Analysis of High-Concentration PM2.5 Episodes during Winter 2019–2020 in Seoul, Korea

PM2.5 is a WHO-designated first-class carcinogen and coping with high-concentration situations with high human risk is becoming more important. In particular, Korea has a high concentration of PM2.5 in winter due to its geographical characteristics, which can be largely divided into foreign inflows and domestic atmospheric stability. To determine this, wind patterns and air pressure data were analyzed representatively and episodes about high concentration phenomena were classified. In this study, high-concentration PM2.5 episodes, where the daily average PM2.5 concentration in Seoul exceeded 35 μg/m3 between October 2019 to March 2020, were analyzed case-by-case. The criteria for the separation of consecutive episodes were established. Then, the designated episodes were classified into four categories based on cause: atmospheric stagnation, combination of causes, penetration from abroad, and accumulation. To identify the causes of the episodes, wind direction, wind speed data, wind roses, and air quality forecast modeling data obtained from “Air Korea” were studied. Fifteen episodes were identified and analyzed and each were given a classification type. Furthermore, the phenomenon of high-concentration episodes was summarized after detailed individual analysis of the episodes. As a result of case analysis, just before there was an inflow from abroad due to strong wind speed, a characteristic of low PM2.5 concentration of air quality as a kind of cleaning effect could be found. In addition, alarm-level PM2.5 concentrations of 75 μg/m3 or higher were often made by external inflow. This will contribute towards identifying the main causes of high PM2.5 concentration episodes in Korea when it is applied over a longer time period.


INTRODUCTION
High concentrations of PM 2.5 can be caused by foreign as well as domestic factors.Domestic factors include the local weather and industrial sources of emission.For efficient management, it is essential to identify recent instances of high-concentration PM 2.5 episodes and analyze their causes through classification.
Episodes occurring during Korea's winter or spring season are related to continental high pressure, which has a dominant impact on the Asian continent from mid-October to late March.If high-concentration PM 2.5 in winter lasts for a long time in a stable atmosphere, it can be extremely dangerous for the elderly and for patients with respiratory diseases.
In this study, high-concentration PM 2.5 episodes in Seoul were analyzed and classified based on their causes.The concentration of PM 2.5 in Korea is primarily related to continental high pressure, which prevails in the winter and spring months from mid-October to late March (Lee and Hills, 2003).However, more recently, cases of highconcentration PM 2.5 originating in China undergoing long-range transport over to Korea by winds and atmospheric diffusion have also occurred frequently (Lee and Hills, 2003).
After the WHO designated PM 2.5 as a first-class carcinogen in 2013, its atmospheric concentration became a major social concern.PM 2.5 is reported to be the pollutant that is most harmful to the human body (Burnett et al., 2014;WHO, 2014).PM 2.5 is also one of the major causes of visual impairment in urban areas.Research is currently being conducted on other health risks it poses, such as respiratory diseases like asthma and reduced lung capacity.A recent study showed that the risk posed by PM 2.5 to human health is relatively high compared to other pollutants, especially in large cities such as Seoul, Incheon, and Busan where traffic is concentrated and the number of construction activities is relatively high, drawing more attention to the need for PM 2.5 management (Shin et al., 2006).Recently, the daily average environmental standard of PM 2.5 has been strengthened from 50 μg/m 3 to 35 μg/m 3 , and management measures for PM are also taking shape.
It is difficult to improve the atmospheric environment in Seoul, given the high concentration of population and cars within such a small area.Seoul's atmospheric environment is affected not only by local emissions from Seoul and the metropolitan area but also by external influences such as emissions from China.In addition, there are noticeable changes in concentration due to interactions between local air pollutant emissions and weather conditions such as humidity according to an analysis of high concentrations of PM 2.5 in Seoul during July, 2005 (Lee et al., 2011).
In this study, high-concentration PM 2.5 episodes, where the daily average PM 2.5 concentration in Seoul exceeded 35 μg/m 3 , were defined by analyzing PM measurement concentrations and meteorological factors in Seoul from October 2019 to March 2020.The character-istics of high-concentration episodes were then examined case-by-case through the process of classification by cause according to certain criteria.

1 Data
The period chosen for analysis was the winter season from October 2019 to March 2020.Air pollution measurement data for 25 districts in Seoul from National Institute of Environmental Research (NIER) were utilized.Weather information data (wind speed and wind direction) provided by the Korea Metropolitan Administration (KMA)' Automated Synoptic Observation System (ASOS) in Seoul were also applied.Forecasting data and analysis weather maps were also used to analyze the causes.

1 Episode Period Definition
The analysis was based on hourly measurement data, it was necessary to define the start and end of high-concentration episodes.Initially, we tried to classify episodes according to the cause of the occurrence, but such a method was unsuitable for episodes associated with multiple causes.If there is a physical time lag between episodes, the cause is likely to be different.Therefore, we decided to use the PM 2.5 concentration value, which is an objective index, over time to distinguish between episodes.Episodes were separated when the PM 2.5 concentration was maintained at less than 33 μg/m 3 for more than 5 hours.The designated value of 33 μg/m 3 was seen to be more flexible than the current Korean PM 2.5 daily air quality standard of 35 μg/m 3 for our data set.This may obviously differ depending on the atmospheric conditions.Also, in this study, the weather conditions most frequently lasted about 5 hours, so this was adopted as a classification criterion.The classification was based on a measured 1 hour concentration unit.If a concentration of 33 μg/m 3 or less lasted more than 5 hours, the previous and later episodes were divided as separate episodes.When episodes occurred "back-to-back" but were separated by "normal" periods of at least 5 hours or more, their designation utilized a sub-numeral such as n-1, n-2, etc. to indicate the adjacency of the episode.

2 Classification Method
Given the limitation of the material available in real time, we have classified episodes based on the decisiontree flow chart in Fig. 1.The major primary weather factors were identified through analysis, and wind patterns and air pressure data were found to be the two main factors influencing the particulate concentrations.However, it is noted that there may be effects from other weather factors that were not considered in this study.The two dominant resulting phenomena, inflow from countries other than the Korean Peninsula and accumulation due to atmospheric stagnation, are the main factors that drive the decision-making.Many different iterations were attempted before settling on this format.The episodes were finally classified into five distinct episode types based on their characteristics with respect to these two main causes.According to the KMA's regional detailed observation data (collected by the Automatic Weather System (AWS)), the pollution level increased around the eastern part of the metropolitan area during this period, so it was not affected by inflows from abroad.In addition, wind   Frequency of counts by wind direction (%)

1 Episode Analysis
speeds were generally found to be stagnant, less than 2 m/s.A strong east wind blew after 2:00 pm on Nov. 3, ending the episode.

1. 2 Episode 2 (10-11 Dec. 2019)
Episode 2 is classified as an Interference Episode.With the concentration of PM 2.5 generated in Korea rising because of the stable atmosphere, an alarm-level concentration situation occurred when the inflow began.
There was continuous atmospheric congestion and associated high PM 2.5 concentrations from December 7 to December 11 until about 1:00 pm.There were short periods of 'normal' concentrations but PM 2.5 from abroad was introduced on Dec. 10, as can be seen by the wind direction (south-west wind).It is believed that PM 2.5 , which was introduced after 1:00 pm on Dec. 10, accumulated in Korea due to repeated atmospheric congestion (Fig. 4).This can also be seen from the shaded wind speed data of Table 2.The atmospheric congestion was maintained until 6:00 pm on December 11, and the episodes were resolved because there were no additional inflows from abroad.Episode 3 is classified a Compound Factor Episode, as it features transitions from stable atmosphere -inflowstable atmosphere.
From 9:00 am on December 22 to 6:00 am on December 23, air congestion continued at low wind speeds, resulting in PM 2.5 accumulation (Fig. 5).From 9:00 am on the 23 rd , the concentration of PM 2.5 reduced slightly because of strong wind speed, but the inflow of high-concentration PM occurred owing to the flow of northwest  winds.This is also confirmed by the wind speed data shaded in Table 3.After that, atmospheric congestion occurred again by the 25 th , and domestic PM 2.5 began to accumulate.The re-formed atmospheric congestion lasted until December 25 th , with a high concentration of PM 2.5 being maintained (Fig. 7).The cause of high PM 2.5 concentration within the period changed twice within a continuing episode, so the episode was defined as a compound factor type.

1. 4 Episode 4-1 (2-4 Jan. 2020)
Episode 4-1 is classified as an Accumulation Episode.Episode 4-1 seems to have started at 11:00 am on the same day after the concentration slowly rose from 3:00 am on January 2. Furthermore, from 1:00 pm on January 2, the east wind began to change to the west wind, and the trend continued until 7:00 pm on the 2 nd .However, from 8 pm, the wind speed weakened to 0.7 m/s, resulting in atmospheric stagnation, which continued until 3:00 am on January 3. Also, as can be seen from the shaded portion of Table 4, the main wind direction is northward.After that, the trend continued to rise and fall, but as the wind speed rose sharply to 4 m/s at 14:00 on the 4 th (Fig. 8), the accumulated concentration was resolved, and Episode 4-1 ended.At the time of Episode 4-2's occurrence, the wind speed was definitely slower than before and after the episode, which can be seen through the wind speed distribution chart below.In particular, most parts of Seoul showed stronger wind speed at the end of Episode 4-2 (Fig. 10).Therefore, Episode 4-2 can be judged to have been caused by atmospheric congestion and was resolved as the wind speed increased.
Episode 5 began when high concentrations of PM 2.5 from China flowed in owing to the west-northwest wind.According to observations in Jung-gu, Seoul, air congestion in areas near Seoul began from 9:00 pm on Jan. 9 to 12:00 pm on Jan. 10, with the average hourly wind speed not exceeding 1.5 m/s.The low wind speeds are shaded in Table 6.From 1:00 pm on January 10 th , the inflow of high concentrations from abroad began again with the west-northwest winds and continued until 7:00 pm on January 11.Since then, the wind direction did not change, but the episode ended owing to rapid wind speed enhancement (7:00 pm average: 1.4 m/s → 8:00 pm average: 4.6 m/s).

1. 7 Episode 6-1 (17-18 Jan. 2020)
Episode 6-1 is classified as a Stable Episode under the influence of high pressure system over the Korean Peninsula.
The weather map was checked from the 16 th , the day before the 17 th high concentration period, to check the weather pattern.The transient high pressure produced from the Siberian high pressure and the deterioration of nearby systems was evident from the 16 th .Usually, during the winter, the transient high pressure from China moves toward the Korean Peninsula carrying with it China's PM 2.5 , and when the transient high pressure settles down, there is very little wind which prevents the dilution of PM 2.5 (Fig. 14).In this episode, the transient high pressure was located near the west coast and above the Kore-  an Peninsula without moving into the Korean Peninsula, making it difficult to determine the presence of significant inflow from just this data.However, the gap between the isometric lines on the 17 th shows that the wind is weak overall, so this should result in atmospheric congestion.
Episode 6-2 was initially analyzed by checking the 1/19 wind speed distribution chart and forecast data.
The wind speed data showed stronger winds from 1:00 pm on the 19 th compared to the two previous days, and it was mostly westerly winds.By checking the situation of PM 2.5 coming from abroad through the forecast data (Fig. 15), it was quite easy to see that Episode 6-2 was caused by a influx of high concentration PM 2.5 from abroad owing to strong westerly winds.

1. 9 Episode 7 (23-25 Jan. 2020)
Episode 7 is classified as a Compound Factor Episode with a stable atmosphere -inflow -stable atmosphere sequence.In Episode 7, PM 2.5 was accumulated in Korea because of atmospheric congestion since the morning of January 23.Afterwards, inflows from abroad were observed on the afternoon of the 23 rd .Based on air quality measurements, PM 2.5 concentrations in the Seoul metropolitan area and Chungcheong Bay were found to increase first (Fig. 17).On January 25, PM 2.5 introduced the previous day from China accumulated in Korea, and there were high concentrations of PM 2.5 in the western half of the country.
During Episode 8, high concentrations occurred owing     to the accumulation of PM 2.5 inflow from overseas as well as domestic PM 2.5 .Based on the air quality concentration forecast on the 31 st , high concentrations of PM 2.5 were moving towards the metropolitan area and Chungcheong Province because of the northwest wind.On January 31, when external inflows were expected, PM 2.5 concentrations in the metropolitan and Chungcheong provinces were generally higher than in other regions.
On the morning of February 2, air congestion also occurred, causing accumulation of both PM 2.5 inflows and domestically generated PM 2.5 .
In Episode 9, from the afternoon of the 11 th , high con-  centrations of PM 2.5 came down from the north.Thereafter, high concentrations remained until the morning of the 12 th owing to atmospheric congestion.On the 12 th , precipitation occurred in the Seoul area, and the concentration of PM 2.5 decreased as the low-pressure base passed through the Korean Peninsula.
In Episode 9-2, there was a slight inflow of PM 2.5 during the morning of the 13 th , but long-term atmospheric stagnation led to high concentrations for three days.From the 16 th , the wind speed increased and the PM 2.5 decreased rapidly because of rain.

1. 13 Episode 10-1 (20-21 Feb. 2020)
Episode 10-1 is classified as a Stable Episode.Since the afternoon of the 19 th , the entire Korean Peninsula was affected by a high-pressure zone, causing the  Episode 10-2 is a simple inflow type.There was a high concentration of external inflow due to the strong west wind.It was also quickly resolved by the fast wind speed (Fig. 25).Episode 11 is classified as a Stable Episode.
In Episode 11, atmospheric stagnation continued because of the high-pressure effect throughout the episode, for three days.Fig. 26 shows that the Korean Peninsula is under relative high pressure.In addition, looking at the wind speed data, shaded in Table 15, the overall low wind speed of 1.5 m/s or less can be confirmed.Low pressure was generated from the afternoon of the 10 th  and the high concentration situation was resolved (Fig. 26).

2 Episode Classification Result
Table 16 summarizes the episodes classified earlier.High-concentration PM 2.5 in Seoul in the winter of 2019-2020 was mainly caused by atmospheric stagnation.

DISCUSSION
Seoul is the capital of South Korea and has a population of approximately 10 million.Management of air pol-  lution is crucial as the population is readily exposed to its ill-effects.Understanding the causes of high levels of pollution can be a good basis for establishing management policies.
In this study, we determined the frequency of highconcentration PM 2.5 cases in Seoul by using a classification system which is a function of cause and sequence.Atmospheric stagnation and inflows from abroad were recognized as the major causes.Individual episodes were identified and classified based on the results of weather maps and forecast models.Most episodes turned out to be high-concentration cases caused by atmospheric stagnation.On the other hand, sudden increases in concentration that do not follow general trends can sometimes be seen due to the effects of rapid external inflows and individual internal sources (Ji et al., 2012;Lee and Hills, 2003).
Episodes featuring overseas inflows tended to show a decrease in PM 2.5 concentration after a short time period because of strong west winds associated with the inflow.The influx of strong winds had the dual effect of polluting and then cleaning the air, resulting in short spurts of high concentration in Seoul.In addition, in the case of high-concentration episodes due to influx from abroad, there were many cases where seasonal anticyclones were present.
However, alarm-level PM 2.5 concentrations of 75 μg/ m 3 or higher are more frequent for episodes that involve external inflows.These extremely high concentrations of air pollution could be more readily observed in inflow episodes than in non-inflow episodes.On the other hand, the influence of other unknown factors were large  enough that the change in wind speed did not affect the concentration in certain situations where the inflow was apparent from abroad.This study has certain limitations and is not meant to be a comprehensive thesis on high PM 2.5 concentration episodes in Korea.Only a relatively short period of time is considered and changes in emissions associated with the changes brought about by the coronavirus pandemic have not been considered.However, the classification methodology introduced in this paper will be tested with more recent data and through accumulation of more data and subsequent analysis we hope to better understand the complex PM 2.5 high concentration phenomena which is currently plaguing our country through continued observations of short-and long-term characteristics of these episodes.

Fig. 10 .
Fig. 10.Wind information and air pollution data near Seoul at 6:00 am on 5 Jan.

Fig. 26 .
Fig. 26.Wind information and air pollution data near Seoul during Episode 11.

Table 1 .
Hourly concentration and meteorological factors of Episode 1.

Table 2 .
Hourly concentration and meteorological factors of Episode 2.

Table 3 .
Hourly concentration and meteorological factors of Episode 3.

Table 4 .
Hourly concentration and meteorological factors of Episode 4.

Table 6 .
Hourly concentration and meteorological factors of Episode 5.

Table 9 .
Hourly concentration and meteorological factors of Episode 7.

Table 10 .
Hourly concentration and meteorological factors of Episode 8.

Table 15 .
Hourly concentration and meteorological factors of Episode 11.