Considering Condensable Particulate Matter Emissions Improves the Accuracy of Air Quality Modeling for Environmental Impact Assessment

: This study examines environmental impact assessment considering ﬁlterable particulate matter (FPM) and condensable particulate matter (CPM) to improve the accuracy of the air quality model. Air pollutants and meteorological data were acquired from Korea’s national monitoring station near a residential development area in the target district and background site. Seasonal emissions of PM 2.5, including CPM, were estimated using the California puff (CALPUFF) model, based on Korea’s national emissions inventory. These results were compared with the traditional environmental impact assessment results. For the residential development area, the seasonal PM 2.5 concentration was predicted by considering FPM and CPM emissions in the target area as well as the surrounding areas. In winter and spring, air quality standards were not breached because only FPM was considered. However, when CPM was included in the analysis, the results exceeded the air quality standards. Furthermore, it was predicted that air quality standards would not be breached in summer and autumn, even when CPM is included. In other words, conducting an environmental impact assessment on air pollution including CPM affects the ﬁnal environmental decision. Therefore, it is concluded that PM 2.5 should include CPM for greater accuracy of the CALPUFF model for environmental impact assessment.

PM can be classified into primary PM (directly emitted from the source) and secondary PM (formed by photochemical reactions in the atmosphere after being emitted in the gaseous phase) [12][13][14]. In Korea, the national emission inventory estimates the annual emissions of major air pollutants, including carbon monoxide (CO), nitric oxide (NO x ), sulfur oxide (SO x ), total suspended particles (TSP), particulate matter less than 10 µm (PM 10 ), particulate matter less than 2.5 µm (PM 2.5 ), black carbon, and volatile organic Sustainability 2021, 13, 4470 2 of 10 compounds (VOCs) based on emission sources and regions [15,16]. The emission inventory for PM (TSP, PM 10 , and PM 2.5 ) considers filterable PM (FPM), which is collected through a filter. However, unlike ambient air, the primary PM emitted from emission sources can be classified into FPM and condensable PM (CPM), and the aggregate of FPM and CPM is considered as the total PM [17]. CPM is in the gas phase under high-temperature conditions and condenses into PM immediately after emitting from the emission source. Corio and Sherwell [18] estimated that CPM accounts for about 76% of total PM10 emission from large stationary emission sources. Although the US EPA recognized the CPM issue in the early 1980s and developed a measurement method for stationary sources, it was not considered as a severe issue [19,20]. Sulfur trioxide (SO 3 ) in flue gas can react with water vapor to form sulfuric acid mist, which has often been misunderstood as CPM formation [21]. In recent years, not only studies on CPM emitted from stationary sources [22][23][24] but also studies using CPM as modeling inputs have been conducted [25]. In particular, Morino et al. [25] revealed the contribution of CPM to ambient PM by measuring CPM at stationary sources and confirmed the improvement in the prediction model for ambient PM by including CPM as a model input. Thus, it is necessary to include CPM in the air quality modeling of environmental impact assessments.
In atmospheric studies, AERMOD (the AMS/EPA regulatory model) is one of the most widely used models for environmental impact assessment [26,27]. It is a steady-state plume model that calculates atmospheric diffusion, based on the concept of turbulent structure and scaling in the atmospheric boundary layer. Moreover, it can be used in simple (planar) district scenarios and complex terrain scenarios [28]. The CALPUFF (California puff) model is another model that is used in the evaluation of large atmospheric emission sources, such as industrial complexes, power plants, and incinerators [26,29]. CALPUFF is a multi-layered, multi-stage, unsteady puff diffusion model that simulates the effects of temporally and spatially varying weather conditions on the transport of pollutants [30]. Furthermore, it can be applied to rough and complex terrains. Although the CALPUFF model is useful, it does not consider photochemical reactions and chemical reactions of secondary pollutants. Therefore, chemical transport models, such as CMAQ [31], which consider atmospheric chemical reactions, have been recently used in atmospheric environmental impact assessments.
To improve the CALPUFF model, this study considered CPM in the emission inventory and applied it to the seasonal environmental impact assessment of the target district.

Materials and Methods
In this study, the 2013 emission inventory provided by the National Center for Fine Dust Information [32] was used as input data for the CALPUFF model. In addition, total suspended particles (TSPs) and PM 2.5 (particulate matter with an aerodynamic diameter less than 2.5 µm) were selected as the target air pollutants. Moreover, CPM emission factors for stationary sources were used, as published by the National Institute of Environmental Research [33] (Table 1). Liquified natural gas (LNG), diesel, and B-C oil were measured in a boiler without a control device. Furthermore, the bituminous coal emission factor was measured at the end of the control devices in a power plant facility. The concentration of CPM emissions was calculated by multiplying the FPM to CPM emission factor ratio with PM 2.5 , obtained from Korea's emission inventory data.  Figure 1 shows a schematic of the CALPUFF model. The CALPUFF model consists of a CALMET module, a meteorological model, and a CALPUFF module, an air pollution model. The CALMET module uses a land cover map, meteorological data, and aerological data as input data to calculate meteorological model results and then inputs the emission inventory to derive the CALPUFF model results. The control file serves to input commands to control each module. To simulate the FPM and CPM emission behavior, a software was developed to calculate the amount of PM emissions from major emission sources in the target district and link this with the meteorological data acquired from the automatic weather system data from the four monitoring stations near the target area. Thus, this software was used to establish a methodology to verify the accuracy of the concentrations of FPM and CPM emissions and understand atmospheric behavior prediction through case studies.
model. The CALMET module uses a land cover map, meteorological data, and aer data as input data to calculate meteorological model results and then inputs the e inventory to derive the CALPUFF model results. The control file serves to inp mands to control each module. To simulate the FPM and CPM emission behavio ware was developed to calculate the amount of PM emissions from major e sources in the target district and link this with the meteorological data acquired f automatic weather system data from the four monitoring stations near the targ Thus, this software was used to establish a methodology to verify the accuracy of centrations of FPM and CPM emissions and understand atmospheric behavior pr through case studies.
The target area was the Bugok residential development district in G Gyeonggi-do, with a project area of 470,000 m 2 (residential area of 200,795 m 2 , com and business area of 6104 m 2 , and public area of 263,101 m 2 ), accommodating a mately 9300 people. To compare and verify the accuracy of the PM concentra model calculations, real-time data provided by the Korea Environment Corp (KEC) monitoring station in and around the target district were used.

Results and Discussion
The CPM conversion factor management program developed in this study p a function for searching and managing conversion factors by fuel type. Additional emissions were calculated by applying the CPM conversion factor for each fuel t line, and surface emission sources. In the CPM emission factor management pro each fuel type, information on the FPM and CPM emission factors of PM2.5 e sources, including construction, buildings, and vehicles, was provided and manag ing the development of the residential area. Thus, it was meticulously configured details on emission factors for each construction instrument, vehicle type, and b

Results and Discussion
The CPM conversion factor management program developed in this study provided a function for searching and managing conversion factors by fuel type. Additionally, CPM emissions were calculated by applying the CPM conversion factor for each fuel to point, line, and surface emission sources. In the CPM emission factor management program of each fuel type, information on the FPM and CPM emission factors of PM 2.5 emission sources, including construction, buildings, and vehicles, was provided and managed during the development of the residential area. Thus, it was meticulously configured so that details on emission factors for each construction instrument, vehicle type, and building energy fuel could be searched. Seasonal variation is one of the most important parameters for environmental impact assessment, so we applied one month of each season [34,35].

Seasonal Results for Environmental Impact Assessment (Winter, January)
Figure 2a-d show the modeling results (5.5 × 5.5 km) by segregating the FPM and CPM emissions generated during the operational stage of the Bugok residential development district based on the area to be analyzed and on whether the emission sources outside the target district were considered. The FPM and CPM emissions were calculated by considering the emission sources in the Bugok residential development district. The FPM concentration was estimated to be 0.0038 µg/m 3 , and the CPM concentration was predicted to be 0.27 µg/m 3 . Meanwhile, PM 2.5 was analyzed by considering emission sources that were outside the target district and excluding the emission sources in the Bugok residential development district. The FPM concentration was predicted to be 10.63 µg/m 3 , and the CPM concentration was estimated to be 14.82 µg/m 3 . Therefore, the TPM (CPM + FPM) concentrations by emission sources inside and outside the target district were calculated to be 0.27 and 25.45 µg/m 3 , respectively.
The modeling results were compared with PM2.5 data acquired from the air pollutant monitoring stations near the target district. Figure 2e shows the results of this study, monthly averages of the monitoring stations, and traditional environmental impact assessment results. In terms of traditional environmental impact assessment, the PM2.5 concentration at the study site was estimated to be 47.28 µg/m 3 . This was the sum of the annual average PM concentration inside the target district and the PM2.5 concentration data acquired from the monitoring station outside the target district. The modeling result of this study was 58.87 µg/m 3 , including the long-range transboundary emissions (33.15 µg/m 3 ), the FPM concentration inside the target district (0.0038 µg/m 3 ), the FPM concentration outside the target district (10.63 µg/m 3 ), the CPM concentration inside the target district (0.27 µg/m 3 ), and the CPM concentration outside the target district (14.82 µg/m 3 ). In particular, the average concentration of TPM, the sum of FPM and CPM, in the target district was 25.72 µg/m 3 , which was approximately two times lower than the concentration measured in winter at the monitoring station (58.87 µg/m 3 ).  Table 2 shows the PM2.5 concentrations in winter in Deokjeok, Seogwipo, and Seosan, which are the national background concentration monitoring stations located in the far The modeling results were compared with PM 2.5 data acquired from the air pollutant monitoring stations near the target district. Figure 2e shows the results of this study, monthly averages of the monitoring stations, and traditional environmental impact assessment results. In terms of traditional environmental impact assessment, the PM 2.5 concentration at the study site was estimated to be 47.28 µg/m 3 . This was the sum of the annual average PM concentration inside the target district and the PM 2.5 concentration data acquired from the monitoring station outside the target district. The modeling result of this study was 58.87 µg/m 3 , including the long-range transboundary emissions (33.15 µg/m 3 ), the FPM concentration inside the target district (0.0038 µg/m 3 ), the FPM concentration outside the target district (10.63 µg/m 3 ), the CPM concentration inside the target district (0.27 µg/m 3 ), and the CPM concentration outside the target district (14.82 µg/m 3 ). In particular, the average concentration of TPM, the sum of FPM and CPM, in the target district was 25.72 µg/m 3 , which was approximately two times lower than the concentration measured in winter at the monitoring station (58.87 µg/m 3 ). Table 2 shows the PM 2.5 concentrations in winter in Deokjeok, Seogwipo, and Seosan, which are the national background concentration monitoring stations located in the far west and south side of Korea. As they are located in the far west and south side of Korea, data collected from monitoring stations are used for evaluation of air pollutants' longrange transportation from polluted areas [36,37]. The observed PM concentrations at the Deokjeok, Seogwipo, and Seosan stations were 31.55, 41.77, and 47.26 µg/m 3 , respectively. The difference between the modeling results and the observed values estimated in this study was considered to be due to the effect of the long-range transboundary emissions from other countries.        Table 2 shows the PM concentrations in winter in Deokjeok, Seogwipo, and Seosan. The observed PM concentrations at the Deokjeok, Seogwipo, and Seosan stations were 43.73 µg/m 3 , 58.67 µg/m 3 , and 44.53 µg/m 3 , respectively. The difference between the modeling results and the observed values in this study was assumed to be the effect of the long-range transboundary emissions from other countries. uated as 0.0002 µg/m 3 and 0.04 µg/m 3 , respectively. By estimating the PM concentration outside the target district, the FPM concentration was predicted to be 8.69 µg/m 3 , and the CPM concentration was 7.67 µg/m 3 . The concentration of TPM from emission sources inside and outside the target district was 0.04 µg/m 3 and 16.36 µg/m 3 , respectively. The TPM concentration around the target district was 16.40 µg/m 3 . Figure 3e shows the predicted results of this study, spring measurements of the monitoring station, and traditional environmental impact assessment results. The modeling result of this study, including the CPM concentration, was 28.65 µg/m 3 , which included the long-range transboundary emissions (12.24 µg/m 3 ), the FPM concentration inside the target district (0.0002 µg/m 3 ), the FPM concentration outside the target district (8.69 µg/m 3 ), the CPM concentration inside the target district (0.04 µg/m 3 ), and the CPM concentration outside the target district (7.67 µg/m 3 ). The concentration of TPM in the target district was estimated to be 16.40 µg/m 3 , which was 1.7 times lower than the average concentration measured at the monitoring station in summer (28.65 µg/m 3 ). Differences in the predicted summer PM concentrations were considered to be due to the effect of the long-range transboundary emissions. Moreover, the observed PM concentrations at the three stations of Deokjeokdo, Seogwipo, and Seosan were 23.45 µg/m 3 , 34.58 µg/m 3 , and 23.79 µg/m 3 , respectively (Table 2).   Figure 3e shows the predicted results of this study, spring measurements of the monitoring station, and traditional environmental impact assessment results. The modeling result of this study, including the CPM concentration, was 28.65 µg/m 3 , which included the long-range transboundary emissions (12.24 µg/m 3 ), the FPM concentration inside the target district (0.0002 µg/m 3 ), the FPM concentration outside the target district (8.69 µg/m 3 ), the CPM concentration inside the target district (0.04 µg/m 3 ), and the CPM concentration outside the target district (7.67 µg/m 3 ). The concentration of TPM in the target district was estimated to be 16.40 µg/m 3 , which was 1.7 times lower than the average concentration measured at the monitoring station in summer (28.65 µg/m 3 ).

Seasonal Results for Environmental Impact Assessment (Autumn, October)
Differences in the predicted summer PM concentrations were considered to be due to the effect of the long-range transboundary emissions. Moreover, the observed PM concentrations at the three stations of Deokjeokdo, Seogwipo, and Seosan were 23.45 µg/m 3 , 34.58 µg/m 3 , and 23.79 µg/m 3 , respectively (Table 2). FPM and CPM concentrations in the study site were predicted to be 21.12 µg/m 3 , whic was 1.7 times lower than the average concentration measured in the autumn at the mon toring station (36.06 µg/m 3 ). The difference in the predicted autumn PM concentratio was considered to be due to the long-distance migration. Moreover, it was confirmed th the PM2.5 concentrations at Deokjeokdo, Seogwipo, and Seosan were 30.45 µg/m 3 , 28.1 µg/m 3 , and 29.53 µg/m 3 , respectively.  (Table 2).

Seasonal Results for Environmental Impact Assessment (Autumn, October)
We included CPM and FPM in the PM prediction model which showed a less tha 5% difference compared to the monitoring station data, while the results of tradition environmental impact assessment showed a difference of 20-40% compared to the mon toring station data. Ghim et al. [38] performed an evaluation of a PM2.5 prediction mod including only FPM emissions, and the predicted value of PM2.5 was found to be 69% o the monitoring station data. Thus, including CPM emissions in the PM prediction mod is one of the ways to increase the accuracy of the model for environmental impact asses ment. Even with the results of measuring CPM and FPM at the large stationary emissio source, the portion of CPM occupies more than 80% [39], which disproves that the PM concentration should be predicted using both CPM and FPM emissions. In addition, th US EPA recommends that the interim guidance for new source review permit program should include the CPM in determining a new major stationary source permission [40 Thus, environmental impact assessment should consider CPM as one of the factors of a quality analysis.  Figure 5e shows the predicted results of this study, spring measurements of the monitoring station, and traditional environmental impact assessment results. The predicted value of this study, including the CPM concentration was 36.06 µg/m 3 , which included the long-distance inflow (14.95 µg/m 3 ), the in-pipe FPM concentration (0.0017 µg/m 3 ), the outside FPM concentration (11.29 µg/m 3 ), the in-pipe CPM concentration (0.20 µg/m 3 ), and the CPM concentration outside the building (9.63 µg/m 3 ). In particular, the average FPM and CPM concentrations in the study site were predicted to be 21.12 µg/m 3 , which was 1.7 times lower than the average concentration measured in the autumn at the monitoring station (36.06 µg/m 3 ). The difference in the predicted autumn PM concentration was considered to be due to the long-distance migration. Moreover, it was confirmed that the PM 2.5 concentrations at Deokjeokdo, Seogwipo, and Seosan were 30.45 µg/m 3 , 28.19 µg/m 3 , and 29.53 µg/m 3 , respectively.
Differences in the predicted autumn PM concentrations were considered to be due to the effect of the long-range transboundary emissions. The observed PM concentrations in the three stations recorded from Deokjeokdo, Seogwipo, and Seosan were 30.45 µg/m 3 , 28.19 µg/m 3 , and 29.53 µg/m 3 , respectively (Table 2).
We included CPM and FPM in the PM prediction model which showed a less than 5% difference compared to the monitoring station data, while the results of traditional environmental impact assessment showed a difference of 20-40% compared to the monitoring station data. Ghim et al. [38] performed an evaluation of a PM2.5 prediction model including only FPM emissions, and the predicted value of PM2.5 was found to be 69% of the monitoring station data. Thus, including CPM emissions in the PM prediction model is one of the ways to increase the accuracy of the model for environmental impact assessment. Even with the results of measuring CPM and FPM at the large stationary emission source, the portion of CPM occupies more than 80% [39], which disproves that the PM concentration should be predicted using both CPM and FPM emissions. In addition, the US EPA recommends that the interim guidance for new source review permit programs should include the CPM in determining a new major stationary source permission [40]. Thus, environmental impact assessment should consider CPM as one of the factors of air quality analysis.

Conclusions
In this study, the PM concentration in the atmosphere was predicted by including CPM emissions in the environmental impact assessment. For the residential development area, the seasonal PM 2.5 concentration was predicted by considering the FPM and CPM emissions in the target area as well as the surrounding areas. In winter and spring, when only FPM was considered, the air quality standards were not breached. However, when CPM results were included in the analysis, air quality standards were exceeded. However, it was predicted that even if CPM is included, air quality standards would not be breached in summer and autumn. This means that air quality forecasting results, including CPM, may alter the results. In addition, the sum of the predicted values of seasonal CPM and FPM was 1.7 to 3 times lower than that of the actual measurement. Compared to the background concentration measurement, it was found to be a result of long-distance travel. Therefore, it is necessary to consider CPM in the emission inventory to carry out environmental impact assessment, air quality modeling, analysis and diagnosis of emissions according to the characteristics of each sector's emission source, and prediction of the PM 2.5 concentration in the surrounding areas. In this study, environmental impact assessment was performed by considering only primary PM using the CALPUFF model. What remains to be undertaken by future research is an environmental impact assessment including secondary PM.