Benefit Analysis of Precursor Emission Reduction on PM2.5: Using CMAQ-RSM to Evaluate Control Strategies in Different Seasons

Received: December 14, 2021
Revised: July 10, 2022
Accepted: July 16, 2022

 Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Cite this article:

Chen, C.R., Lai, H.C., Hsiao, M.C., Ma, H.W. (2022). Benefit Analysis of Precursor Emission Reduction on PM_2.5: Using CMAQ-RSM to Evaluate Control Strategies in Different Seasons. Aerosol Air Qual. Res. 22, 210381. https://doi.org/10.4209/aaqr.210381

HIGHLIGHTS

• Applying RSM model to quantify the impact of precursors emission on PM_2.5.
• Controlling primary PM_2.5 emissions is the most effective to reduce ambient PM_2.5.
• Managing the emission of NH₃ has greater benefits compared with SO_x and NO_x.

Keywords: Fine particulate matter, Air quality management, Emission reduction, Response model

1 INTRODUCTION

PM_2.5 pollution is a major environmental and human health problem, and long-term exposure increases the risk of cardiovascular disease acute and chronic respiratory disease dementia and depressive responses (Laden et al., 2006; Pope III and Dockery, 2006; Wu et al., 2015; Carey et al., 2018; Chu et al., 2019). In 2019, 99% of the world population was living in places where the WHO air quality guidelines levels were not being met (5 µg m^–3 annual mean; 15 µg m^–3 24-hour mean). The combined impact of outdoor and indoor air pollution leads to an estimated eight million premature deaths each year (WHO, 2021).

PM_2.5 is composed of primary aerosol and indirectly formed secondary aerosol. Compared with the direct emission of primary aerosol, secondary aerosol is produced from primordial pollutants through atmospheric chemistry. These atmospheric chemical mechanisms include photochemical reactions, homogeneous gas phase reactions, heterogeneous liquid phase reactions, and gas-solid phase reactions. The precursors mainly involved in the reaction include NO_x, SO_x and NH₃. As the relevant atmospheric chemistry mechanism involves the transformation of primary pollutants, it has a considerably strong relationship with the concentration of precursors and seasonal climatic conditions (Li et al., 2016; Tseng et al., 2016; Tian et al., 2018; Thunis et al., 2021).

In Taiwan, where industry and commerce are developed and the population is dense, there are serious PM_2.5 pollution problems in most areas. It is urgent to implement effective air quality management strategies to reduce the health burden caused by this air pollution. Besides domestic pollutant discharge, the atmospheric environmental conditions and Taiwan's special terrain are also important factors contributing to the poor air quality. The atmospheric environmental structure changes seasonally and produces unique meteorological conditions at different times, resulting in a complex array of causes behind air pollution (Lai and Lin, 2020). Taiwan is located in the subtropical monsoon region of the West Pacific, in which blow the northeast and southwest monsoons with the alternation of the seasons. The northeast monsoon prevails for seven months from October to April of the following year, and the southwest monsoon prevails from mid-June to mid-September, lasting only about three months. The prevailing period of the northeast monsoon is not only longer but also stronger than that of the southwest monsoon. Air pollution in Taiwan is more serious in spring and winter when the northeast monsoon prevails.

It is necessary to provide further systematic solutions to how to estimate the reduction ratio of various emission sources under different meteorological conditions and the improvement benefits after reduction. The traditional approach of reducing emissions from various sources based on emission inventory is usually effective when only primary pollutants are involved. When secondary pollutants are involved, however, the relationship between source reduction and ambient concentrations becomes complicated and causes uncertainty about the effect of the reduction (Chen et al., 2017). The relationship between the generation of secondary pollutants and the discharge of precursors can be greatly affected by the impact of the terrain and climate, and the region where the reduction benefit occurs cannot be determined due to the pollutant transport factor (Hsiao et al., 2021).

To put forward a more effective planning policy, a complex air quality model is often used for scenario analysis, to predict the distribution of pollutant’ spatial-temporal information and simulate the effect of the reduced emissions of air pollutants. For example, Derwent et al. (2009) used a mobile air mass trajectory model to estimate PM concentrations in rural areas and proposed a chemical kinetic description of how the main particles formed. Based on a sensitivity study of a 30% reduction in SO₂, NO_x, NH₃, VOC and CO emissions, it was found that PM_2.5 generation in a rural environment is mainly limited to ammonia, suggesting that policy-makers should consider focusing on the abatement of NH₃ to get the largest PM_2.5 reduction.

In a study by Smith et al. (2016), the CMAQ–Urban and WRF models were used to estimate outdoor air pollutant concentrations and exposure to air pollutants in various environments for different age groups. The mean annual exposure to outdoor sources was estimated to be 37% lower for PM_2.5 and 63% lower for NO₂ than at the residential site. These smaller estimates reflect the effects of reduced exposure indoors, the amount of time spent indoors, and the mode and duration of travel in London.

To better characterize NH₃ emissions in an air quality model simulation, Hsu et al. (2019) applied dynamic NH₃ emissions parameterization to improve the temporal profile of NH₃ emissions from livestock operations, synthetic nitrogen fertilizers, and standing crops. Based on the emissions inventory of Taiwan (TEDS 8.1, 2010), the dynamic NH₃ emissions approach was applied to improve estimation of the diurnal and seasonal variations and reduce the simulation bias. The dynamic approach and CMAQ simulation with reduced NH₃ sewage discharge level suggested that the existing Taiwan emissions inventory may overestimate NH₃ sewage emissions.

To evaluate the effectiveness of air quality management policies and reduce the time and computational cost of multi-scenario simulations, the U.S. Environmental Protection Agency (U.S. EPA) developed the Response Surface Model (RSM), which can be used to assess the air quality conditions in real time by inputting meteorological data and emission source positions. The RSM system can also be used to simulate changes in the concentrations of pollutants such as aerosols and particulate matter. The model uses statistical techniques to efficiently analyze the relationship between emission data and air quality simulation results, and has often been used to study the benefits of emission control measures (Zhu et al., 2015; Long et al., 2016; Hsiao et al., 2021).

Xing et al. (2019a) used the CMAQ-RSM model to determine the monthly concentrations of PM_2.5 and ozone precursors in China and found that the reduction of NO_x emissions reduced PM_2.5 and ozone concentrations simultaneously. The average PM_2.5 concentration decreased by 1.2 µg m^–3 for every 10% reduction in NH₃ emissions. This shows that the RSM system is suitable for simulating a variety of pollution sources to reduce emissions, which helps save simulation time and allows timely data to be provided to optimize pollution control strategies.

At present, most of the relevant studies on air quality models focus on the impact of point source precursor emissions on local areas and there are few studies on the impact of compound emissions on PM_2.5 concentration over a large range. The reduction in the sensitivity of PM_2.5 concentration to precursors also needs to be clarified; however, due to the bottleneck in pollution prevention and control technology, the reduction of NO_x and SO_x has become more difficult and the cost has been increasing in the face of more active control targets. Therefore, it is desirable to seek additional reduction strategies to improve effectiveness and reduce costs (Pinder et al., 2007; Shahzad Baig and Yousaf, 2017; Xing et al., 2019b).

The influence of precursor emissions on the concentration of PM_2.5 is an important basis of relevant management strategies and an effective and rapid assessment tool is needed. This research is based on the Taiwan emission inventory TEDS9, using a mature air quality model to simulate the impact of precursor emission reductions on the ambient PM_2.5 concentration, and also develop a high-resolution air product model, RSM, that can be used to quickly test various scenarios and allow comparisons to be made about the contribution of primary PM_2.5, NO_x, SO_x and NH₃ to provide a scientific basis for formulating effective control strategies.

 
2 METHODS

 
2.1 Mesoscale Weather Model and Air Quality Model

We used Operational Global Analysis data (NCEP) for the weather model and four-dimensional data integration technology (Grid-nudging) for the initial and boundary conditions in the Weather Research and Forecasting Model (WRF) (Srivastava et al., 2015; Xing et al., 2019a). There are four domains in this study. Domain1 used the ASCII file of vertical Profiles of CMAQ. From domain2 to domain4, the previous domain concentration value is used as the initial value. For example, the initial data of domain4 is based on domian3. This was suitable for high-resolution numerical forecasts of different terrain and climate processes that can improve the resolution and accuracy of weather forecasts.

The emission data used in the air quality model simulation was divided into Taiwan and East Asia. The Taiwan Emission Data System (TEDS9) with the base year 2013 was used for the former, and the MIX Asian emission inventory was used for East Asia. The emission inventory contains MEIC (China), PK-NH₃ (China's ammonia emission Inventory), CAPSS (South Korea), ANL-India (India), REAS2 (Japan) for Asian regional emission inventories as input data. The MIX anthropogenic source emissions inventory includes ten main atmospheric chemical components : SO₂, NO_x, CO, NH₃, NMVOC, PM₁₀, PM_2.5, BC, OC and CO₂. The chemical mechanisms are divided into CB05 and SAPRC-99 (Zhang et al., 2012; Li et al., 2017), providing grid data for five emission sources: power industry, industry, people's livelihood, transportation source, and agriculture, which can provide air quality model requirements at multiple scales.

This study uses the Models-3/CMAQ version 5.1 air quality model developed by the U.S. EPA. The CMAQ universal multi-scale air quality model is the core of the third-generation air quality model, Models-3. It integrates the emission, chemical changes and transmission of air pollutants. The process involves a numerical simulation based on observational results and theoretical principles. The complex solution process requires a huge amount of computing power, but it can simulate the reaction of multiple pollutants in the atmosphere at the same time. Therefore, the results of the CMAQ simulation can fully express the contribution of precursor emissions to PM_2.5 formation.

This study conducted simulations with four-level nested domains, as shown in Fig. 1(a). The simulation settings used four layers of grids, from the first layer (81 kilometers) to the fourth layer (3 kilometers). The resolution of the evaluation model was 3 km, and the fourth layer contained the entire island of Taiwan. The number of grids was 90 × 135 = 12,150 and the simulation period had the base year of 2013. January, April, July and October represented the four seasons, and the average of the four months was set as the annual average concentration. Other research related to Taiwan, such as Hsu et al. (2019), Chen et al. (2019) and Lai et al. (2019) also used the same settings. In terms of Taiwan’s topographic characteristics and the advantages of WRF mesoscale weather system calculations, it is sufficient and reasonable to use 3 km × 3 km grid points to simulate Taiwan’s air quality.

 
Fig. 1. Model simulation field (a) Configuration of the four-level nesting domains and (b) Air Basins in Taiwan.

 
2.2 Response Surface Model and Software of Model Attainment Test

CMAQ can generate results for the ozone, suspended particulates and dry deposition in one simulation; however, modifying related parameters such as the emission ratio for repeated simulations is time-consuming, which causes difficulties for decision-makers when evaluating the simulation results using the air quality model. To meet policy maker demands for emission analysis, the US-EPA developed an air quality instant response system so that the Response Surface Model (RSM) could be used to simulate the climate field model and air quality model via simulations of various emission changes. The RSM statistical formula is used to establish the air quality results of various emission changes, and the calculation errors of RSM and CMAQ simulations can be analyzed using the system validation formula to ensure the accuracy of the RSM data (Ashok et al., 2013). In this study, WRF and CMAQ were firstly used to conduct a large number of scenario simulations to build the RSM database. After the pollutant concentrations of several scenarios were simulated using the RSM system, the Software for the Model Attainment Test (SMAT) was used to calibrate the RSM results with the observed values to enhance their consistency (Long et al., 2016; Li et al., 2019). SMAT used air quality models of baseline and pollution reduction scenarios to construct relative response factors (RRFs), and then adjusted the design values at baseline by applying Voronoi neighbor averaging to the RRFs.

The RSM statistical formula calculates the air quality results of each pollution concentration using the high-dimensional Kriging interpolation method, and a set of correlation regression curve formulae found as Eq. (1):

represents the RSM prediction; f is the d × 1-dimensional vector of the y regression function;  is the unknown d × 1-dimensional regression coefficient; F is the n × d-dimensional matrix of the sample regression function; and z(x) is the covariance of a Gaussian random process. For further derivation of z(x),  is the n × 1-dimensional correlation coefficient vector of  and the variational functions  and Yⁿ are process simulations of Y_x0 and y_n, respectively.

Since the circulation of air pollution in Taiwan crosses county and city administrative division lines and also has obvious regional characteristics and seasonal cycles (Yu and Chang, 2001), the Taiwan-EPA has divided Taiwan into seven Air quality zones (Air Basins) based on topography, climate, wind direction, and pollution dispersion. According to TEDS9, the baseline emissions (unit: t year^–1) of primary PM_2.5, SO_x and NO_x in each air basin: the Northern:13 934, 21 556 and 90 159; Chu-Miao: 5 457, 3 321 and 28 288; the Central: 14 941, 26 356 and 82 715; Yun-Chia-Nan: 11 968, 17 495 and 64 344; Kao-Ping: 14 479, 38 429 and 85 034; Yilan: 3 568, 1 172 and 12 920; Hua-Tung: 9 770, 4 911 and 22 637. The emission sources in Taiwan are mainly located in the west (Hsu et al., 2019), and so the monitoring values of five major air basins in the west of Taiwan were used for performance comparison (Fig. 1(b)). From north to south they are: the Northern Region, Chu-Miao, the Central Region, Yun-Chia-Nan and Kao-Ping.

Daily mean values obtained from monitoring stations and simulated values were used for statistical PM_2.5 quantitative analysis, which included three quantitative indicators: the Mean Fractional Bias (MFB) being within 35%, the Mean Fractional Error (MFE) of the matching values being less than 55% and the correlation coefficient (R) being greater than 0.5.

The observational data were analyzed using the automatic station data from the Taiwan-EPA monitoring website with regard to the following four quantitative indicators: the Maximum peak normalized Bias (MB) being within ± 10%, the Mean Normalized Bias (MNB) being within ± 15%, the Mean Normalized Error (MNE) of paired values being within 35% and the correlation coefficient being greater than 0.5.

 
3 RESULTS AND DISCUSSION

In order to evaluate the benefits of emission reduction on PM_2.5 and the changes across seasons and regions, we used CMAQ to establish the RSM, and set a reduction scenario based on the year 2013 to reduce emissions by 10–100% for each species.

The RSM model was used to calculate the reduced concentration of PM_2.5 in the perimeter under the reduction of primary PM_2.5, NO_x, SO_x, and NH₃ emissions, and then to estimate the impact of emission reduction on PM_2.5 concentration per ton. The impact of the temporal and spatial changes of these reduction benefits on the formulation of policies was also discussed. According to the Taiwan EPA, the main source of SO_x is SO₂ from the sulfur-containing fuels, mainly from the burning of industrial fuels (77%) and transport except for the road (19%); the main source of NO_x is NO + NO₂ when fuel is burned at high temperature, which mainly comes from vehicles (51%), industry (36%) and transport except for the road (8%); NH₃ mainly comes from animal husbandry (38%), sewage treatment (35%) and about 10% from biological sources (TEDS9.0, 2015; Taiwan EPA, 2021).

The simulation results show that the maximum tonnage reduction of primary PM_2.5 was 74 117 t year^–1 and the maximum improvement in the ambient PM_2.5 annual average concentration was 6.99 µg m^–3. The maximum reduction of NO_x tonnage was 386 097 t year^–1 and the maximum improvement in the ambient PM_2.5 annual average concentration was 3.13 µg m^–3. The maximum tonnage reduction of SO_x was 113 240 t year^-1 and the maximum annual improvement in the ambient PM_2.5 concentration was 0.70 µg m^–3. The maximum reduction of NH₃ tonnage was 187 721 t year^–1 and the maximum improvement in the ambient PM_2.5 annual average concentration was 3.36 µg m^–3. In terms of the benefit of emission reductions per ton, the highest was 9.79 ´ 10^–5 µg m^–3 t of primary PM_2.5, followed by 1.83 ´ 10^–5 µg m^–3 t of NH₃, 8.84 ´ 10^–6 µg m^–3 t of NO_x and 7.73 ´ 10^–6 µg m^–3 t of SO_x.

 
3.1 Model’s Simulation Verification

A comparison between the simulated and observed values in the base year showed that the MFB, MFE and correlation coefficient were respectively –29%, 34% and R = 0.7, which were all in line with the model simulation specifications. The coincidence rate of the MFB, MFE and correlation coefficient conforming to their specified values in the simulation area were respectively 64%, 94% and 99%.

Table 1 presents the simulation results of the PM_2.5, NO₂ and SO₂ air quality models for the five air basins in western Taiwan and the simulation results of the Taiwan-EPA monitoring stations. In 2013, the average MFB of PM_2.5 in each air basin ranged from –36.4% to –23.0%. The month that best met the simulation specification was in the Northern Region (86%), followed by the Central Region (82%). The worst compliance rate was in Yun-Chia-Nan, with a compliance rate of only 18%. The MFE of PM_2.5 was between 29.0% and 40.4%, the coincidence rate of the Northern region, Chu-Miao and Yun-Chia-Nan all reached 100%, and the Central region was the lowest with 91%. For PM_2.5, the total average MFB in 2013 was –28.9%, MFE was 33.9% and the correlation coefficient was 0.71.

The average MFB of NO₂ in each air basin ranged from –11.7% to –1.5%, and the compliance rate of the Northern region, Chu-Miao, Central region, Yun-Chia-Nan and Kao-Ping was 100%, which is in line with the simulation specification. The MFE was between 7.5% to 14.8%. The coincidence rates of the Northern region, Chu-Miao, Central region, Yun-Chia-Nan and Kao-Ping all reached 100%. For NO₂, the total average of MFB in 2013 was -8.8%, MFE was 13.2% and the correlation coefficient was 0.90.

The average MFB of SO₂ in each air basin ranged from –23.3% to –8.7%, the compliance rate of the Northern region, Chu-Miao, Central region and Yun-Chia-Nan was 100%, and of the Kao-Ping it was 92%, which is in line with the simulation specification. The MFE was between 21.5% to 40.3%. The coincidence rates of the Northern region, Chu-Miao, Central region and Yun-Chia-Nan all reached 100%, and that of the Kao-Ping was 92%. For SO_2, the total average MFB in 2013 was –8.8%, MFE was 13.2% and the correlation coefficient was 0.90. The verification parameters of the above species are within the specification.

 
3.2 Seasonal Variation of Concentration Decline

We simulated a series of reduced emission scenarios to analyze the response of PM_2.5 concentration to the reduction of NH₃, NO_x and SO_x. Before analyzing the impact of emission changes on concentration, it is necessary to understand the baseline concentration distribution in the field. As shown in Table 2, the annual average of the simulated base year was 23.3 µg m^–3, the maximum value was 29.64 µg m^–3 in spring, and the minimum value was 11.79 µg m^–3 in summer. The high concentration areas were located in the southwest regions and coastal areas.

The simulated precursor emission reduction scenario results ranged from 10% to 100%, and the highest tonnage emission reductions were primary PM_2.5 74 117 t year^–1, NO_x 386 097 t year^–1, SO_x 113 240 t year^–1 and NH₃ 187 721 t year^–1. The simulation results show that when the emission reduction was 100%, the seasonal average contributions of primary PM_2.5, NO_x, SO_x, and NH₃ were 5.87–8.02 µg m^–3, 0.58–5.07 µg m^–3, 0.49–0.96 µg m^–3 and 0.58–6.09 µg m^–3 respectively. The decreasing concentrations of NO_x and NH₃ were significantly different across seasons, whereas the trend for SO_x was not.

The seasonal variation of the PM_2.5 Baseline concentration shows that the high annual average value is mainly due to the contribution of the spring and winter conditions. Summer is the time of year when PM_2.5 levels are lowest. The contribution rate of primary PM_2.5 is maintained at a high level in all seasons, and even reaches 53% in summer; however, the actual contribution of primary PM_2.5 is between 5.87–8.02 µg m^–3 and does not change significantly across the four seasons because good diffusion conditions and low relative humidity in summer are not conducive to the generation of secondary PM_2.5 (Cheng et al., 2015; Viatte et al., 2021), resulting in lower PM_2.5 concentration in summer.

The seasonal spatial distribution of concentration reduction can be observed in Fig. 2. In spring, the concentration reduction is evenly distributed among the five air basins in western Taiwan, and in summer, the reduction benefit mainly occurs in the central land and northern coastal areas. In autumn and winter, the reduction benefits are mainly distributed in the southwest. According to the emission inventory, the emission sources of primary PM_2.5 and precursors are mainly distributed in the west of Taiwan, but the simulation results show that the areas with high reduction benefits will occur downwind with seasonal changes. This means that the generation and transmission of air pollution will be affected by the monsoons (Chen et al., 2021), that is, high emissions do not necessarily represent high concentrations, and the regions producing high concentrations will vary with seasonal changes. We also observe that the location of the reduction benefit of NH₃ is farther from the emission source, with many high-efficiency areas extending to the sea. The reduction benefit of primary PM_2.5 is mainly located in urban areas on land, which highlights the diffusion characteristics of secondary PM_2.5. This shows that regional emission reductions do not necessarily produce equivalent reduction benefits locally (Hsiao et al., 2021).

Fig. 2. Seasonal variation of decreased PM_2.5 concentration distribution caused by 100% reduction in primary PM_2.5 and NH₃.

Comparing the decreased concentrations of each season (Fig. 3(a)), the reduction benefit of primary PM_2.5 in summer and autumn is lower than that in spring and winter. In reality, however, the emission of primary PM_2.5 does not show obvious seasonal variation, which reflects the better tropospheric diffusion conditions in summer and autumn. Opposite in winter, because the mixed layer is low, and often accompanied by the inversion layer, the diffusion conditions of PM pollutants deteriorate (Tian et al., 2018; Xu et al., 2019; Nidzgorska-Lencewicz and Czarnecka, 2020). The decreased PM_2.5 concentrations of NO_x and NH₃ are quite close to the seasonal trend, and the reduction rate of NH₃ is even better than NO_x in spring and winter. The annual average reduction rate of SO_x is the lowest, but it is worth noting that the reduction rate of SO_x is better than that of NO_x and NH₃ in summer.

Fig. 3. Benefit of decreased PM_2.5 for each season with 100% emission reduction. (a) Decreased PM_2.5 concentration in each season with 100% emission reduction and (b) Proportion of the decreased PM_2.5 concentration in each season with 100% emission reduction.

 
3.3 Effects of Precursor Control on Reducing PM_2.5

From the perspective of the variation in the resulting decreased PM_2.5 concentration, it can be observed from Table 3 and Fig. 4 that precursor emission and decreased PM_2.5 concentration show a non-linear positively correlated response. The unit benefit (decreased concentration/ reduction in tonnage) of PM_2.5 decreases slightly with the increase of the emission reduction rate, among which SO_x has the most obvious non-linear response. There was no significant difference in unit efficiency between species in different scenarios, but the change was obvious between seasons. With the increase of the emission reduction ratio, the difference of the individual decreased concentration also became larger. Regarding the annual average value, the reduction of primary PM_2.5 had the greatest benefit, followed by NH₃ and NO_x, and the lowest benefit was from reduced SO_x. When the emission reduction rate was 100%, the annual average decreased concentration of primary PM_2.5 reached 6.99 µg m^–3, which is about ten times that of SO_x, the maximum annual decreased concentration of NH₃ was up to 3.36 µg m^–3, which is about five times that of SO_x, and the annual average decreased concentration of NO_x was up to 3.14 µg m^–3, about 4.5 times that of SO_x.

The precursors show different trends in the proportion of the reduced concentration in each season (Fig. 3(b) and Fig. 4). Primary PM_2.5 is still the main species in summer, though generally having a low level of PM_2.5, but the proportion of the reduced concentration of SO_x at that time reached 8.35%, which exceeded NO_x and NH₃. In the spring and winter, generally having a high level of PM_2.5, the proportion of the reduced concentration of NO_x and NH₃ increased sharply.

Fig. 4. Proportion of decreased PM_2.5 concentration caused by different emission reduction scenarios for each season.

The reduction benefit of SO_x had a trend roughly opposite to those of NO_x and NH₃. In summer, the proportional benefit of sulfate reduction reached a peak, whereas nitrate showed a relatively higher proportional benefit in spring and winter. It is speculated that the saturated vapor pressure of HNO₃ is high and so most nitrate exists in gaseous form, such as nitric acid, at high temperatures. At low temperatures, most nitrate exists as particles, such as ammonium nitrate. Low temperature and high relative humidity increase particulate NH₄NO₃ concentration and decrease gaseous HNO₃ and HN₃ concentrations, and so there is often abundant content of nitrate in aerosols in winter (Hueglin et al., 2005; Qian et al., 2013; Shimadera et al., 2014).

We divided the reduced concentration of each scenario by the number of tonnages that needed to be reduced to obtain the impact (benefit) per ton of reduction as shown in Fig. 5. The reduction benefit of primary PM_2.5 was between 9.43–9.79 × 10^–5 µg m^–3 t^–1, of NO_x it was between 8.12–8.84 × 10^–6 µg m^–3 t^–1, of SO_x it was between 6.15–7.45 × 10^–6 µg m^–3 t^–1 and of NH₃ it was between 1.78–1.83 × 10^–5 µg m^–3 t^–1. The reduction benefits per ton of NO_x and SO_x were quite similar.

Fig. 5. Decreased PM_2.5 concentration caused by emission source reduction per ton (µg m^–3 t^–1).

The reduction benefit of primary PM_2.5 was the most significant, being more than 11 times that of NO_x, whereas the reduction benefit per ton of NH₃ was only about twice that of NO_x and SO_x. The simulation results show that PM_2.5 concentration was highly sensitive to Primary PM_2.5 and NH₃, as well as that the reduction benefit of NH₃ was superior to that of NO_x and SO_x.

 
4 CONCLUSION

We provided a method to quantify the benefits of emission reduction on PM_2.5 concentration, and the model results illustrated concentration differences across time and space, providing a basis for future air quality management strategies. Assuming equal proportion reduction of all emission sources, the reduction benefit of each precursor has different trends with the seasons. The simulation results also showed that PM_2.5 concentration is highly sensitive to primary PM_2.5 and NH₃. In the past, emission pipelines have been mainly regulated by laws and regulations based on the reduction of NO_x and SO_x, which has been effective for a long period of time; however, both technology and costs have run into bottlenecks in recent years, and it has been found that controlling primary PM_2.5 and NH₃ is more effective at present. By using CMAQ-RSM, we can clarify the reduction benefits of precursors in each season, which helps to quickly evaluate the effects of PM_2.5 reduction measures and adjust control strategies.

The follow-up application of this research model could be its development into a decision support system for quantitative targets. When air quality targets are drawn up for different regions, the corresponding reductions of each precursor could be obtained through RSM calculations, which would be conducive to flexible control of different pollutants in different seasons. Strategies such as the drafting of primary PM_2.5 concentration specifications for emission pipelines, and focusing on NH₃ emission reduction in spring and winter would thus have more significant effects. Therefore, when formulating long-term control strategies, regional or cross-administrative reduction targets should be taken into consideration.

 
ACKNOWLEDGMENTS

The authors gratefully acknowledge the Taiwan-EPA for providing data for the modeling.