Characteristics of Ground ‐ Level Ozone from 2015 to 2018 in BTH Area, China

: With the ground ‐ level ozone pollution problem increasingly prominent in recent years in China, it is particularly important in basic researches on ozone contamination characteristics. In this study, 13 cities in Beijing ‐ Tianjin ‐ Hebei (BTH) area were examined to determine the characteristics of surface ozone (O 3 ) from 2015 to 2018. Due to the photochemical oxidation of ozone precursors (such as nitrogen oxides and carbon monoxide) along with the presence of sunlight and characteristics of local emission sources, the O 3 and oxidant ( OX) concentrations showed obvious seasonal variation and daily variation. It implicated that the O 3 concentrations reached the maximum during summer. The concentrations of O 3 were higher at daytime than those measured at nighttime. The ozone weekend effect was estimated by the difference and deviation, which exhibited that the difference between weekday and weekend were related to the concentrations of ozone precursors and PM, vehicle emissions, and solar radiation. Moreover, the O 3 concentrations decreased with the increase of other air pollutants by correlation analysis. The ozone pollution was easily formed at light and moderate polluted periods when compared to other air quality levels.

percent of light and moderate polluted periods was 43.5%. The percent of serious and very serious polluted periods was about 6%. However, more researches were about the characteristics of pollution at heavy pollution or specific periods, few researches were about the characteristics of pollution at other air quality levels.
Therefore, the aim of this study is to analyze the temporal and spatial characteristics of O3 and the properties of O3 at different air quality levels and its corresponding influence factors, and discuss the relationship between the changes in ozone and its precursors.

Air Quality Data
Real-time hourly ambient air quality index (AQI) and six air pollutants concentrations (PM2.5, PM10, NO2, O3, SO2, and CO) were downloaded from (http://beijingair.sinaapp.com/), which is the National thousand pieces of hourly ambient AQI and pollutant concentrations data. The missing rates of data are less than 1% among 13 cities in BTH. The average missing rate is 0.41%. These data were carried out by statistical analysis to deal with abnormal values. First of all, the z-score method was used for the series of hourly raw data standardization. If the z-score data met the following three conditions, it would be removed and replaced with the time-interpolated method. The three conditions were as follows: (1) The absolute z-score value is greater than 4 (| | 4), (2) the increment between the current time and previous time is greater than 6 ( 6), and (3) the z-score divided by its central moving average of order 3 is greater than 2 (3 / 2). Further details can be seen in previous studies [12,52].

Definition of Statistical Index
The ozone weekend effect (OWE) were caused by changes in anthropogenic emissions. In order to describe OWE, the difference (W) and deviation (Dev) between weekend and weekday were used in this study. The weekend is Saturday and Sunday, and the weekday includes Monday, Tuesday, Wednesday, Thursday, and Friday. The deviation equations are as follows [36,40,53].
W (1) Dev 100% (2) where is the O3 concentration on weekend; is the O3 concentration on weekday.

Ozone Characteristics
The Significant seasonal variation could be seen in Figures 2 and 4a. The trend was a single peak distribution. The average concentrations of O3 reached a peak value (109.27 ± 11.42 μg•m −3 ) in June in BTH area which met the ambient air quality standards (AAQS, GB 3095-2012) of China. It indicated that the O3 concentrations in summer were much higher than other seasons, which was similar to other previous studies [4,14,38,55]. One of the possible reasons was prone to the occurrence of photochemical reaction with the long sunshine time in summer, which was conducive to the formation of O3 [12,24,56]. Another reason was that NOx was easy to convert to nitrate with the conditions of high temperature and humidity due to the rainfall increasing significantly in summer in BTH area, which should have a decrease in O3 removal by vapor-phase chemistry [57].
Diurnal trends of O3 concentrations were basically consistent with each city in BTH area (in Figure 4b). There was trough at 6:00-8:00, while a peak at 15:00-16:00. The peak values ranged from 88.82 to 115.39 μg•m −3 , which were similar to the sparely populated region and densely populated region of China [14], and higher than Brazil (55.96 μg•m −3 ) [58]. The O3 concentrations at daytime (from 7:00 to 20:00) ranged from 1.28 to 2.25 times than at nighttime (from 0:00 to 6:00 and from 21:00 to 23:00). The analysis results agreed with what had been observed in other studies [18,24,29]. The diurnal trends were a possible result of the increase of photochemical reactions of ozone precursors with the increase of solar radiation from sunrise to noon [12,24]. Therefore, there was an increase in O3 concentrations at daytime.

OX Concentrations
The temporal and spatial changes of OX (OX = NO2 + O3) can be seen in Figure 5. The average OX concentration was 106.15 μg m −3 in this study. By comparison, it was much greater than Canoas (74.23 μg m −3 ) and Esteio (70.64 μg m −3 ) in Brazil [58]. The OX concentrations varied in different months and seasons in BTH area. There were obvious peaks of OX concentrations from May to June. Overall, the trends of average OX concentrations were summer (125.25 μg m −3 ) > spring (118.68 μg m −3 ) > autumn (92.83 μg m −3 ) > winter (87.85 μg m −3 ) in BTH area. Regionally, the average OX concentrations were in order as BD > TS > CZ > XT > HS > LF > SJZ > HD > BJ > ZJK > TJ > QHD > CD. The highest concentration of OX in BD were 1.26 times those monitored at CD. The difference between the two cities was 23.62 μg m −3 , which indicated that the atmospheric oxidation was relatively stronger in BD and TS. Moreover, the diurnal variations of OX concentrations were distinctive ( Figure 6). There was a tendency to ascend from 8:00 and descend at 18:00. The peak values were 118.8-147.93 μg m −3 at 15:00-17:00, which was similar with the results of Tiwari et al. [16] and Notario et al. [24]. The variations of OX concentrations were consistent with O3 concentrations, whereas they were opposite to that of NO2 concentrations (Figure 7a and Table 1). Results of Spearman correlation analysis showed that OX correlated closely with O3 (R 2 = 0.895, p < 0.01), while it had a negative with NO2 (R 2 = −0.291, p < 0.01) in BTH area. Owing to the enhanced photochemical activity during the afternoon, the O3 and OX maximum was obtained [24]. Thus, the tendency of OX concentrations was more affected by diurnal variations of O3 concentrations, which was associated with the previous study [16].

Ozone Weekend Effect (OWE)
The diurnal variations of O3 and other air pollutant concentrations showed similar trends on the weekend and weekday in Figure 7. Owing to non-normal distribution of O3 concentrations, Friedman was applied in this study, implicating that there were significant differences in the concentrations of O3 between weekend and weekday (p < 0.05). The OWE varied in different regions. The absolute value of the O3 differences (W) and deviations (Dev) ranged from 0.05-1.39 μg m −3 and 0.02%-2.41%, respectively ( Table 2). Except in BJ, QHD, BD, ZJK, and CD, the differences (W) and deviations (Dev) of O3 were positive at other observed cities. These pollutant concentrations changed dramatically at different times during the day ( Figure  7). The average hour concentrations of O3 were correlated with OX, whereas they were contrary to NO2, CO, PM2.5, PM10, and SO2. For O3, the deviations were much higher at 7:00-9:00, exhibiting that the O3 concentrations were higher on weekend than weekday, whereas the deviations of NO2, SO2, and CO were relatively low. As previous studies mentioned, O3 concentrations were linked to a lot of ozone precursor emissions, solar radiation, and vertical transport in the boundary layer [1,12,40,55,59]. Jing et al. [60] found that the traffic volume reached an early peak at 7:00-9:00 on the weekday, while on weekend the peak appeared 2 h later. Additionally, the emissions factors of NOx and CO for the vehicle exhaust had two emission peaks at 8:00-9:00 and 17:00-18:00 which were close to traffic volume; and the emissions factors had an obvious difference between weekend and weekday. The deviation of NOx for vehicle emissions and O3 showed similar trends during the day, as illustrated in Figure 7. In addition, the UV radiation also had a difference on weekend and weekday. Wang et al. [40] indicated that the UV radiation was higher on weekend (12.7 W•m −2 ) than on weekday (12.3 W•m −2 ). Therefore, the higher concentration in the morning was most likely due to the reduction of vehicle emissions on weekend. The deviations of O3 were much less than zero at 12:00-17:00, exhibiting that the O3 concentrations were relatively low on weekend, while deviations of PM2.5, PM10, NO2, SO2, and CO reached the peak values. At 0:00-4:00, the deviations of O3 were also below zero, while the deviations of other air pollutants were greater than zero. Moreover, the deviations of OX were much higher than other times. In the whole, the deviations of OX were almost greater than zero, indicating that the photochemical oxidant were higher on weekend. As can be seen from the above, the concentrations of ozone precursors (such as NO2 and CO) were higher on weekend than on weekday at 12:00-17:00 and 0:00-4:00, which resulted in ozone depletion. Moreover, PM would reduce the photolysis frequencies of J(O3) and J(NO2) [61], which showed a reason for lower surface ozone concentrations. PM2.5 and PM10 concentrations were higher on weekend in generally ( Table 2). The differences (W) were 2.81 and 3.11 μg•m −3 for PM2.5 and PM10, respectively. The higher PM concentrations on weekend may be as a result of an increase emission or meteorological condition. This pattern may be responsible for the lower concentrations of ozone on weekend.

Relationships between Ozone and Other Air Pollutants
From the perspective of region, the correlation analysis results were basically the same between the ozone and other air pollutants. The correlation analysis results were slightly different in BJ and ZJK. There was a significant negative correlation among ozone, NO2, and CO in the current complex pollution in BJ. However, the O3 concentrations had a positive correlation with PM10, NO2, and CO in ZJK. It may be due to the higher PM10 concentrations and local meteorological conditions. If the primary components in ambient air are greatly affected by regional transport of air mass, it may undergo photochemical reactions which could accelerate the O3 production process at this location [45].
As a whole, the O3 concentrations had a significant negative correlation with particulate matter, NO2, SO2, and CO concentrations ( Figure 8 and Table 3) in BTH area. Similar results were obtained by He et al. [12] and Wang et al. [29] that demonstrated the O3 was significantly negatively correlated with its precursors, such as NO2 and CO. Furthermore, aerosols reduced the actinic flux photolysis rates under absorption and scattering, so the surface ozone production decreased [44,38,61]. Nishanth et al. [45] elucidated that it was a negative correlation between daily values of PM10 and O3 concentration. Li et al. [62] demonstrated that O3 concentration was strongly inhibited under the condition of high PM2.5 concentration using the daily PM2.5 and O3 monitoring data.   Table 3. Spearman correlation coefficients between O3 and other air pollutants.  According to the O3 concentrations distribution (Figure 9), over 90% of monitoring data were up to the concentration limits of AAQS, respectively. However, the differences of O3 concentrations were obvious at different air quality levels (Figures 10 and 11). The O3 exceeding standard rate were mainly in the order as light polluted periods (16.27%) > moderate polluted periods (12.38%) > good (3.76%) > severe polluted periods (2.50%) > very severe polluted periods (0.83%). In addition, it was zero at excellent. Furthermore, the maximum of O3 exceeding standard rate was 32.47% at ZJK at light polluted periods. The average rates were less than 6% at good periods and severely polluted periods. Overall, the O3 exceeding standard rates were much higher at light and moderate polluted periods than at other air quality levels.  Except ZJK and CD, the O3 concentrations were mainly up to the highest at light polluted periods in BTH area, which ranged from 63.67 to 87.21 μg•m −3 . Secondly, the O3 concentrations were much higher at excellent, good, and moderate polluted periods, ranging from 38.13 to 76.29 μg•m −3 . While the O3 concentrations were much lower at severe and very severe polluted periods, which were less than 40 μg•m −3 . The O3 concentrations had the highest values (109.16 μg•m −3 ) with the highest O3 exceeding standard rate among thirteen cities at light polluted periods at ZJK, while the PM2.5 concentrations were relatively lower. Moreover, the O3 concentrations at ZJK were basically much higher than other cities at different air quality levels, especially at severe and very severe polluted periods. For CD, the O3 concentrations were much higher at light polluted periods, moderate polluted periods, and very severe polluted periods. Additionally, the PM2.5 concentrations were relatively lower. It was more likely to form an ozone pollution in ZJK and CD, especially at light polluted periods.

Ozone Pollution Characteristics at Different Air Quality Levels
Although PM was the primary pollutant in severe and very severe pollution events in BTH area, there is a shift tendency from PM to O3 pollution, especially at light and moderate pollution events. Therefore, the direction and measures for air pollution control should be adjusted accordingly.

Conclusions
This research mainly discussed the characteristics of O3 in 13 cities of BTH area from 2015 to 2018. For O3, it varied throughout temporal and spatial. The O3 concentrations were much higher in summer due to increased photochemistry reaction and low ozone destruction rate. The diurnal trends were unimodal distribution with maximum values at 15:00-16:00. There was a lower O3 concentration on weekday and a higher concentration on weekend in BTH area. In addition, the difference (W) and deviation (Dev) were applied to elucidate the ozone weekend effect. These results indicated clearly that vehicle emissions, PM concentrations, and solar radiation also contributed to the difference between weekday and weekend except the ozone precursors. Geographically, O3 concentrations showed similar seasonal and diurnal variation trends. However, it was much higher in ZJK and CD.
In addition, the OX concentration also showed seasonal and diurnal variations, which agreed with the trends of O3 concentrations. Additionally, a positive correlation was found between OX and O3 by Spearman analysis, whereas it was negative between OX and NO2. The O3 concentrations had a significant negative correlation with other air pollutants. The aerosol effect was found to decrease O3 concentration because aerosol effects reduced J (NO2) photolysis frequencies. The O3 exceeding standard rate and O3 concentration generally reached the highest values at light polluted periods. Based on the region distribution, the values were much higher at light polluted periods at ZJK and CD than other cities along with lower PM2.5 concentrations, which indicated that the formation of ozone pollution more easily happened at the two cities.