PM 2.5 and PM 10 in the atmosphere is influenced by local emissions as well as regional and long range transport phenomena when air masses move through polluted pathways towards the target sites (Lv et al., 2015).
PM 2.5 and PM 10 have been found to have adverse effects on human health, cause deterioration of visibility as well as influencing global climate change (Pope, 1996;Diaz, 2008;Bell et al., 2009;Bell, 2012;Choi et al., 2012;Chang et al., 2013;Chow et al., 2015;Wang et al., 2015).In fact, the World Health Organization (WHO) Global Burden of Disease (GBD) project estimated that about 3.2 million premature deaths that occurred worldwide in 2010 were related to ambient particulate matter (Gao et al., 2015).More importantly, as result of its size, PM 2.5 has received a lot of attention recently.Its fine size allows it to penetrate the gas-exchange membranes and even transfer across the circulatory system causing collapse of organs (Bell, 2012;Li et al., 2012;Xu et al., 2012;Huang et al., 2015;Ma et al., 2015;Hwang et al., 2016).
Recently, Lu et al. (2016) described Tainan as coastal city with a mean population density of approximately 860 people km -2 and a maximum of 14,072 people km -2 in some regions.The area has about 9,000 factories and a vehicular (motorcycles, cars, buses, and trucks) count of approximately 1.94 million.In the northern Tainan, there is a large agricultural region comprising of about 7,000 acres, which is a major source of particulate matter from open burning.Monsoon winds coming from the North and South during the dry seasons of fall and spring as well as the wet season of summer contribute to the PM distribution in Tainan via long range transport pollution from South East Asia and China.In Taiwan, the new PM 2.5 standards in the ambient air were announced in 2012, with 35 and 15 µg m -3 as 24-hour mean and yearly average standards, respectively.As for PM 10 , the 24 hour mean standard and yearly average standards are 35 and 65 µg m -3 , respectively.
The aims of this study were to understand the current air quality status and evaluate conformity with National Air Quality Standards and provide useful information for the decision making, especially, the establishment of air quality control strategies.Additionally, this study aimed at assessing the effect of long-range transport on the air quality in Tainan City.
Generally, the short term air quality data, i.e., daily and diurnal data are easily affected by changing metrological conditions.Therefore, in order to model correctly the local and trans-boundary sources of PM 2.5 and PM 10 contributions over Tainan, the previous air monitoring data from 2005 to 2014 were collected, and analyzed for the long-term variation to evaluate the air quality trends.
The yearly averages were used to calculate basic statistical data and combined with Box Plot and Time Series Analysis to analyze the long-term spatial and temporal variations.The ratios of PM 2.5 /PM 10 were analyzed to resolve the characteristic of spatial distribution and the relationship between PM 2.5 /PM 10 and air quality.Additionally, the distributions of PM 2.5 and PM 10 over Tainan City were simulated by using a combination of MM5-CMAQ air quality models to evaluate both trans-boundary and local sources, and project the best available control strategies.

Long-Term Variation of PM 2.5 and PM 10
The air quality monitoring data for PM 10 and PM 2.5 were collected in Tainan from 2005 to 2014 from four air quality monitoring stations at Tainan, Annan, Shanhua and Xinying which are shown in Fig. 1.Using this data, Box Plots and Time Series Analysis were used to analyze the long-term temporal variation of PM 2.5 and PM 10 .
In this study, a combination of Mesoscale Modeling System Generation 5 (MM5) and Community Multiscale Air Quality Model (CMAQ) chemistry model was used to simulate the effect of regional transport on the air quality in Tainan including the local sources (point, line and area source), which influence the concentration of PM 2.5 and PM 10 in Tainan City.
In this study, the pollutant reduction objectives from TWEPA for 2020 (target year) as well as the target reduction objectives for the year 2014 from each county were used to evaluate the control strategies for emission reduction of each pollutant including PM 2.5 , PM 10 , SO x , NO x and NMHC.
The MM5 is defined as a limited area, non-hydrostatic and terrain-following-sigma-coordinate model that simulates meso-scale and regional scale atmospheric circulation (Zhou et al., 2012).The MM5 provides 4D metrological parameters to CMAQ, which in turn generates output data.In order to successfully couple MM5 and CMAQ for simulation, a meteorology-chemistry interface processor (MCIP) is used (Gehrig and Buchmann, 2003;Mao et al., 2006).CMAQ is an Eulerian grid model developed by the US EPA that is used to simulate the physical and chemical processes like advection, diffusion, atmospheric chemical reactions, cloud effects and dry deposition that affect the transport and fate of pollutants in the atmosphere (Jiménez-Guerrero et al., 2008;Wong et al., 2012;Zhou et al., 2012).The results of the simulation can output 3-D air pollution concentrations and 2-D dry deposition by hour.In order to investigate the effect of regional pollution on the air quality levels in Tainan, a wide spatial range was simulated by using four domains of nested grids of a ratio 1:3 with a domain origin centered around 22°98'44.02"N120°20'33.81"Eas shown in Fig. 2. Assuming pollution emissions from China, the parent domain designated as domain 1 contained coarse grids of 81 km × 81 km, which covered the larger Eastern Asia into which the partial fine grids domains covering Taiwan and Tainan were fitted.The target domain, domain 4 simulated a scale of 3 km × 3 km covering specific air basin in Taiwan.
Prior to modeling, the MM5 model generated 4-D meteorological fields and meteorological parameters, which were input into the model.The initial and boundary conditions were obtained from data or assumptions.The anthropogenic emission data from other eastern Asian countries were obtained from the Intercontinental Chemical Transport Experiment-Phase B (INTEX-B) database.The base year of this database is 2006 and in this study the assumption was that there was no change in the pollutant emission levels between 2006 and 2010.On the other hand, the anthropogenic emission data from China was extracted from the Multiresolution Emission Inventory for China (MEIC) estimated by Tsinghua University and whose base year was 2010.The pollutants obtained from these two databases were SO 2 , NO x , CO, VOC, PM 2.5 , PM 10 , BC and OC, whose sources included power plants, industries, residential areas, traffic and agriculture.The East Asia Biogenic Emissions Inventory System (EABEIS) was used to estimate the biogenic emissions from Eastern Asia including NO and 33 kinds of VOCs (isoprene, monoterpenes etc.).The Taiwan Emission Data 8.1 system (TED8.1)was used to represent the anthropogenic emissions in Taiwan, while the biogenic emissions were estimated by biological emission estimating model.
Emission data extracted from emission inventory, Taiwan Emission Data System 8.1 (TEDS8.1),and emission data from Eastern Asia were input into MM5-CMAQ to simulate the air quality levels over Tainan for 4 months (January, April, July and October), representing the winter, spring, summer, and fall seasons, respectively in 2010 (i.e., basic case).The schematic procedure of MM5-CMAQ analysis is as shown in Fig. 3 Performance Assessment and Quantitative Analysis of the Model After modeling the basic case, the reduction case for 2014 was simulated for Tainan City by using the emission reduction strategies proposed by Taiwan EPA, to evaluate the effect of emission reduction strategy.These modeled results were further compared with measured data to make sure the accuracy of the model.The measured data is taken from TWEPA monitoring sites and the criteria for performance assessment followed the guidelines provided by TWEPA in "The Standard of Air Quality Models" (TWEPA, 2015).
The performance assessment was done using the Mean Fractional Bias (MFB) and the Mean Fractional Error (MFE), which were used to evaluate the accuracy of simulated results by quantitative analysis of air quality results from stations Mean Fractional Error (2) whereby P i,k represents the simulated concentrations for i th day and k th station ; O i,k represents the monitored concentrations for i th day and k th station ; and N is the number of modeled days, while M stands for the number of the sampling stations (excluding traffic stations and specific stations)

Long-Term (Decennial) Trend of PM 2.5 and PM 10
Short-term air quality presentations are not dependent, since they are easily affected by the constantly changing meteorological conditions.Therefore, in this study the decennial trends of the PM 2.5 and PM 10 for Tainan City were presented by box plots for the period 2005 to 2014.Additionally, the annual trend for 2014 was presented to evaluate the seasonality of PM 2.5 and PM 10 in Tainan City.The box plots are advantageous in that they can present the median, average, upper and lower quartile data all at once (Ferreira et al., 2016).The yearly average concentrations are shown in Table 1 while the decennial trends for PM 2.5 and PM 10 are as shown in Figs. 4 and 5, respectively.The purple lines in the figure represent the median values, while the rhombic marks stand for the average values and the orange lines are the trend lines.The longer box means that the variation is large.
From the time series presented in Fig. 4 it is evident that there is decreasing trends of PM 2.5 levels for all the stations in Tainan City during the period 2005-2014.The levels of PM 2.5 decreased consistently but still were higher than the PM 2.5 air quality standard of 15 µg m -3 .This observation on PM 2.5 decreasing trend mirrors that reported in our recent study (Lu et al., 2016).
The decennial PM 10 trend lines in Fig. 5 show a slight decreasing trend for Shanhua, Annan and Tainan stations while there were small variations observed for Xinying station.Over the ten year period, the average annual concentrations were all above the PM 10 air quality standard of 65 µg m -3 as observed in Table 1.

Annual Trends for PM 2.5 and PM 10
The annual trends for PM 2.5 and PM 10 for the year 2014 are represented in Figs. 6 and 7 based on mean monthly concentrations of PM 2.5 and PM 10 , respectively.For both PM 2.5 and PM 10 , the trends are similar showing that lowest pollution levels were observed in summer (June-August) and the increase steadily through Autumn and Winter (September-January), but decreased in February before rising again in March and decrease in Spring (April and May).Recently Lee et al. (2016), showed the similar seasonal trend for PM 2.5 and PM 10 for data collected in Kaohsiung City adjacent to Tainan City on the south.Similarly, Chen et al. (2013) showed that PM 2.5 and PM 10 have similar spatial trends, with corresponding highs and lows.
Lower particulate levels observed in summer were possibly due to abundant precipitation, higher wind speeds and stronger convection that encourage dispersion of particulates (Marcazzan et al., 2001).On the other hand, in winter, greater particulate levels can be attributed to reduced dispersion due to lower height of mixing layer and temperature inversions (Terzi et al., 2010) and trans-boundary contribution by northeastern monsoon wind.Additionally, Eldred et al. (1997) noted that higher organic content in winter may contribute to higher particulate levels.To confirm this, Lu et al. (2016) observed that the elemental carbon (EC) and organic carbon (OC) contribution in PM 2.5 in Tainan City (Xinying, Shanhua, Annan and Tainan stations) was greatest in winter than in summer.

Ratio of PM 2.5 /PM 10
To understand the contribution of the coarse and fine   matter the ratios of PM 2.5 to PM 10 were used and the annual mean averages were tabulated in Table 2 for the period 2005-2014.The monthly PM 2.5 /PM 10 ratios were depicted in Fig. 8 showing the seasonal variations.
Table 2 shows at Xinying station with a decennial average ratio of 0.49, the contribution of coarse particles was dominated, while for other station which had PM 2.5 / PM 10 ratios of greater than 0.50 the fine particulate were dominant in the ten year period.Tainan station which is in the urban area had the highest ratios in the observation period probably due to influence of traffic and vehicle emissions which are major sources of PM 2.5 in the atmosphere (Lu et al., 2016).On the other hand, when comparing Shanhua, Xinying and Annan, Shanhua had the highest    average ratio (PM 2.5 /PM 10 = 0.52) possibly because it is more remote while Shanhua and Xinying are more likely to be influenced by local sources (Parkhurst et al., 1999).Overall, in all the stations, the highest contribution (PM 2.5 /PM 10 ratio = 0.57) by fine particulate was observed in 2007, while the lowest was observed in 2012 (PM 2.5 /PM 10 ratio = 0.48).
In Fig. 8, the trend lines for all the seasons show that the contribution of PM 2.5 was greatest in Tainan station followed by Shanhua station, Xinying station and lowest in Annan station in that order.Additionally, it is clear the PM 2.5 contribution was least in summer and spring and highest in the onset of fall.Gehrig and Buchmann (2003) noted that lower contribution of PM 2.5 in spring maybe due to higher presence of coarser biogenic particles.The higher contribution in winter can be associated with conditions favoring increased rates for secondary aerosol formation resulting from oxidation of SO 2 to SO 4 2-, NO 2 to NO 3 2-and NH 3 to NH 4 + (Parkhurst et al., 1999;Lu et al., 2016).

The Spatial Analysis of Air Quality Modeling in Basic Case
The spatial distribution of monthly mean concentration for PM 2.5 , PM 10 , SO x , NO x and NMHC were simulated for the basic case in January, April, July and October in 2010 and are as shown in Fig. 9. From the Fig. 9, it is evident that the distributions of PM 2.5 and PM 10 are similar, and highest month is January and lowest in July.In January the concentrations of PM 2.5 and PM 10 are higher than 45 µg m -3 and 75 µg m -3 , respectively, in southern Taiwan.On the other hand, Western Taiwan has higher particulate levels compared to Eastern Taiwan.Urban areas like Kaohsiung and Taoyuan cities show greater PM concentrations than other areas.
Similarly, the NO x and SO x levels in western Taiwan exceeded those recorded for eastern Taiwan possibly due to higher number of highways and freeways in the western part of Taiwan.The highest levels (> 30 ppb) and 7 ppb for NO x and SO x , respectively, are observed in January (winter) and lowest in July and are concentrated in urban areas of Taoyuan City, Taichung City and Kaohsiung City.
Pingtung County shows the highest NMHC concentration greater than 60 ppb during the simulation period.Other areas with high NMHC levels are Yunlin and Changhua Cities.Similar to other pollutants, the highest levels of NMHC are observed in January and lowest in July.As for NH 3 , the temporal variation is different in that the highest episodes are observed in October with levels of over 300 ppb and more concentrated the northern cities of Taipei, New Taipei and Taoyuan.

Performance Analysis of the Basic Case Modeling
Following the guidelines of Standard of Air Quality Modeling Provided by TWEPA, the modeled results for the basic case were qualitatively and quantitatively evaluated.The qualitative analysis involved correlation analysis between the results obtained and actual observed data obtained from the air quality monitoring stations in Tainan City.The metrics used for quantitative analysis were Mean Fractional Bias (MFB) and Mean Fractional Error (MFE), which were used to determine the accuracy of simulated results by com paring with data from neighboring counties.The results of correlation analyses for the observed and simulated values were as shown in Fig. 10.The correlation analyses for each pollutant were done for daily average during simulated period (January, April, July and October) in Tainan Station and showed high confidence when comparing the simulated results to actual data.The correlations are high for PM 2.5 (R = 0.76) and NO 2 (R = 0.66), but low for SO 2 (R = 0.46).The results show that simulated values of PM 2.5 and SO 2 were underestimated.On the other hand, the values of NO 2 were overestimated at low concentration conditions, and underestimated during high concentration conditions.
Table 3 shows the corresponding results of four month quantitative analysis computed using MFB and MFE for Tainan and the neighboring cities or counties, while Table 4 shows the MFB and MFE results for the each month.
Results from both tables show that the model simulations pass the guidelines for model performances with acceptable accuracy.According to Boylan and Russell (2006), the criteria for model performance is when the MFE and MFB are less than or equal to 75% and 60%, respectively, while the model goals are met when the MFE and MFB are less than or equal to 50% and 30%, respectively.Based on the Boylan and Russell (2006) recommendations, the model performance of this study was met, but for the performance goals only the PM 2.5 at Xiaogang station from neighboring Kaohsiung City did not meet the requirements.Additionally, the Taiwan EPA has evaluation guidelines for MFB and MFE for PM 2.5 , SO 2 and NO 2 as shown in Table 3 and similarly only the PM 2.5 at Xiaogang Station in Kaohsiung City did not meet the criteria.Obs.

Simulation of the Effects of Long-Range Transport on the PM 2.5 in Tainan City
To understand the extent by which the PM 2.5 was affected by the long-range transport in Tainan, two cases were simulated.One case evaluated was non-eastern Asian emission from Taiwan, while the other scenario was no emission in simulated region.These results were compared with the results of basic case to understand the level of effect by long-range transport using the following relationship for the four stations and four seasons in 2010.

Long-range Transport = Basic case (A) -Case which was non-eastern Asian emission (B) + Case which was no emission in simulated region (C)
The results of quantitative of long-range transport were as shown in Table 5.The most affected station was Xinying Station (9.23 µg m -3 ), and the lowest was Shanhua Station (8.75 µg m -3 ).Similarly, the highest affected ratio was Xinying Station (36%) and the lowest one was Shanhua Station (33%).The average contribution by long-range to PM 2.5 in Tainan is 9.1 µg m -3 , and the affected ratio is 33%.In terms of seasonality, the contribution ratio in the 4 seasons from the highest to the lowest were as following: winter > autumn > spring > summer.The highest one of affected ratio of long-range transport was winter (39%), the last one was summer (10%).The highest contribution from longrange transport was experienced in winter (16.3 µg m -3 and the lowest in summer (1.1 µg m -3 ) as shown in Table 6.These observations in respect to seasons were similar to the results reported by Chen et al. (2014) who noted that northwestern monsoon promotes high long-range transfer of particulate matter from East Asia, while for summer the majority of PM is from local sources in Taiwan.
Fig. 11 shows the spatial distribution of long-range pollution effect to the PM 2.5 levels in Tainan city.It is clear from annual average, that the influence of long-range transport were highest in western area in Tainan City.The highest influence is intersected with Chiayi City (about 9-10 µg m -3 ).The variation of concentration is decreased from coastal to inland.The contribution fraction of long-range transport is about 30-45%.Similar to this study Chuang et al. (2008), the long-range transport influence was greater compared to other sources in Taipei, northern Taiwan.Other studies (Chuang et al., 2008;Fu et al., 2009;Chen et al., 2014) noted, long range transport of pollutants from the continental Asia and Mainland China has undermined the efforts by TWEPA to improve the air quality in Taiwan.

Simulation of the Effects of Primary and Secondary Aerosols on the PM 2.5 in Tainan City
The effect of primary and secondary PM on the PM 2.5 concentrations in Tainan were investigated from 3 kinds of cases: case of non-eastern emission, case of non-emission and case of primary PM emission in Taiwan, respectively.The results of simulation were compared with each other to understand the level of interaction.The relationships used for comparison are as follows: The effect of primary PM in Taiwan = Case of primary PM emission in Taiwan (D) -Case of non-emission (C); The effect of secondary PM in Taiwan = Case of noneastern emission (B) -Case of primary PM emission in Taiwan (D); and The effect of primary and secondary PM in Taiwan = Case of non-eastern emission (B) -Case of non-emission (C).
The results of the simulation for primary and secondary PM are as shown in Table 5 for stations and seasons, respectively.The highest contribution by primary PM emission in Tainan was recorded at Tainan Station (11.7 µg m -3 ) with a corresponding ratio of 39%, while the lowest was at the Xinying Station (8.01 µg m -3 ) with a corresponding ratio of 31%.Overall, the average contribution by primary PM emission in Tainan was 9.6 µg m -3 with a mean ratio of 35%.When comparing the effect of primary PM in terms of season concentrations the order was winter > autumn > spring > summer, but the order of the ratios was summer > spring > autumn > winter.This diffrence is because of the variation of the long-range transport decreased in summer, so the contributed ratio of primary emission for PM 2.5 in Taiwan were consequently increased.Additionally, the PM 2.5 concentration is much lower than other seasons in the summer, so the variation of contribution ratios were higher than other seasons.
As for the contribution of secondary PM shown in Table 5, Tainan station (9.40 µg m -3 ) had the highest amount, while Xinying (8.33 µg m -3 ) had the least.The order of secondary PM effcet in the stations was Tainan > Shanhua > Annan > Xinying, while in terms of contribution ratios, it was Shanhua > Xinying > Annan = Tainan.Comparisons for the seasonal Table 5.The contribution long-range transport and primary and secondary fractions to PM  effcet of secondary PM on PM 2.5 in Tainan was greatest in winter and least in summer.Lower secondary PM in summer can be attributed to scavenging of primary PM and precursors such as NO x , SO x and NH 3 from the atmosphere by precipitation (Li et al., 2016) and higher potential for secondary aerosol formation in winter (Lu et al., 2016).The spatial distribution of the effect of primary and secondary aerosols were as shown in Figs. 12 and 13, respectively, while Fig. 14 indicates the total contribution of both primary and secondary aerosl from Taiwan to PM 2.5 concentrations in Tainan City.Fig. 12 indicates that the emission trend of primary aerosol increases from central Tainan to southern Tainan.The most affected amount is in urban area in Tainan City (about 10-12 µg m -3 ), and the contribution fraction is 32-40% in Tainan City.On the other hand, the trend of secondary aerosol are different in that the contribution effect decreased from central Tainan to the surrounding regions.The affected amount is 6-12 µg m -3 , and the highest affected amount is 10-12 µg m -3 in Tainan City intersected with Kaohsiung City.Recently, Chen et al. (2014) reported that from CMAQ modelling, the primary PM was mostly from local sources in Taiwan, while the secondary PM (53%) is majorly from long range transport.

The Effects of Local Emission Sources and Trans-Boundary Pollution on PM 2.5 Levels in Tainan City
Four cases were simulated in order to evaluate the contribution from local sources and trans-boundary emission from neighboring counties to the PM 2.5 levels in Tainan.The four scenarios were Non-point emission in Tainan, Non-line emission in Tainan, Non-area emission in Tainan, and Emission from Tainan only and domain 4 containing As for the point sources and line sources emissions, the diagram of showing annual average indicates that the influence in southwest area in Tainan was higher due to high concentration of urban areas in this part.The average contribution was about 3.5 µg m -3 and 2.5 µg m -3 , while the fractions in the PM 2.5 were in the range of 9-16% and 9-11%, for point and line sources, respectively.In the case of area sources, the influence increased from the south towards the north parts of Tainan.The contribution amount was higher than 3.75 µg m -3 in the southwestern area, and the fraction of area sources in the PM 2.5 was higher than 15%.As seen in Fig. 18, the influence of trans-boundary emissions on Tainan PM 2.5 was from east to west with higher contribution fractions in the east.This is because that the west is the ocean and the east is where Tainan has boundary with Kaohsiung and Pingtung Counties from where emissions might be coming from.
In terms of spatial distribution the contribution amounts and ratios were presented for different monitoring stations in Table 6.In terms of amounts the highest effect was from trans-boundary emissions followed by area emissions, line emissions and point emissions.The average contributions were 9.71, 5.28, 2.20 and 1.87 µg m -3 for trans-boundary, surface, line and point source emissions, respectively.The influence of point sources in terms of both contribution amounts and ratios was highest at Shanhua due to the surrounding factories and industries, while that of the line sources and surface emissions affected the PM 2.5 more at Tainan, as a result of concentrated road network and residential areas in the city center.Due to its proximity to Chiayi County, Xinying station had more contribution from trans-boundary emissions (over 42%) than other stations, which are located seaward and to the west and far from borders of other counties.Overall, among the local sources, area sources had the greatest influence since it encompasses road dust, livestock and agricultural emissions as well as residential areas.Therefore, to reduce PM 2.5 , more attention shroud be given to area sources.
Table 7 shows the distribution of the influence of point, line, area, and trans-boundary emissions in terms of seasons.In terms of ratios, local sources including point sources (16.4%),line sources (14.0%) and surface emissions (32.6%) had the highest influence in summer compared to other seasons.Conversely, in terms of contribution ratios, summer had the least influence simulated for trans-boundary sources.Overall, trans-boundary emissions were greater than local sources similar to the case in Beijing as shown the study done by Lang et al. (2013), who proposed that in this case a target needs cooperation from neighboring regions in order to reduce PM 2.5 levels in the atmosphere.

CONCLUSIONS
The results of this study indicated that the yearly of average PM 10 and PM 2.5 in Tainan were above the air quality standards of 15 and 35 µg m -3 , respectively, but the long-term analysis show a decreasing trend in the PM 10 and PM 2.5 emissions for each monitoring station.The high correlation values of PM 2.5 show that by using modeling, the exposure estimates can be achieved to evaluate longterm health effects of PM 2.5 in Taiwan and also provide data where there are gaps in in monitoring and measurement activities.The results of MM5-CMAQ simulation indicated that the contribution of PM 2.5 concentration in Tainan includes long-range transport (32.9%), the other counties of Taiwan (34.2%) and local emissions from Tainan (32.9%).Since it is not possible to control long range transport, stricter regulations should be imposed on local pollution emission sources thus creating an allowance for long range pollution contribution.In terms of local sources, the highest influence is from area sources (18.6%),followed by line sources (7.7%) and point sources (6.6%).Thus, to control PM 2.5 in Tainan, focus should be on construction, road dust and residential activities.Additionally, since approximately 32.9% of PM 2.5 in Tainan is contributed trans-boundary emissions from other counties, control strategies should be done such that they are in tandem with the rest of Taiwan and just not in isolated county scenarios.

Fig. 1 .
Fig. 1.Four sites of air quality monitoring in Tainan (edited and adapted from Lu et al. (2016)).

Fig. 2 .
Fig. 2. Four Domains of Nested Grids used in Air Quality Modeling.

Fig. 4 .
Fig. 4. Time series and Box Plot of PM 2.5 annual mean concentration.

Fig. 5 .
Fig. 5. Time series and Box Plot of PM 10 annual mean concentration.

Fig. 8 .
Fig. 8.The monthly and seasonal mean variation of PM 2.5 /PM 10 ratios for each station in Tainan City.

Fig. 9 .
Fig. 9.The spatial distribution of monthly mean concentration in January, April, July and October.

Fig. 10 .
Fig. 10.The scatter chart of observed and simulated daily mean for each pollutant during for months (January, April, July and October) in 2010.
2.5 concentrations for each station in Tainan City (average for 4 months) and each season in 2010.Primary and secondary PM from Taiwan which affected Tainan

Fig. 12 .
Fig. 12.The spatial distribution of primary PM effects from Taiwan on PM 2.5 concentrations for Tainan City in 2010 (for each season and average for 4 months).

Fig. 13 .
Fig. 13.The spatial distribution of secondary PM effects from Taiwan on PM 2.5 concentrations for Tainan City in 2010 (for each season and average for 4 months).

Fig. 14 .
Fig. 14.The spatial distribution of primary and secondary PM effects from Taiwan on PM 2.5 concentrations for Tainan City in 2010 (for each season and average for 4 months).

Table 1 .
The annual mean concentrations of PM 2.5 and PM 10 for the period 2005 to 2014 (µg m -3 ).

Table 3 .
The table of simulated performance of mean for each pollutant during 4 months for each station surrounded Tainan in 2010.The evaluation guidelines for performance are newest version released by TWEPA(TWEPA, 2015).

Table 4 .
The table of simulated performance of monthly mean for each pollutant for each station surrounded Tainan in 2010.The evaluation guidelines for performance are newest version released by TWEPA

Table 6 .
Spatial contributions of local emission sources and trans-boundary pollution on PM 2.5 levels in Tainan City.

Table 7 .
Seasonal contributions of local emission sources and trans-boundary pollution on PM 2.5 levels in Tainan City.