Long-term Assessment of Ozone Nonattainment Changes in South Korea Compared to US, and EU Ozone Guidelines

We conducted spatiotemporal assessments of ozone in South Korea from 1990–2020 to evaluate trends and compare changes in compliance based on South Korean, US, and EU standards. Observational data from nationwide air-quality monitoring stations were collected and converted to the maximum daily 8 hr ozone average (MDA8O3). Seasonal ozone variations displayed an overall increase across most of South Korea and a noticeably high rate of 0.86 ppbv/yr in Seoul, with an even higher rate 1.2 ppbv/yr for the fourth-highest MDA8O3. Recent air-quality regulations to reduce NOx emissions have been estimated to weaken NO titration effects, leading to higher ozone levels for VOC-limited urban areas in South Korea while decreasing ozone concentrations elsewhere. In recent years, nearly all monitoring stations have exceeded the South Korean MDA8O3 standard, leading to debate regarding the adequacy of current standards for monitoring changes in nonattainment. Comparison with EU and US standards showed that implementing these could significantly lower nonattainment events due to the easing of target threshold values by either percentile or concentration values. Relative distances in nonattainment percentages between South Korean and other standards indicated that the EU ozone guideline was most suitable for tracing recent ozone changes not apparent when using the South Korean or US standards.


INTRODUCTION
Atmospheric pollution is a complex problem in urban areas in South Korea.Although primary pollutants have been successfully controlled by effective policies, secondary pollutant reduction has been less efficient due to multiple precursors and complex reactions (Shin, 2015;Lim, 2012;Thi Nguyen and Kim, 2006).Tropospheric ozone (O 3 ) is a representative secondary air pollutant produced by nonlinear photochemical interactions in polluted air masses containing nitrogen oxides (NOx) and volatile organic compounds (VOCs) (Monks et al., 2009).This is an active radical known for its role in the initiation of photochemical oxidation processes and is of growing concern given the substantial threat posed to human health and vegetation (Fleming et al., 2018;Mills et al., 2018;Monks et al., 2015;Kinney, 2008).Ozone also affects climate change as the third-largest anthropogenic greenhouse gas (Zhang et al., 2016).
Rapid population growth and industrialization in East Asia have driven significant increases in ozone precursor emissions, leading to high ozone concentrations during the South Korean spring and summer.Effective strategies to reduce the impact of ozone require knowledge of ozone precursor emission sources and interactions between the precursors.Consequently, the regulation of ozone precursor emissions has been a major emphasis for many years.However, despite consistent efforts, South Korean ozone levels have steadily increased in past decades (Kim et al., 2018) and no monitoring sites have been compliant with the country's ambient air quality standards for recent years (NIER, 2020).
Although South Korea's ozone standards were first established in 1983, they were last modified in 1993 to establish the current 99 th percentile of maximum daily average 8 hr ozone (MDA8O3) as ≤60 ppbv and the 999 per mill standard for 1 hr ozone as <100 ppbv.These strict guidelines (especially for MDA8O3) mean that all South Korean monitoring sites in recent years have exceeded the national standard (NIER, 2020), creating a growing need to reevaluate the standards by comparison with other countries.For example, China's ozone standard for urban regions (established in 2012) is a daily 8 hr mean of 160 μg/m 3 and an hourly mean of 200 μg/m 3 (Li et al., 2018) and Japan's 1 hr mean photochemical oxidant (Ox) target (established in 1973) is 60 ppbv (You et al., 2017).In the US, the Clean Air Act established an air-quality management framework in 1970 and the National Ambient Air Quality Standards for ozone have been tightened three times since 1997.The current US ozone standard matches the fourth highest of MDA8O3 (≤70 ppbv), averaged across three consecutive years (US EPA, 2015), while the EU Ambient Air Quality Directive sets the target threshold value for MDA8O3 as ≤120 μg/m 3 for more than 25 days/yr averaged over three years (European Commission, 2018;González Ortiz, 2013).The EU's long-term objective value is that MDA8O3 may not exceed 60 ppbv, which is equivalent to the 93.2 percentile value each year (EEA, 2020;Glatthor et al., 2007).These differences between national standards reflect differing methods for setting exceedance days, averaging periods, or even standard concentration levels (Park et al., 2016).Although the public-health benefits of stricter air-quality guideline values is clear, the adequacy of the current South Korean MDA8O3 standard has been a persistent subject of debate for many years.
In this study, we conducted a spatiotemporal analysis of South Korean ozone observations from 1990-2020 to characterize their trends and assess changes in ozone compliance with respect to South Korea, US, and EU standards.We calculated annual nonattainment percentages based on these standards and demonstrated how the US and EU standards were able to address hidden changes in ozone nonattainment in South Korea, which has been saturated for the last ten years.

1 Data
The number of urban air quality stations in South Korea rapidly increased from 54 in 1990 to 473 in 2020 (Table 1) after the first national long-term air quality monitoring plan and operation guidelines were established (Kim and Park, 2020).Air-quality monitoring stations in South Korea are classified into 11 different types according to the main targeted air-quality issue (urban, roadside, background, rural, atmospheric trace metal, photochemical, hazardous air pollutant, acid deposition, air-quality super, PM 2.5 speciation, and comprehensive).Overall, South Korea has 4 major air-quality stations covering urban, rural, roadside, and background sites, forming the densest monitoring network in the world (Fig. 1).We selected hourly ozone data from urban, rural, and background sites that were appropriate for long-term analysis.
To calculate MDA8O3 for a given calendar day, we selected the highest value among 24 possible 8 hr rolling mean concentrations from the hourly ozone data observed at each station.If >25% of hourly data for a given day were not valid, MDA8O3 was not calculated for those days and was excluded from further analysis.Yearly and seasonal MDA8O3 averages were calculated from daily values.Any sites with <75% MDA8O3 data completion for a given season or year were also excluded.
Although each air-quality monitoring station had an associated suit of meteorological sensors, in urban locations these are often surrounded by buildings and tend to inadequately capture local weather conditions (especially wind direction and speed).We therefore used hourly meteorological data from 17 synoptic weather stations operated by the Korea Meteorological Agency to calculate daily maximum temperature, wind direction, morning and afternoon wind speed, relative humidity, sea-level pressure, solar irradiation, and precipitation.The MDA8O3 values for each air-quality station were matched with meteorological variables from the nearest weather station for long-term ozone analysis.

2 Trend Analysis
Most meteorological variables play significant roles in spatiotemporal ozone variations, making it necessary to conduct a meteorological adjustment analysis prior to assessing long-term ozone changes.Many statistical analyses have estimated temporal ozone trends as a function of various meteorological variables (Botlaguduru et al., 2018;Camalier et al., 2007;Ghim et al., 2001;Thompson et al., 2001).We employed the generalized linear model (GLM) to filter the varying effects of weather from the selected long-term yearly MDA8O3 data (Camalier et al., 2007).The detailed and identical weather adjustment method can be found in the other studies (Kim et al., 2018;Shin et al., 2017).GLM is a statistical linear regression model that utilizes the relationship between the response variable (MDA8O3) and predictor variables (weather parameters) via a link function.A separate model was prepared for each of South Korea's 17 first-tier administrative divisions (nine provinces and eight major cities) using GLM written in R software (R Core Team, 2021;Camalier et al., 2007).

3 Spatial Analysis
Kriging is used to distinguish the spatial distribution of variables at an unmeasured location from observed values at nearer locations (Tyagi and Singh, 2013), and is a useful tool for analyzing the geostatistical distributions of air pollutants in South Korea (Kim and Song, 2017;Kim et al., 2014;Park, 2005).Ordinary kriging assumes that the mean and variance of the variable are constant across the spatial domain.However, this stationarity assumption is often not valid for air pollution variables, which are largely affected by local influences and emission sources.As the covariances of our ozone data were more likely to be nonstationary, the simple ordinary kriging variogram method was not appropriate.Variograms are plots of the covariance between a pair of variable points (lags).Those for our ozone data indicated a varying covariance with a power function within 3-4 points.Thus, we applied kriging with a power function variogram model to estimate the spatial variation of the fourth-highest MDA8O3 in South Korea.

4 Non-attainment Analysis
In this study, non-attainment was defined as any area that does not meet the primary ozone standards for ambient air quality.In South Korea, the 95 th percentile of MDA8O3 should be less than the guideline value of 60 ppbv averaged for three years to satisfy national ozone attainment (MOE, 2015).However, this guideline is specialized for the designation of non-attainment in air quality regulation areas.Therefore, we used the national ambient air quality standard implemented in all areas.We assessed the exceedance (non-attainment) of the MDA8O3 by applying the guidelines of South Korea, US, and EU to 30 years of data in order to address the ozone compliance observed in South Korea.

1 Long-term MDA8O3 Trends
For long-term analysis, the 1 hr ozone concentration data from 1990-2020 were converted to MDA8O3 and grouped by administrative division.In Seoul, daily maximum temperature and solar irradiation had positive effects on MDA8O3, while humidity, precipitation, and cloud cover had negative effects (Fig. 2).High MDA8O3 (>100 ppbv) was consistently observed for westerly winds (between southwest and northwest), which could be explained by either long-range transport or local circulation patterns such as the sea-land breeze (Peterson et al., 2019;Park, 2018;Choi et al., 2014).Although not shown in Fig. 2, ozone concentrations were nonlinearly related to wind speed, generally peaking during low wind speeds in the morning (1-2 m/s) and afternoon (2-4 m/s).Higher ozone concentrations in Seoul under these calm conditions have been extensively discussed in previous studies on regional or local transport regimes (Choi et al., 2014;Ghim et al., 2001).
Regionally grouped MDA8O3 was modeled with meteorological parameters, using GLM to filter out weather effects on yearly long-term ozone variations.The deviance residuals, which represents the difference between the results, calculated from GLM implemented results was significantly low (-0.0338)indicating good fitting among the simulated results.During the ozone season in Seoul (May-August), the weather-adjusted variability tended to be smoother than the observed variability (Fig. 3), but the differences were not significant except in a few years.The largest difference occurred in 2020, when observed ozone declined significantly but weather-adjusted MDA8O3 showed a clear increase, implying that decreased ozone levels during summer 2020 were not primarily due to reduced anthropogenic activities related to the COVID-19 pandemic but were more likely due to weather effects.Seasonal MDA8O3 steadily increased from the mid-20s ppbv to >50 ppbv over the last 30 yr at a rate of 0.86 ppbv/yr.Although the increase of ozone concentration is noticeable in the 90s, this does not imply that the air quality in South Korea is declining.This is one of the highest ozone increase rates reported in the world and is comparable to those previously reported for long-term ozone variations in Seoul (Kim et al., 2018;Susaya et al., 2013).
Steady and rapid increases in ozone were found in all parts of South Korea.Weather-adjusted yearly averages of MDA8O3 and MDA8Ox (maximum daily 8 hr average Ox (O 3 + NO 2 )) in seven major cities and Gyeonggi Province (Fig. 4) showed that, despite large interannual variability, distinct long-term trends were clear, with similar rising rates throughout the country.This ubiquitous increase across South Korea has been attributed partially to rapidly increasing ozone in northeast Asia (Chang et al., 2017;Nagashima et al., 2017) and reduced NOx emissions, especially in urban areas in South Korea (Kim and Lee, 2018;Shin et al., 2017).Reducing ozone precursors such as NOx and VOCs should be the top priority when establishing air-quality policies in South Korea.However, it is first necessary to assess the current ozone standard, especially the 8 hr guideline, in order to implement and monitor effective ozone control policies.
Ox is often regarded as a more adequate and stable indicator of atmospheric oxidant levels, as it is not affected by the titration effects of ozone due to local NOx emissions.As shown in Fig. 4, seasonal Ox has tended to decline since 2016, in contrast to the continuous increase in ozone over the same period.While Ox in Gyeonggi Province showed a consistent upward trend, declining Ox was most distinct in major cities.This may be related to the Euro 6 automobile emission standard adopted in 2014 to reduce NOx emissions, which was enforced starting in 2016, especially with respect to diesel vehicles in South Korean cities.Most South Korean urban areas are VOC-limited regimes for ozone formation (Kim and Lee, 2018).The reduction of local NOx emissions and the sequential reduction of the ozone titration effect might have had a substantial impact on increasing seasonal MDA8O3 in urban areas in recent years.The recently revised air quality strategy includes a target for reducing NOx emissions by 64% by 2024, compared to the 2016 levels (MOE, 2019).Based on this plan, we expect only a 32% VOC reduction within the same period and predict that ozone levels in South Korea will rise, especially in urban areas, since ozone production is more limited by NOx as East Asia has slowly shifted toward a more NOx-sensitive regime (Souri et al., 2017).A recent study confirmed that the reduction of NOx greater than VOCs would increase ozone concentration till significant percentage of reduction was achieved in China (Wang et al., 2019).Temporal changes in the spatial distribution of ozone were also very clear (Fig. 5).The nationwide mean value of the fourth-highest MDA8O3 was 53 ppbv in 1990, which increased to 89 ppbv in 2020 at 1.2 ppbv/yr, higher than the seasonal average (0.86 ppbv/yr) of MDA8O3.In general, due to prevailing winds over the region, higher ozone tended to be located in downwind (eastern) regions.The most drastic increase in fourthhighest MDA8O3 occurred in the Seoul Metropolitan Area (SMA), which consists of Seoul, Incheon, and Gyeonggi Province.This is one of the most densely populated areas in the world, containing ~25 million people within 11,704 km 2 .The highest increase rate for ozone was observed in the southern SMA, where the highest population growth also occurred during the analysis period.Two isolated low-ozone areas occurred in central-northeastern South Korea (blue arrows, Fig. 5d).Space-and ground-based NO 2 monitoring analyses identified these areas as having large cement production facilities (Kim et al., 2020a, b), demonstrating that NOx emissions from this industry could increase MDA8O3 by up to 4 ppbv in surrounding regions.As locations near such industries would be influenced by ozone titration effects with respect to local NOx emissions, they   appeared on the map as local minima.In contrast, Jellabokdo Province (red arrow, Fig. 5d) is known for especially rapid growth of ozone and particulate matter (Kim et al., 2021), but the exact causes of this isolated ozone peak are not known and require further research.In Fig. 5(E-H), the decennial spatial distributions of 26 th highest MDA8O3, which is the equivalent to EU standard, were also listed to compare the degree of guideline com-  pliance between two standards.

2 Long-term Nonattainment Analysis
MDA8O3 compliance based on South Korean, US, and EU ozone standards was tested from 1990-2020.The number of days with MDA8O3 >60 ppbv was counted for each monitoring station every year for the South Korean standard while a three-year average was used for the EU standard and the annual three-year mean for the fourth-highest MDA8O3 was used for the US standard (Fig. 6).The number of non-attainment cases for MDA8O3 in South Korea increased steadily in recent years and was highest with the South Korean standard.Median exceedance day values for the South Korean standard reached 60 days in 2014, indicating that more than half of monitoring stations in South Korea observed MDA8O3>60 ppbv for >2 months.Every station exceeded the guideline value since 2013, except for 2020.This fact has been subject to severe debate regarding the significance of ozone pollution and the adequacy of current guidelines.
Non-attainment events in South Korea were much less common using the EU standard (Fig. 6b), with 25 days of permitted exceedance under EU standard and 4 days for South Korean standard.In addition, interannual variations in exceedance days were noticeably reduced in the EU case due to the three-year averaging effect, and this procedure allowed the long-term trend to be easily interpreted with little noise.Non-attainment cases were noticeably improved in recent years, while it was difficult to discern differences with other standards.The EU standard would therefore be a useful index for analyzing nonattainment changes, especially in South Korea.
More than half of the ozone monitoring sites did not meet the US ozone standard (Fig. 6c).However, the degree of nonattainment by the US standard fell between that for the South Korean and EU standards.In the US standard, non-attainment areas were further classified into six categories based on the degree of exceedance (marginal, moderate, serious, severe-15, severe-17, and extreme).In general, non-attainment areas in South Korea were classified as moderate (81-93 ppbv) and serious (93-105 ppbv) by the US standard.Although extreme areas (≥163 ppbv) were not observed in South Korea, 15-25% of monitoring sites were listed as severe, and ozone control efforts should be focused on these areas.

3 Standard Noncompliance Distance Analysis
The percentages of non-attainment monitoring stations among all stations for the South Korean MDA8O3 standard remained at ~40% in the early 1990s and rapidly increased in the late 1990s, reaching nearly 100% by 2010 (Fig. 7).Yearly percentages of ozone nonattainment by US guidelines tended to be ~20% lower than those of the South Korean standard until the early 2000s, though this also reached nearly 100% in the 2010s.The variation in non-attainment percentages was most widely distributed for the EU standard, with most monitoring stations being compliant in the 1990s and steadily increasing to a peak of nearly 100% in 2016.Non-attainment conditions for the EU standard actually improved after 2016, unlike the other two.The previously discussed strict  NOx control rule implemented in 2016 reduced NOx emissions significantly, such that urban ozone levels might have increased with a reduced NO titration effect but also improving attainment in rural locations where NOx was limited.
Although nonattainment outcomes were considerably different under the three standards, it was difficult to directly compare standards as the target values and statistical methods varied.Therefore, we tried to find the degree of similarity between changes in the current South Korean MDA8O3 standard and the others, considering either altering the percentile value of 99 or adjusting the standard concentration of 60 ppbv.We introduced a relative measure that quantified the similarity in compliance statistics between the modified South Korean ozone standard values (percentile and concentrations) and the other two, calculating their root sum squared (RSS) values from 1990-2020.
When the percentile values were modified, the minimum distance for the EU standard was reached at 92.3 percentile, as compared to 96.7 percentile for the US standard (Fig. 7), implying that ozone nonattainment statistics in South Korea were best matched to US and EU guidelines when the percentile value was reduced from 99 to 96.7 and 92.3, respectively at a fixed concentration value of 60 ppbv.Park et al. (2016) assessed the degree of compliance strictness by South Korean, US, and EU ozone standards from 2000-2004, using air quality monitoring data from Suwon City in Gyeonggi Province.They estimated that the equivalent levels of MDA8O3 standards in the US and EU were 97.5 and 95.7 percentile of the South Korean standard with a 60 ppbv concentration value, respectively.Our results were somewhat lower than these, especially for the EU standards, which may be attributed to the statistical methods used to address the differences between standards and in   We then applied the 96.7 and 92.3 percentiles to the South Korean ozone standard and compared its agreement with the US and EU standards with respect to nonattainment percentages.Those at 96.7 percentile were well-matched with the US standard (Fig. 7b), confirming that a simple modification of percentile value in the South Korean standard would make it equivalent to the US standard in terms of nonattainment percentage.However, the yearly South Korean standard at the 92.3 percentile showed larger interannual deviations compared to the three-year EU standard, as indicated by the larger relative distance (Fig. 7a).
The relative distances for nonattainment percentages of changing South Korean MDA8O3 standard concentration levels corresponding to the US and EU standards were lowest at 68 (US) and 76 (EU) ppbv (Fig. 8a).The minimum distance against USA standard along with concentration changes in Korean standard was lower than that with percentile changes.This implies the concentration adjustment in the South Korean standard was more appropriate for matching the USA standard.The yearly variation in nonattainment percentages at a 68 ppbv South Korean standard was highly correlated with that of the US standard (Fig. 8b).While modifications at a South Korean standard of 60 to 76 ppbv were successful in deriving the long-term changes in nonattainment percentages estimated by the EU standard, these showed much higher inter-annual variations.As most monitoring stations have not been compliant with the current standard throughout the last decade, it is difficult to assess the effectiveness of current air-quality policies, such as by using nonattainment values at a national level.An encouraging and daunting national plan recently proposed would significantly reduce NOx by 64% by 2024; this could increase the ozone concentration in urban settings but reduce it in other parts of the country.With the current South Korean standard, non-attainment changes would not be apparent for many years.In addition, adoption of the US standard would not improve the degree of saturated nonattainment.We found that the EU standard was the most suitable for tracking changes in nonattainment under present conditions.
However, simply changing the percentile or concentration in the current South Korean standard to correspond with the EU standard did not provide steady and reliable non-attainment results, unlike the original EU standard.Revising the primary standard is a great challenge, espe-cially when weaker limits are considered.WHO proposes interim target levels to nations where air pollution levels far exceed the guideline levels, in order to assess the progress towards meeting the national standard (WHO, 2006).The specialized guideline for the designation of air quality regulation areas in South Korea has been implemented since 2015.Therefore, the revision of the current specialized guideline value to EU MDA8O3 scheme would be easily adapted.However, the addition of a secondary or interim target value has to be considered in the Korea ozone standard for temporary and defined periods.This would be mainly oriented toward administrative purposes in order to better evaluate the implemented air-quality policy without compromising the purpose of public health protection.

CONCLUSIONS
The long-term temporal trend of MDA8O3 increased throughout South Korea, with an alarmingly high rate of 0.86 ppbv/yr in Seoul.In 2020, a distinct decrease in ozone levels was influenced by weather variables rather than by a pandemic-related reduction in precursor emissions.Ox trends indicated a reduction in NOx emissions due to strengthened policies established in 2016, but in VOC-limited urban regions, ozone levels were expected to rise due to weaker ozone titration effects.The nationwide averaged fourth-highest MDA8O3 had an even higher rate of growth (1.2 ppbv/yr) with the sharpest increase in the SMA.However, two isolated regions in central-northern South Korea had low ozone distributions that could be related to local cement production although this link has yet to be proven.
Ozone nonattainment areas covered nearly 100% of South Korea in recent years, indicating an alarming increase in ozone concentration.In contrast, nonattainment level was lower in the US, especially with EU guidelines.The EU standards showed a distinct improvement in nonattainment, due to the difference in days allowed for exceedance using three-year averaging.Adjusting the South Korean standard to either 96.7 percentile or 68 ppbv (to match US standards) did not achieve a clear change in nonattainment when compared with that of the original South Korean standard.Compatibility with the EU standard required modification of current Korean standard to either 92.3 percentile or 76 ppbv, but the resulting nonattainment displayed significant interannual variability compared to the original EU guidelines.We concluded that the current EU method was the most useful for identifying recent changes in ozone nonattainment in South Korea, which was not discerned by other standards.We suggest a secondary and interim value for Korea ozone standard to assess the progress and effectiveness of ozone control policy, as WHO set the Interim target-1 level of ozone in nations where ozone levels frequently exceeded the guideline.The secondary Korean MDA8O3 standard equivalent to the EU MDA8O3 target would be the proper choice to evaluate the effectiveness of implementing air-quality policy without infringing on the current strict Korean ozone standard to protect the public health.

Fig. 1 .
Fig. 1.Location of the 473 operational air quality monitoring sites in South Korea in 2020.

Fig. 4 .
Fig. 4. Yearly variations in (a) spatial means of daily maximum 8 hr average ozone and (b) daily maximum 8 hr average Ox in selected cities and one province.

Fig. 6 .
Fig. 6.Number of days with MDA8O3>60 ppbv based on (a) yearly South Korean standard and (b) three-year average by EU standard, while (c) three-year mean by fourth-highest value of MDA8O3 in South Korea was used to conduct comparison with US standard.Horizontal lines indicate compliance value for each standard.

Fig. 8 .
Fig. 8. (a) Relative distance (root squared sum) between nonattainment by modified MDA8O3 concentration values under the South Korean ozone standard and US and EU guidelines and (b) yearly variations of nonattainment percentages based on South Korean, US, and EU standards.

Fig. 7 .
Fig. 7. (a) Relative distance (root squared sum) between nonattainment by modified MDA8O3 percentile changes in South Korean ozone standard and those by US and EU guidelines and (b) yearly variations of nonattainment percentages based on South Korean, US, and EU standards.

Table 1 .
Number of operational air quality monitoring stations over time in 17 major cities and provinces in South Korea.