Dual-height Distribution of Ozone and Nitrogen Oxides during Summer in Urban Tianjin: An Observational Study

Measurements of gaseous pollutants, including ozone (O3) and nitrogen oxides (NOx), were simultaneously conducted at 220 m (via the installation of an air flow drainage system on a 255-m meteorological tower) and 3 m above the ground in urban Tianjin during summer 2018. The observed O3 concentrations at the two altitudes exhibited similar diurnal variations but distinctly different values, with higher levels near the surface during the day and the opposite trend at night. Generally higher concentrations of NO and NO2 were found at 3 m than 220 m, and the difference in concentration between the two altitudes for the latter pollutant was smaller during daytime and highest at night. Ox (O3 + NO2) concentration near the surface during the day, but the difference was negligible at night. Based on the higher NOx level at 3 m, the photochemical production of O3 (Ox) at low altitudes intensified during the day, suggesting that the O3 surface concentration was mainly influenced by local photochemical production. Additionally, by measuring the reactive nitrogen (NOy) near the surface and calculating NOz (NOy – NOx), the ozone production efficiency (OPE; Ox/NOz) in urban Tianjin was assessed for the first time and determined to be 6.0 ± 0.4. Compared to the values measured during summer 2010, lower levels of NOx but significantly higher ones for O3 were observed during the same season in 2018.


INTRODUCTION
Tropospheric ozone (O 3 ) exerts a large impact on air quality, environment, and climate change (Tang et al., 2006). Along with rapid economic growth and urbanization, emissions of gaseous pollutants including O 3 precursors have largely increased in China. As a result, O 3 pollution has become increasingly serious at the urban and regional scales. In the densely populated and economically developed regions of China, such as the Yangtze River Delta, Pearl River Delta, and Beijing-Tianjin-Hebei region, O 3 has become a major air pollutant during summer and autumn (Lam et al., 2005;Wang et al., 2006;Ding et al., 2008;Ran et al., 2009;Lu et al., 2010;Tang et al., 2011).
Tropospheric O 3 is mainly formed by photochemical reactions of its precursors, nitrogen oxides (NO x ), and volatile organic compounds (VOCs) in the presence of sunlight (Haagen-Smit, 1952;Sillman, 2002;Zou et al., 2015). Variations in surface O 3 and NO x concentrations are not only influenced by precursor emissions and photochemical processes but also by transport processes and boundary layer evolution. The advective transport of air pollutants is influenced by the dominant wind direction and topography. Vertical exchange of air pollutants includes both downward transport from high altitude and upward turbulent mixing from the ground. Information on the vertical exchange of air pollutants only based on measurements near the surface can be difficult to obtain. Therefore, obtaining vertical measurements is necessary to understand the characteristics of O 3 more thoroughly and its precursors within the boundary layer.
Given the limitations of the current techniques, continuous online monitoring of the vertical distribution of air pollutants within the urban boundary layer is usually based on meteorological towers. Some studies on the vertical profiles of air pollutants such as O 3 , NO x , carbon monoxide (CO), sulfur dioxide (SO 2 ), particulate matter (PM) and peroxyacetyl nitrate (PAN) in the North China Plain have been based on either the 325-m meteorological tower in Beijing (Liu and Hong, 2000;Meng et al., 2002;An et al., 2003;Ding et al., 2005;An et al., 2007;Ma et al., 2007;Sun et al., 2010) or the 255-m meteorological tower in Tianjin (Han et al., 2009;Huang et al., 2009;Sun et al., 2010;Qiu et al., 2019;Yao et al., 2019). A difference in the concentrations of air pollutants at different heights was noted within the boundary layer. However, due to the limited space and uncontrollable conditions on the tower, the online instruments placed at different heights of the towers are usually not sufficiently maintained to ensure adequate data quality. Additionally, data quality is also restricted by the fluctuation of instrument response with time and the parallelism between instruments. Passive sampling devices are relatively simple, but present low time resolution as a drawback.
Tianjin is located in the North China Plain, where tropospheric O 3 is found to be at a considerably high level . O 3 pollution has become more severe during the past decades Lu et al., 2019). Urbanization, economic development, and rapid increase in the number of vehicles (more than 2.88 million in 2018 in Tianjin) caused significant anthropogenic ozone precursor emissions. Therefore, conducting long-term observations of O 3 and its precursors at different heights in Tianjin area is crucial. Such studies can provide a scientific basis for taking reasonable and effective measures to mitigate O 3 pollution. In this study, an air flow drainage system at a high altitude was built on the 255-m meteorological tower in urban Tianjin. Online instruments on the ground were able to sample air from high altitudes almost without any loss. Concentrations of O 3 and NO x were simultaneously measured at a height of 220 m and 3 m above ground level to study the characteristics of these gases at different heights.

METHODS
The observational site is located in the yard of Tianjin Atmospheric Boundary Observatory (TABO) of the China Meteorological Administration (39°06ʹN, 117°10ʹE, 2.2 m a.s.l.), which is in the southern part of urban Tianjin. It is surrounded by residential and commercial areas and is approximately 100 m away from a freeway to its north. An air drainage system ( Fig. 1) with the inlet placed at a height of 220 m above ground level was situated on the 255m-high meteorological tower in TABO. Air from a height of 220 m could be drained into a Teflon tube and eventually entered the online instruments in an air-conditioned room on the ground. The air drainage system mainly contains a large-caliber Teflon tube (Shanghai Huzhuang Rubber Plastic Products Co., Ltd., China) with an I.D. of 3.2 cm and an O.D. of 3.6 cm and a rotary vane pump (GAST1423-101Q-G626X; Gast Manufacturing, Inc., USA) with a free air flow of 20 m 3 h -1 under an air pressure of approximately 1 bar. To prevent the entry of rain and insects into the pipeline, a protection cover was equipped in front of the inlet. The Teflon pipe was wrapped using aluminum foils to shield it from sunlight. The retention time of the air in the drainage system was less than 35 s. An air flow distributor with eight branch pipes was installed between the pipe and the pump. Through the branched pipes, air could be sampled and analyzed using different instruments. A pressure meter probe was also placed in the distributor to monitor the air flow pressure. Difference in pressures between the inlet (at 220 m high) and the distributor was less than 20 hPa. Condensed water (if any) in the buffer bottle could be automatically drained off by a little pump (Shenzhen Sypda Technology Co., Ltd., China) to prevent water from accumulating in the pipe and entering the analyzers. The rotary vane pump was protected by a high-volume particulate filter. Using a time controller and 3-way solenoid valve, air at 220 m, surface air, zero gas, and span gases could be subsequently switched in the analyzers. Accordingly, gas pollutants at different heights could be conveniently measured near the surface. Instrument calibration and maintenance could also be easily performed on the ground. This system helps in overcoming the space limit of the tower.
Field observations of O 3 , NO 2 , and NO at two heights were obtained using this air flow drainage system from June 1 to August 31, 2018. The air-sample collection was switched every 15 min between 220 m and 3 m. Reactive nitrogen (NO y ) was only measured near the surface because NO y compounds other than NO must first be transformed into NO before they can be measured using chemiluminescence. They were converted to NO by using a molybdenum converter heated to approximately 375°C. NO y includes all reactive oxides of nitrogen (i.e., NO, NO 2 , NO 3 , N 2 O 5 , HNO 2 , HNO 3 , PAN, organic nitrates, and aerosol nitrates). To minimize the loss of NO y prior to measurement, an external molybdenum converter was used to limit sample transport time and surface contact area. Thus, the external converter was only deployed near ground level. Data were recorded every 1 min. Information on the gas analyzers is presented in Table 1.
Three aspects of quality control were considered during the measurement: (1) To determine the effect of gas loss in the pipeline, O 3 loss test was performed. From August 16 to September 30, 2017, O 3 concentrations at the inlet and outlet of the pipeline were tested in parallel by using two O 3 analyzers. Comparison between measurement results (Fig. 2) from two instruments revealed the negligible loss of O 3 in the pipeline.
(2) Standard gases were used to calibrate the instruments.
The O 3 analyzer was calibrated using an O 3 calibrator (Model 49i-PS; Thermo Fisher Scientific, USA). An NO/N 2 mixed reference gas (Beijing Huayuan Gas Chemical Co., Ltd., China), gas dynamic calibrator (Model 146i; Thermo Fisher Scientific, USA), and zerogas generator (Model 111i; Thermo Fisher Scientific, USA) were used for multipoint calibration of the NO y instrument and for the NO 2 analyzer through the gasphase titration method. (3) Data correction and rejection were performed. Original data was corrected using multipoint calibration results. Each time the sampled air was switched from one height to another, a balancing time of 2 min was required, and the corresponding data was eliminated. The corrected data were further processed into hourly averages for subsequent analysis.     Table 2 presents the descriptive statistics of O 3 , NO 2 , NO, and O x concentrations at two sample heights. A one-to-one correspondence between the data for the two heights was adopted to improve the reliability of the comparison.

Variations in O3, NO, and NO2 at Different Heights
The statistical analysis of O 3 concentrations at different heights is also presented in Fig. S1. Surface O 3 level was greater than that at high altitude toward the higher end of the concentration distribution (e.g., the 95% highest value was 109.2 ppb near the surface and 104.1 ppb at 220 m). By contrast, surface O 3 level was lower than that at high altitude toward the lower end of the concentration distribution (e.g., the 5% lowest value was 10.3 ppb near the surface and 15.8 ppb at 220 m). The average (53.2 ± 30.6 ppb) and median (48.8 ppb) values of surface O 3 concentrations were slightly lower than the average (53.8 ± 27.7 ppb) and median (49.9 ppb) values at 220 m. The difference between the daily mean O 3 concentrations at 3 m and 220 m ranged from -18.0 ppb to 12.0 ppb. Variations in daytime O 3 concentration were mainly influenced by processes such as transport (advection or vertical mixing), photochemical reaction, and deposition. High O 3 concentrations were mainly noted during the afternoon, and low O 3 concentrations were mainly noted during the night. Surface concentration was lower than that at 220 m during the night, indicating that near-surface O 3 may be consumed through NO titration more easily than that at 220 m (Han et al., 2009). This result differs from that reported by Sun et al. (2010), who measured O 3 concentration at three heights (40 m, 120 m, and 220 m) in the same tower from August 18 to September 22, 2006; they found that O 3 concentration always increased with height. The difference in the vertical distributions between 2006 and 2018 may have been due to the substantial changes in population, urbanization, and pollutant emission in Tianjin.
Figs. 4 and 5 display the time series of hourly mean NO 2 and NO concentrations at different heights during the observational period. Obvious diurnal variations were observed, with a higher level during the night than in the daytime. NO peaked almost at the same time at different heights, whereas NO 2 peaked at different times, with a later appearance of the maximum concentration at 220 m. According to the statistical results shown in Table 2 and Fig. S2, NO 2 concentration near the surface was significantly higher than that at 220 m for all conditions. NO concentration near the surface was obviously higher than that at 220 m toward the higher end of the concentration distribution, but remained comparable to that at 220 m toward the lower end of the concentration distribution.
The average value of the surface NO 2 hourly mean (±1 standard deviation) was 15.36 ± 8.15 ppb, with a median value of 14.17 ppb. The mean value of NO 2 at 220 m was 10.98 ± 6.34 ppb, with a median value of 9.40 ppb. The average value of the surface NO was 1.65 ± 3.22 ppb, with a median value of 0.92 ppb. The average value of the highaltitude NO was 1.13 ± 1.40 ppb, with a median value of 0.86 ppb. A two-sided reduced major axis regression was conducted for NO 2 and NO at different heights. The slopes of NO 2 /NO were 2.35 ± 0.07 (R 2 = 0.156, P < 0.05) near ground level and 4.45 ± 0.12 (R 2 = 0.244, P < 0.05) at 220 m. Higher NO 2 /NO values indicated more photochemically aged air mass at 220 m. The average value of surface NO was significantly higher than the median value and considerable amounts of newly emitted NO were observed near ground level, suggesting the effects of local pollution. Fig. 6 depicts the average diurnal variations of O 3 , NO 2 , O x (O 3 + NO 2 ), and NO at 3 m and 220 m during the observational period. In general, similar diurnal variations were found at different heights for each of the gases.

Diurnal Variations
O 3 increased rapidly after sunrise and maintained a high value between 12:00 and 16:00. Thereafter, the concentration   began to decline until the early hours of midnight. During the day (08:00-17:00), O 3 concentration was higher near ground level than at 220 m (the average difference was 4.0 ppb, and the maximum difference was approximately 6.0 ppb; Fig. 6(a)). The daytime variations of vertical O 3 distribution differed from those reported in previous studies. Further exploration involving measurements of VOCs is warranted. At night (21:00-06:00), surface O 3 concentration was lower than that in the upper level (an average difference was 7.0 ppb and the highest was close to 9.0 ppb), although the difference seems less than that in 2006 (30-50 ppb; Han et al., 2009;Sun et al., 2010). During early morning (around 07:00) and in the evening (17:00-20:00), O 3 concentrations at different heights were nearly the same. As seen in Fig. 6 concentration during the daytime. The diurnal variation in NO 2 concentration ( Fig. 6(b)) shows morning and evening peaks around 07:00 and 19:00, respectively. This was particularly evident at 220 m. Surface NO 2 concentration remained relatively high after 19:00. NO 2 concentration gradually decreased before noon and reached the lowest during 12:00-16:00, contrary to the variation in the O 3 concentration. In general, NO 2 concentration near the surface was higher than that at 220 m, but the two were comparable during the day. The difference between NO 2 concentrations at different heights increased during the night, with the maximum exceeding 9 ppb, possibly due to the accumulation of emitted NO 2 near the ground and weak vertical mixing.
From Fig. 6(d), the diurnal variation of NO is noted to be characterized by a unimodal type, with a peak during 06:00-10:00 and a higher level at night than in the day. The time when NO concentration peaked during early morning was slightly different for different heights. The concentration of NO near the surface was higher than that at 220 m, with the largest difference during the day and the smallest at night. Therefore, O 3 was significantly affected by chemical titration of NO at night, especially near the surface, leading to a much higher level at 220 m than near the surface. In 2006, the average diurnal change of NO x is with an evident peak around 14:00 (Sun et al., 2010), but this afternoon peak is not found in 2018.
In cities, O x (O x = O 3 + NO 2 ) is usually used to represent the total oxidant in the atmosphere. Being more conservative than O 3 , O x can better characterize the photochemical process, as rapid NO x -O 3 cycle (Liu, 1977;Nunnermacker et al., 1998) need not be considered. As seen in Fig. 6(c), O x showed similar variation as O 3 , except the small difference in O x at different heights during the night (especially from midnight to early morning). The concentration increased rapidly after sunrise and reached the maximum at approximately 16:00. The peak of the mean concentration was 87.6 ppb at 220 m and 92.6 ppb near the surface. During the day, the O x concentration was considerably higher near ground level than at 220 m, with the maximum difference exceeding 6 ppb. As shown in Table 2, for the 1550 pairs of hourly mean O x concentrations, the average (69.4 ± 27.0 ppb), median (65.1 ppb), minimum (14.9 ppb), and maximum (185.1 ppb) values of surface O x concentration were 3-4 ppb higher than the average (65.9 ± 27.0 ppb), median (62.8 ppb), minimum (11.4 ppb), and maximum (181.3 ppb) values at 220 m. Therefore, total oxidant production was considerably stronger near ground level than at 220 m during summer in urban Tianjin.
According to the aforementioned analysis, the concentration of O 3 precursor, NO x , was higher near the surface than at 220 m in urban Tianjin. During the day, photochemical generation of O 3 (O x ) was greater than that at 220 m. Thereafter, the rapid increase in the O 3 concentration was mainly affected by local photochemical formation and not by downward mixing of the residual layer. From 07:00 to 14:00, O 3 concentration on an average increased to approximately 50 ppb near the surface and 46 ppb at 220 m. Fig. 7 depicts the time series of O 3 and NO 2 hourly mean concentrations during summer at the same site in Tianjin in different years (2010 and 2018); the observation data in 2010 are available in Ran et al. (2012). The measurements in 2010 and 2018 were performed at the same location. O 3 concentrations at both times were measured using UV photometric O 3 analyzers. These analyzers were calibrated using a primary standard UV photometric O 3 calibrator, which is traceable to the standard reference photometer maintained at the World Calibration Centre of the World Meteorological Organization in EMPA, Switzerland. NO analyzers were operated according to the principles of chemiluminescence. In contrast to the direct measurement process implemented in 2018, that adopted in 2010 involved first converting NO 2 into NO by using a molybdenum converter heated to approximately 325°C and then performing measurements using chemiluminescence. The NO/N 2 standard mixtures for in situ calibrations were compared with the same NO standard produced by Scottgas (https://industry.airliquide.us/scott-gas-mixtures), which can be traced to the National Institute of Standards and Technology (Lin et al., 2011). NO was converted into NO 2 through gas-phase titration for NO 2 calibration. Therefore, comparison of data from 2010 and 2018 was sound. Because insufficient data prevented meaningful statistical analysis (only 381 hourly overlapping NO observations), NO and NO x were not compared for these periods.

Comparison of O3 and NO2 Concentrations between 2010 and 2018
Compared with the data in 2010, NO 2 concentrations in 2018 were significantly lower, whereas the range of O 3 concentrations was similar but with higher O 3 peaks in 2018. From the statistical data (Table 2 and Fig. S3), hourly mean NO 2 concentration was found to be 15.70 ± 0.27 ppb and the median value was 14.58 ppb in 2018, whereas the hourly mean was 25.51 ± 0.39 ppb and the median was 24.54 ppb in 2010. The difference in NO 2 mean was close to 10 ppb. The comparability of pollutant concentrations in different years depends not only on changes in source emissions but also on weather conditions. The weather influences atmospheric diffusion, transmission, and photochemical reactions, and conditions vary from year to year. During the comparison periods, the average temperatures were 25.7°C were reclassified and averaged according to temperature, relative humidity, wind speed, and wind direction, as suggested by Lin et al. (2012). Thus, the differences were evaluated for various meteorological categories, and the results are shown in Figs. 8 and 9. It is worth noting that less proportions of data could be found in the lowest or the highest values of bins. For example, the frequency of wind speed higher than 3 m s -1 only took up a factor of less than 6% in 2018. Temperatures between 22°C and 35°C comprised 93.5% and 94.4% of the recordings in 2018 and 2010, respectively. As shown in Fig. 8, the concentration of O 3 increased with temperature; as RH and wind speed increased, O 3 concentration increased and then decreased. The differences between the O 3 concentrations in 2010 and 2018 were noticeable for the RH and wind direction categories, with considerably higher concentrations in 2018 than in 2010. However, the differences were more complex for the temperature and wind speed categories, possibly as a result of the interaction effect of temperature and wind speed. Generally, higher temperature resulted in higher wind speeds, and high wind speed could reduce the ground-level O 3 concentration.
As shown in Fig. 9, the concentration of NO 2 increased with RH and decreased with the increasing temperature and wind speed. Because limited data were available from 2010 for temperatures higher than 34°C, the concentrations of NO 2 were considerably higher in 2010 than in 2018 for various meteorological categories. Therefore, emission reduction should effectively reduce NO 2 concentrations. This can be confirmed by decreasing tropospheric NO 2 column density in Tianjin since 2010 (You et al., 2016), according to satellite observations, and the decreasing trend observed in routine NO 2 measurements in the 2018 Tianjin Ecology and Environment Statement. In summary, O 3 concentration during summer in Tianjin increased from 2010 to 2018, whereas NO 2 concentration decreased.

Ozone Production Efficiency
Based on ground measurements, the relationship between O x (O x = O 3 + NO 2 ) and NO z (NO z = NO y -NO x ) was calculated and ozone production efficiency (OPE) was obtained from the slope of the correlation. Fig. 10 depicts the correlation between O x and NO z from 11:00 to 16:00 during the observational period. Linear regression was performed using a reduced major axis regression method. A significant correlation was noted at the confidence level of 0.05, and the correlation coefficient (R) was 0.434. From the slope of the correlation, OPE in summer in urban Tianjin was determined to be 6.0 ± 0.4. This value falls within the OPE values measured in other urban areas (Ge et al., 2010).
A lack of measurements of VOCs makes ozone sensitivity difficult to discuss. Previous studies have suggested that O 3  is more sensitive to VOCs than to NO x in Tianjin (Han et al., 2013;Han et al., 2015). On the basis of the present results, we could only conclude that O 3 increased considerably while NO 2 decreased between 2010 and 2018. Lower NO chemical titration can increase the mean concentration of O 3 and reduce its range, although this does not hold for daytime peak O 3 . Further work is needed to understand if the changes in the emission of VOCs and the ratio of VOCs to NO x have caused the increase in surface O 3 concentration.

CONCLUSIONS
An air flow drainage system utilizing a long Teflon pipe was built on a 355-m meteorological tower in urban Tianjin to obtain simultaneous measurements of gaseous pollutants at a high altitude (220 m a.g.l.) and near the surface (3 m a.g.l.) during summer 2018. O 3 loss in the pipeline was tested and found to be negligible, confirming that the instruments installed 3 m above the ground could accurately measure the concentrations of pollutants at 220 m.
Similar variations in the concentrations at 3 m and 220 m were measured for both O 3 and NO x , but noticeable differences between the two heights during the day and the night. Specifically, the diurnal O 3 concentrations were higher near the surface than at the high altitude, whereas the nocturnal ones displayed the opposite trend. Also, the NO 2 concentrations at 3 m generally exceeded those at 220 m by an amount that was slight during daytime but far more prominent during nighttime. The concentrations of NO were higher near the surface, with the largest and smallest differences between the two altitudes occurring during the day and the night, respectively. Finally, the O x concentrations displayed significant diurnal differences but minimal nocturnal ones between the ground and the upper layer.
The concentrations of the O 3 precursor NO x were higher at 3 m than 220 m, which, along with the much-increased photochemical generation of O 3 (O x ) near ground level during daytime, indicates that the surface O 3 was primarily affected by local photochemical formation during the summer in urban Tianjin. concentration increased on average by approximately 50 ppb near the surface and 46 ppb in the upper layer. The measurements collected during summer 2010 reveal that the NO x concentrations significantly decreased between this period and summer 2018, whereas the O 3 concentrations increased, and the vertical structures for these pollutants also greatly changed from 2006 till 2018. Using field measurements, the summer ozone production efficiency measurement, OPE (O x /NO z ) was reported as 6.0 ± 0.4 during the summer in urban Tianjin for the first time.