Crop Residue Burning Emissions and the Impact on Ambient Particulate Matters over South Korea

: In the study, crop residue burning (CRB) emissions were estimated based on ﬁeld surveys and combustion experiments to assess the impact of the CRB on particulate matter over South Korea. The estimates of CRB emissions over South Korea are 9514, 8089, 4002, 2010, 172,407, 7675, 33, and 5053 Mg year − 1 for PM 10 , PM 2.5 , OC, EC, CO, NO x , SO 2 , and NH 3 , respectively. Compared with another study, our estimates in the magnitudes of CRB emissions were not signiﬁcantly different. When the CRB emissions are additionally considered in the simulation, the monthly mean differences in PM 2.5 (i.e., (cid:52) PM 2.5 ) were marginal between 0.07 and 0.55 µ g m − 3 over South Korea. Those corresponded to 0.6–4.3% in relative differences. Additionally, the (cid:52) PM 10 was 0.07–0.60 µ g m − 3 over South Korea. In the spatial and temporal aspects, the increases in PM 10 and PM 2.5 were high in Gyeongbuk (GB) and Gyeongnam (GN) provinces in June, October, November, and December.


Introduction
The air quality in South Korea has been gradually improving over the past decades [1]. The recent concentrations of the particulate matter showed a rapid decline, in particular during the period of the coronavirus COVID-19 pandemic [2,3]. Nevertheless, there is still apprehension about serious air quality, and the annual average levels of PM 10 and PM 2.5 do not meet the WHO air quality guidelines [4]. The South Korean government planned to reduce the PM 2.5 emissions by 35.8% by 2022 from 2014. To reduce the concentrations of particulate matter, the government has established policies (e.g., Special Act on Fine Aerosol) for each ministry and has implemented reduction measures as follows [3]: - In energy production, conversion from coal-fired power plants to liquefied natural gas (LNG)-fired plants (i.e., coal-to-gas switching). - In the industry section, extending the environmental management policy for total amount control of air pollutants from the Seoul Metropolitan area to the whole country. -In transportation, early scrappage of old diesel vehicles and supply of low-emission vehicles such as electric and hybrid vehicles.
In the agricultural section, the South Korean government also strengthened the enforcement actions against illegal open burning activities. Nevertheless, as agricultural practices, crop residue burning (CRB) activities have been conducted in the field after harvest to manage the large quantities of the crop residues and to provide some nutrients from the residues (e.g., ashes) into the soil [5,6]. However, there are few studies to quantify the CRB emission and examine their impacts on particulate matter in South Korea.
Similarly, the CRB has often occurred in China, India, Myanmar, Thailand, Laos, Cambodia, and Vietnam [7][8][9][10][11]. The emissions from crop residue burning (or biomass burning) activities are closely related to radiative effects, air pollution, and public health [12][13][14][15]. Crutzen and Andreae reported that gaseous species of CO, CH 4 , NO x , and VOCs from biomass burning influence the oxidation capacities of the atmosphere by reacting with OH radicals or elevating the levels of O 3 [13]. Particulate species emitted from the CRB can play a role in condensation nuclei leading to changes in radiative balance. In terms of air pollution, the peak concentrations of PM 2.5 due to the enhanced biomass burning emissions reached~500 µg m −3 in May 2012 in Chengdu, China [16]. Importantly, many investigators have attempted to estimate the biomass burning emissions based on bottom-up, top-down, and hybrid approaches on regional and global scales, particularly in China and India, as summarized in Table 1.
Generally, the emission fluxes are calculated by the multiplication of emission factors and dry masses of burned crop residues. In a conventional way (i.e., bottom-up (B) approach), the information on the dry mass of crop residues, agricultural waste burning, and its areas is obtained from statistical data and field surveys to farmers. The approach can lack accurate spatial and temporal allocations of the statistical and survey data. On the other hand, the remotely sensed observations (e.g., MODIS, SEVIRI, GOES) have been used to quantify the biomass burning emission [23,25,27,33,35,38,40,42,[50][51][52]. The top-down emissions of biomass burning are estimated, multiplying several parameters of (i) burned areas, (ii) biomass load (FRP, fire radiative power), (iii) combustion efficiency (or conversion factor), and (iv) emission factors [53,54]. The typical products of the biomass burning (BB) emissions calculated from the MODIS sensor are the Fire Inventory from NCAR (FINN) and Global Fire Emissions Database (GFED) v3-v4, which are inventories, used worldwide to consider the BB emissions in the 3D-chemistry transport model simulations [36,[55][56][57]. Figure 1 shows the spatial distributions of the annual carbon monoxide (CO) and black carbon (BC) emission fluxes from the FINNv1.5 inventory over East Asia for 2016. CO and BC are the main products of incomplete combustion of biomass burning.  As shown in Figure 1, no hot spots were observed in South Korea, unlike many points in the biomass burning emissions of CO and BC over China. Thus, we believe that the top-down CRB emissions were poorly characterized in South Korea. This absence is closely related to the lower detection of the MODIS sensor for the small and short-interval fires due to its moderate spatial resolution and sparse observation frequency [44,58]. To improve the capability of the detections for small fires, many investigators have used multiple/geostationary satellite retrievals [37,51,59,60] or the hybrid MODIS-Landsat method with Google Earth Engine [43] in Table 1. Despite many efforts to improve the capability, the estimation algorithm based on satellite observations (i.e., top-down approach based on multi-sensors) is still hampered by cloud or hazy conditions [61]. Therefore, both estimations through field survey and remote sensing satellite observations compensate for mutual BB emissions [43,44].
Despite many uncertain issues in the bottom-up emission inventory, the objective is to build the bottom-up emission database of crop residue burning over South Korea. For the estimation, a field survey and combustion experiment were carried out for the 19 types of crops considered in South Korea. Thus, these activity data and emission factors are originated from actual situations in South Korea. The second objective of the study is to examine the impact of CRB emissions on the concentrations of particulate matter in South Korea.

Experiments and Methods
The impact of agricultural residue burning on the concentrations of particulate matter in South Korea was investigated using two different WRF-CMAQ simulations with and without the emissions of crop residue burning (CRB), denoted as "SM1" (or default) and "SM2" simulations in Figure 2, respectively. The conceptual framework of the study is given in Figure 2. For the SM2 simulation, the CRB emissions were calculated by the activity data and emission factor and analyzed spatially and temporally in detail in Section 2.2.

WRF-CMAQ Model Simulations
A 3-dimensional Eulerian Chemistry Transport Model (3D-CTM), US-EPA Models-3/CMAQ is one of the most powerful tools to manage the air quality for environmental policymakers and understand atmospheric phenomena for atmospheric scientists. The CMAQ model has been used worldwide to simulate the atmospheric levels of multi-pollutants, particularly focusing ozone and particulate matter, on local and regional scales because it is well built up and parameterized in gas-phase/heterogeneous chemistry and aerosol dynamics/thermodynamics (refer to CMAQ user's guide for more details [62,63]). In this study, we also utilized the CMAQ v5.2.1 model to simulate the air quality and examine the impact of crop residue burning (CRB) on the levels of particulate matter in South Korea during the year of 2016 [63]. The spatial domain of the simulations consists of 273 rows and 204 columns of 15 × 15 km 2 resolution with 27 vertical layers from the surface to~50 hPa over East Asia. The CMAQ model simulations were configured with Stateside Air Pollution Research Center-07 (SAPRC-07) for the gas-phase chemistry [64,65] and sixth-generation model CMAQ aerosol model (AERO6) for the aerosol dynamics [66,67]. For the horizontal advection in the CMAQ simulations, we utilized Yamartino (YAMO) scheme [68], showing good performance in terms of mass conservation [69,70]. Additionally, the Multiscale and ACM2 schemes were used for the horizontal and vertical mixing processes [71][72][73].
The CMAQ model was derived by the meteorological fields produced from the WRFv3.8.1 with the ARW dynamic solver [74]. The initial and boundary data were the National Center for Environmental Prediction (NCEP) Final operation global analysis (ds083.3) obtained from the NCAR (https://rda.ucar.edu accessed on 14 February 2022). These input data have a spatial resolution of 0.25 • × 0.25 • with a 6 h interval. The WRF simulations employed the single-moment 6-class for microphysics scheme [75]. Other physical methods were Yonsei University (YSU) PBL parameterization [76], Rapid Radiative Transfer Model (RRTM) scheme for long-and short-wave radiations [77], and Kain-Fritch scheme for cumulus physics [78]. The chemical and physical parameterizations are given in Table 2.

Emission Inventory
Atmospheric levels of gaseous and particulate species are significantly influenced by emission data [83]. For anthropogenic emission, we utilized the 0.1 • × 0.1 • resolved KORUS v5.0 emission inventory prepared from the SMOKE-Asia processing [84] over the entire domain except for South Korea. The emissions fluxes of CO, NO x , SO 2 , NH 3 , PM 2.5 , PM 10 , and total VOCs from the KORUS v5.0 were replaced with those from the 1 km × 1 km resolved Clean Air Policy Support System (CAPSS) inventory for South Korea. However, the KORUS v5.0 inventory was also used for information on the VOC speciation because the CAPSS inventory provided only total VOC emission fluxes instead of the speciated VOCs. For biogenic emissions, the Model of Emissions of Gases and Aerosol from Nature-Monitoring Atmospheric Composition and Climate (MEGAN-MACC) [85] was obtained from the monthly Emission of atmospheric Compounds and Compilation of Ancillary Data (ECCAD, https://eccad.aeris-data.fr/ accessed on 14 February 2022). For the fire emission, we also took into account the daily Fire Inventory from NCAR (FINN) v1.5 (https://www.acom.ucar.eud/Data/fire/ accessed on 14 February 2022) [55]. These emission data were spatially re-gridded into the model grids for the SM1 (default) simulation. The spatial distributions of emission fluxes used in the SM1 simulation were shown over South Korea and East Asia in Figure 3 and Figure S1, respectively. In this study, monthly emissions of CO, NO x , SO 2 , NH 3 , PM 2.5 , PM 10 , and speciated VOCs from open burning of crop residue were estimated over South Korea using several activity data and emission factors. The monthly emissions (E ij ) of pollutants (i) from residue burning of nineteen crops (j) were quantified from Equation (1) based on the methodology suggested by the studies of Delmas et al. [86], California Air Resources Board, US [87], and NSW Environment Protection Authority, Australia [88]. For the estimation, we considered nineteen crops, including rice (in the rice fields and ridges), barley, sweet potato, corn, soybean, pepper, sesame, perilla, peanut, garlic, onion, watermelon, apple, persimmon, citrus, pear, grape, and peach.
where, M j represents masses of the residue of crop type j burned in the open field (unit: kg month −1 ) and is defined as the multiplication of P j (production of crops, kg month −1 ), R j (conversion factor for crop residue production or residue-to-crop ratio), D j (correction factor for dry matter from the crop residues), and F j (fraction of crop residue burning in the field). The production of crops (P j ) was obtained from Statistics Korea [89]. The conversion factors (R j ) were obtained from a report of the National Institute of Agricultural Sciences (NAS) in Korea [90] and summarized in Table S1. For other activity data, we carried out surveys (e.g., face-to-face, mail, and internet surveys) on the open burning of crop residue, targeting approximately 1000 farmers. The interviewees were selected considering their distributions of the agricultural areas and farmworkers by region. The questionnaire for the survey includes (i) what fraction of each crop residue is burned in the open field, (ii) how frequently, and (iii) which season the burning occurs in. In the study, to estimate the emission factors (EF i,j ) of chemical species for 10 crops of rice, garlic, onion, sweet potato, barley, corn, peanut, watermelon, persimmon, and citrus fruit, the combustion test facility was built for the open combustion of crop residues under well-controlled conditions, which was designed based on the EPA method 5G [91]. The facility is composed of five components: (i) combustion chamber in which the crop residue is burned, (ii) inlet and outlet pipes, (iii) filter-based sampling part for PM 10 , PM 2.5 , EC, OC, and ion species, (iv) measuring part for VOCs, and (v) measuring part for other gases (e.g., NO x , CO, SO 2 , NH 3 ). The schematic diagram for the combustion test facility is given in Figure S2. Then, particulate matters (PM 10 and PM 2.5 ) and gaseous species are analyzed qualitatively and quantitatively based on several analysis methods in Table S2. The emission factors of gaseous and particulate species were summarized in Table S3. The emission fluxes from the crop residue burning were estimated based on the municipal-level divisions (e.g., Si for the city, Gun for the county, and Gu for the district) in Korea. Finally, the emission fluxes based on the Si-Gun-Gu levels were spatially allocated to the modeling grids using the software of the Geographic Information System (GIS) and Spatial Allocator (SA). The spatial distributions of emission fluxes from the crop residue burning (CRB) were presented in the right two columns of Figure 3. The CRB emissions were then added into the default emission for the SM2 simulation.

Observations
The performance of the chemical model was evaluated by comparing the concentrations of O 3 , PM 10 , and PM 2.5 observed at~1400 sites in China and 320 sites in South Korea. These data were collected for the entire year of 2016 from both the China National Environmental Monitoring Center (CNEMC, https://quotsoft.net/air/ accessed on 14 February 2022) [92] in China and the national air quality monitoring network (AirKorea, https://www.airkorea.or.kr/ accessed on 14 February 2022) [93] in South Korea (Figure 4). In addition to the ground data, satellite-observed data were also collected from TEMIS (www.temis.nl accessed on 14 February 2022) [94] for evaluation over the entire domain. The Ozone Monitoring Instrument (OMI) observation data retrieved from the Quality Assurance for Essential Climate Variables (QA4ECV) algorithm provides the tropospheric vertical NO 2 columns on a daily basis. We extracted and utilized only valid data according to several flag criteria (e.g., processing_error_flag = 0; solar_zenith_angle < 80; snow_ice_flag < 10 or snow_ice_flag = 255; amf_trop/amf_geo > 0.2; cloud_radiance_fraction_no2 < 0.5) suggested in the document for the QA4ECV product [95]. Averaging kernels were also considered in the CMAQ-simulated NO 2 columns for direct comparison with the OMI observations [96].

Emission Fluxes (E) Estimated from Residue Burning of Crops over South Korea
As discussed previously in Section 2.2, the CRB emission fluxes (E i,j ) were estimated from several emission factors (EF i,j ) and activity data (M j ) related to crop residue burning. The annual emission fluxes over South Korea estimated as 9514, 8089, 4002, 2010, 172,407, 7675, 33, and 5053 Mg for PM 10 , PM 2.5 , OC, EC, CO, NO x (as NO 2 ), SO 2 , and NH 3 , respectively (in Table 3), which corresponded to 4. 16, 7.93, 8.21, 11.96, 10.27, 0.47, 0.01, and 1.86% of those from anthropogenic sources. The contributions from the crop residue burning to total emissions were not small, particularly for EC and CO, above 10%. However, such contributions can vary considerably depending on the analysis regions defined in Figure 3(a2). From a regional perspective, most species showed the highest emission fluxes in the order of Gyeongbuk (GB), Gyeongnam (GN), Chungbuk (CB), Chungnam (CN), and other provinces in Table 3. The high emissions were also clearly seen in the spatial distributions of CRB emission in Figure 3. GB province, the largest emitter of CRB, accounted for 33-41% and 14-37% for particulate and gaseous species, respectively. On the other hand, the CRB emissions in SEO, BS, DG, IC, GJ, DJ, US, and SJ were insignificant due to the limited cultivation areas in the metropolitan cities [89]. This shows a clear spatial contrast between the CRB and anthropogenic emissions (refer to Figure 3).
To evaluate the approximate magnitudes of emission fluxes in this study, we compared our estimations with data of Kim et al. (hereinafter referred to as KIM) [97], conducted for South Korea based on a similar bottom-up approach. Figure 5 showed scatter plots between KIM and our estimates for available species of PM 2.5 , PM 10 , CO, NO x , SO 2 , and NH 3 . In this analysis, emission fluxes in CN province also included those of SJ province. As shown in Figure 5, many data points were significantly scattered along the 1:1 line. Those were possibly due to two aspects. For the first, we considered seven more crops of persimmon, citrus fruit, sweet potato, rice, garlic, watermelon, and onion than those in KIM, leading to data points distributed above the 1:1 line (i.e., higher estimates). Secondly, the emission factors (EF) for barley were significantly different. The EF PM2.5 , EF PM10 , and EF CO of KIM were 6-15 times larger than our experiment values. In contrast to the first aspect, data points can be distributed below the 1:1 line. In the scatter plot for NH 3 , the large biases were determined dominantly by the high EF NH3 for watermelon, considered only in our estimate (Table S3). Nevertheless, both estimates were not significantly different in their magnitudes. Additionally, in both studies, GB province was ranked the highest in common for PM 2.5 , PM 10 , CO, and NO x . Annual emission fluxes from our estimate and KIM were summarized in Table S4.     Figure 6 shows monthly variations of CRB emissions of gaseous and particulate species over South Korea. As shown in Figure 6, while the CRB emission fluxes were generally high in the warm season, particularly in June for the gaseous species of CO, SO2, and NH3, they were great in the cold season for the particulate species of PMs, EC, and OC. Such differences are significantly related to the harvest time for the specific crop with its different emission factors for the chemical species. Accordingly, different harvest times and emission factors for the crops also led to significant differences in the monthly contributions. For example, the contributions were generally marginal within 0.02, 1.37, and 6.10% for the gaseous species of SO2, NOx, and NH3 (in the right column of Figure 6), respectively. On the other hand, the maximum contributions reached 14.25, 14.58, and 24.78% for PM2.5, OC, and EC (in the left columns of Figure 6), respectively. The emission fluxes of EC were relatively small compared to those of PM10 and PM2.5, despite it being the highest contribution.   Figure 6 shows monthly variations of CRB emissions of gaseous and particulate species over South Korea. As shown in Figure 6, while the CRB emission fluxes were generally high in the warm season, particularly in June for the gaseous species of CO, SO 2 , and NH 3 , they were great in the cold season for the particulate species of PMs, EC, and OC. Such differences are significantly related to the harvest time for the specific crop with its different emission factors for the chemical species. Accordingly, different harvest times and emission factors for the crops also led to significant differences in the monthly contributions. For example, the contributions were generally marginal within 0.02, 1.37, and 6.10% for the gaseous species of SO 2 , NO x , and NH 3 (in the right column of Figure 6), respectively. On the other hand, the maximum contributions reached 14.25, 14.58, and 24.78% for PM 2.5 , OC, and EC (in the left columns of Figure 6), respectively. The emission fluxes of EC were relatively small compared to those of PM 10 and PM 2.5 , despite it being the highest contribution. For anthropogenic and CRB sources, we also examined the ratio of PM 2.5 to PM 10 emissions in this study. The ratios can be used to identify the sources because the particulate matter in different sizes possibly originates from different sources [98,99]. In the anthropogenic emissions from the CAPSS inventory, the monthly PM 2.5 /PM 10 ratios were almost constant, ranging between 0.41 and 0.44. For the CRB emission, the average ratio was 0.85 in the range of 0.82-0.90, indicating relatively large amounts of PM 2.5 emitted through the agricultural residue burning. Our averaging ratios for the CRB emission are similar to values from observations during the periods of biomass burning events at Gwangju Korea and Pathumthani Thailand [100,101].

Modeling Performances
We evaluated the general performances of the default CMAQ simulation (i.e., SM1) by comparing the tropospheric NO 2 columns over the entire domain. The hourly mixing ratios (unit: ppm) of NO 2 at each layer from the simulations were converted to the vertical column concentrations (unit: molecules cm −2 ) by integrating vertically from surface to approximately 10 km in altitude for direct comparison [102].
The first and second columns of Figure 7 showed spatial and seasonal distributions of the CMAQ-calculated and OMI-retrieved NO 2 columns over East Asia. The differences and scatter plots between the two columns were presented in the third and fourth columns of  Figure 7 (also refer to the statistical values and slopes in the fourth column of the figure), indicating the underestimation of NO x emissions in these areas. According to the study of ozone production efficiency (OPE) by Oak et al. [103] using KORUSv5 emission data, the OPE characteristics are best simulated when 50% of NO x emission increase in the chemistrytransport modeling. Such results also indicate the underestimation of KORUSv5 NO x emissions used in our study. Nevertheless, our CMAQ simulations using the same emission inventory well captured the satellite-observed spatial distributions and seasonal variations over East Asia. The correlation coefficients were quite good, ranging between 0.85 and 0.88. During the summer episode, the lower concentration observed from the OMI sensor was also well simulated by the CMAQ model in the second row of Figure 7. The low NO 2 columns in summer are attributed to the active NO x chemical losses via the reaction of NO 2 with high concentrations of OH radicals [104,105].
In addition to satellite, in situ ground observations from the CNEMN and AirKorea were utilized to evaluate the modeling performance. The hourly observation was compared to the simulated data averaged for the pixels corresponding to the CEC, CEC2, and SK regions defined in Figure 4. Figure 8 showed the time series of hourly surface concentrations of O 3 , PM 10 , and PM 2.5 from observations (black lines) and simulations (blue lines) for CEC, CEC2, and SK in the first, second, and third columns, respectively.
In the upper panel of Figure 8, the simulated O 3 concentrations were higher by 20-25%, compared to the observation. The high O 3 mixing ratios simulated from the CMAQ simulation are strongly related to the underestimation of NO x emissions over East Asia discussed in the NO 2 column analysis in Figure 7 because of less O 3 being titrated by NO. The correlation coefficients were good, generally ranging from 0.72 to 0.91. In Figure S3 In the bottom panel of Figure 8, the simulated concentrations of PM 10 were small by −50.8, −25.5, and −38.7%, compared to the observations over CEC, CEC2, and SK. Nevertheless, the correlation coefficients between the simulated and observed PM 10 were moderately good (0.76 and 0.77) over CEC and CEC2. The observations captured the Asian dust in China and its impact on PM 10 in South Korea in May and June 2016. However, the high concentrations of PM 10 were not adequately simulated in the CMAQ simulations, in which the dust module for dust emission was off. The lower correlation coefficient in SK was mainly due to some inconsistency, particularly during April-June (refer to Figure S4). It is well known that the concentration of PM 10 is closely related to the Asian dust event [106].
On the other hand, the temporal variations and magnitudes of observed PM 2.5 were well simulated, as shown in the middle panel of Figure 8. Accordingly, the correlation coefficients for PM 2.5 were good as 0.87, 0.84, and 0.70 over CEC, CEC2, and SK, respectively. Although the simulated PM 2.5 concentrations were also small by −27.6, −0.4, and −20.1% over the analysis regions, PM 2.5 was better simulated than PM 10 in terms of several statistical parameters between simulations and observations (i.e., absolute and relative differences, correlation coefficient, and slope in the scatter plots). Additionally, the spatial distributions of PM 2.5 concentrations were well consistent with those of in situ observation in Figure S5. Therefore, we more focused on PM 2.5 for the impacts of crop residue burning. Figure 7. Spatial distributions of CMAQ-simulated and OMI-retrieved tropospheric NO 2 in the first and second columns. Their differences and scatter plots between two columnar NO 2 in third and fourth columns, respectively. In statistical analysis, R, MB, NMB, S, and N represent correlation coefficient, mean bias (molecules cm −2 ), normalized mean bias (%), slope, and the number of data, respectively. Collectively, the performance for SM1 simulation showed underestimation of PM 10 and O 3 in the CEC region, which affects the air quality in South Korea. Nevertheless, there were relatively consistent for PM 2.5 . From the statistical analysis between the simulated and observed O 3 and PM 2.5 concentrations, we believe that the SM1 simulation has sufficient performance to investigate the effect of the CRB emission on particulate matter in South Korea.

Impact of Crop Residue Burning on Particulate Matters
The impact of the crop residue burning on the concentrations of particulate matter was examined in terms of absolute (unit: µg m −3 ) and relative (unit: %) differences using Equations (2) and (3) as follows: A. Di f f = C SM2 − C SM1 (2) R.
where, C SM1 and C SM2 represent the concentrations of PM 2.5 or PM 10 from the SM1 and SM2 simulations, respectively. Figures 9 and 10 show the spatial distributions of absolute and relative differences of PM 2.5 between the SM1 and SM2 simulations over South Korea.
The positive values of the absolute differences colored in red in Figure 9 indicate the elevated PM 2.5 originated from the crop residue burning. The absolute differences were lowest at 0.07 µg m −3 in July and were generally high in June, October, November, and December (highest at 0.55 µg m −3 in October). The monthly absolute difference increased to a maximum of 1.82 µg m −3 at a specific pixel of GN province (defined in Figure 9) in June. The relative difference at the same specific pixel point in GN province also increased by 12.91% in June. From a temporal viewpoint, the relative differences were highest (4.33%) in October and lowest (0.60%) in March. From a spatial point of view, the relatively high elevations occurred in the rural areas of the northern GB, central GN, northern CB, southern CN areas. For PM 10 in this study, we also examined the absolute and relative differences in Figures S6 and S7. As shown in the analysis, the temporal and spatial features for PM 10 were not much different from those for PM 2.5 .
From such analysis, most increases in the PM 10 concentration were attributed to the increase in the PM 2.5 concentration. The result was consistent with the PM 2.5 /PM 10 emission ratios discussed in Section 3.1. Because PM 2.5 has a more severe impact on cardiorespiratory disorders than PM 10 , the high ratios are more important from a public health perspective [107]. We summarized the impacts of CRB on the monthly mean and maximum concentrations of surface PM 2.5 and PM 10 in terms of absolute (AD) and relative differences (RD) in Tables 4 and 5. Collectively, our simulations showed that over the entire period of 2016 in South Korea, the monthly PM 2.5 (and PM 10 ) concentrations originated by the crop residue burning were elevated within 0.55 µg m −3 (and 0.60 µg m −3 ). Table 4. Impacts of crop residue burning on the monthly concentrations of particulate matters over South Korea, in terms of absolute (AD) and relative differences (RD).  The contributions from the CRB emissions to the concentrations of particulate matter estimated in this study are still subject to some uncertain issues. The uncertainty is related to the diurnal variations of the CRB emissions. Although the contributions in this study were estimated every month, the diurnal variations of CRB emissions can be randomly large. Open crop residue burning activity has often been carried out surreptitiously in the dark evening in South Korea to avoid surveillance for the illegal burning [50,108]. Thus, the peak contribution occurs possibly between late evening and early night during the day. As the planetary boundary layer (PBL) decreases at that temporal point, the surface concentrations of particulate matter or its contribution can further increase [109]. However, the time of open burning during the day varies spatially, depending on country-specific agricultural practices [44,50]. Additionally, as input data for the estimation, the activity data still has room for improvement through the precise representativeness of the survey population to the reality. Table 5. Impacts of crop residue burning on the maximum concentrations of particulate matter at certain pixels in South Korea, in terms of absolute (AD) and relative differences (RD).

Summary and Conclusions
In this study, we attempted to quantify the impacts of crop residue burning (CRB) on the concentrations of PM 10 and PM 2.5 in South Korea. For the investigation, we built the CRB emission database on the basis of the bottom-up approach over South Korea. Then, two different one-year WRF-CMAQ simulations with and without the CRB emissions were conducted for 2016.
The monthly CRB emissions over South Korea were estimated using (i) several activity data of nineteen crops from the online and offline surveys with statistical data in Korea and (ii) emission factors for the gaseous and particulate species through a combustion test facility. In our estimation, the emissions of CRB over South Korea are 9514, 8089, 4002, 2010, 172,407, 7675, 33, and 5053 Mg year −1 for PM 10 , PM 2.5 , OC, EC, CO, NO x (as NO 2 ), SO 2 , and NH 3 , respectively. The contributions of the CRB emissions to the anthropogenic emissions were generally small, except for EC and OC, which have more than 10% contributions. In the regional comparison with the KIM study [97] over South Korea, our estimates were not significantly different except for NH 3 and SO 2 species. In both studies, Gyeongbuk (GB) province was the largest emitter of CRB emission in common for PM 2.5 , PM 10 , CO, and NO x . The CRB emissions in GB accounted for approximately 33-41% of the total amounts for the particulate species. In terms of seasonal variation, the crop residue burning emissions were high in the cold season for PM 10 , PM 2.5 , EC, and OC. On the other hand, they were higher in the warm season for CO, SO 2 , and NH 3 .
From two different WRF-CMAQ simulations, we found the monthly average concentrations of PM 2.5 (PM 10 ) over South Korea were elevated between 0.07 and 0.55 µg m −3 (between 0.07 and 0.60 µg m −3 ) when the CRB missions are considered in the CMAQ simulation. The elevations in PM 2.5 (PM 10 ) were corresponded to 0.6-4.3% (0.5-3.1%) in relative differences. The increases in PM 2.5 were high in June, October, November, and December. Spatially, the relatively high elevations occurred in the northern Gyeongbuk (GB), central Gyeongnam (GN), northern Chungbuk (CB), and southern Chungnam (CN) provinces in South Korea.
Despite large uncertainties in the activity data, our crop residue burning emission from this study is expected to be an alternative database for the areas where the satellite observations for the top-down estimation are not available, in part, for small and shortlasting fires or cloudy conditions. The emission data can also be applied to the air quality modeling to evaluate the associated disease/public health burden and impact on climate change. Finally, our analysis contributes to determining the effectiveness of the reduction measure when the Korean government establishes the reduction policy to reduce the concentrations of PM 10 and PM 2.5 . From the study, the impact of the CRB emissions on PM 10 and PM 2.5 were marginal over South Korea. Nevertheless, the impacts could vary significantly from a temporal and spatial point of view. Thus, further studies should investigate the in-depth diurnal/temporal variations of the crop residue burning emissions based on the combined top-down and bottom-up approaches.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/atmos13040559/s1, Figure S1: Annual emission fluxes of (a) PM 10 , (b) PM 2.5 , (c) OC, (d) EC, (e) CO, (f) NO, (g) SO 2 , and (h) NH 3 combined from the KORUS v5, MEGAN, and FINN v1.5 inventories used in the SM1 simulation; Figure S2: Schematic diagram of combustion test facility; Figure S3: Spatial distributions of O 3 mixing ratios from the SM1 simulations over South Korea. The colored circles represented the observed values from the AirKorea network (unit: ppb); Figure S4: Spatial distributions of PM 10 concentrations from the SM1 simulations over South Korea. The colored circles represented the observed values from the AirKorea network (unit: µg m −3 ); Figure S5: Spatial distributions of PM 2.5 concentrations from the SM1 simulations over South Korea. The colored circles represented the observed values from the AirKorea network (unit: µg m −3 ); Figure S6: Absolute differences for PM 10 ; Figure S7: Relative differences for PM 10 ; Table S1: Residue-to-crop ratios for crops considered in this study; Table S2: Analytic methods for gaseous and particulate species; Table S3: Examples of emission factors (E i,j ) of particulate and gaseous species for the crop residue (Unit: g/kg); Table S4: Estimated CRB emission and their comparison with Kim et al. [97] (unit: g/kg). Additional information are explained in [110,111]