Causes of a continuous summertime O 3 pollution event in Ji ’ nan , a central 1 city in the North China Plain 2

In summer 2017, measurements of ozone (O3) and its precursors were carried out at an 18 urban site in Ji’nan, a central city in the North China Plain (NCP). A continuous O3 pollution 19 event was captured during August 4-11, with the maximum hourly O3 reaching 154.1 ppbv. 20 Model simulation indicated that local photochemical formation and regional transport 21 contributed 14.0±2.3 and 18.7±4.0 ppbv/hr to the increase of O3 during 09:00-15:00 local time 22 (LT) in this event, respectively. For local O3 formation, the calculated OH reactivities of volatile 23 organic compounds (VOCs) and carbon monoxide (CO) were comparable between O3 episodes 24 and non-episodes (p>0.05), so was the OH reactivity of nitrogen oxides (NOx). However, the 25 ratio of OH reactivity of VOCs and CO to that of NOx increased from 2.0±0.4 s -1 /s -1 during non26 episodes to 3.7±0.7 s -1 /s -1 during O3 episodes, which resulted in the change of O3 formation 27

emission and outdated production capacity. It has been repeatedly confirmed that air pollution in 118 these types of cities in the NCP has seriously deteriorated air quality in Beijing (Lin et al., 2008; 119 Wang et al., 2010). Thus, O 3 pollution in Ji'nan is also a regional issue. Contradictory to the      The inlet was located ~1 m above the rooftop of the 7-story building (~22 m a.g.l.). O 3 and 151 NO/NO x were detected with a UV photometric based analyzer and a chemiluminescence NO-152 NO 2 -NO x analyzer, respectively. NO 2 was calculated from the difference between NO and NO x . 153 Studies indicated that NO 2 monitored with chemiluminescence was generally overestimated due suburban and mountainous sites in China, Xu et al. (2013) suggested that the chemiluminescence 161 monitors overestimated NO 2 by less than 10% in urban areas with fresh emission of NO x . As 162 described in section 2.1, our sampling site was located in the urban area of Ji'nan and was only 163 ~50 m to a main road. Therefore, we infer that NO 2 might not be significantly overestimated in 164 this study. However, the influences of the overestimation on the findings were still discussed 165 qualitatively where necessary. 166 The hourly concentrations of sulfur dioxide (SO 2 ) and CO were acquired from a nearest AQMS 167 of CNEMC which is ~1 km from our sampling site. Year-round monitoring of inorganic trace 168 gases was conducted at this AQMS, where the air was drawn into the analytical instruments at a 169 flow rate of 3 L/min through an inlet, ~1 m above the rooftop of a 5-story building (~ 16 m a.g.l.).
170 Table S1 provides the details of the trace gas analyzers used in this study, including the 171 instrumental model, resolution, accuracy, precision and detection limit. The hourly 172 concentrations of O 3 and NO 2 (NO data was not available on CNEMC website) measured at our 173 sampling site agreed well with those reported at the AQMS, with the slope of 1.04 (R 2 = 0.82) 174 and 1.13 (R 2 = 0.71) for O 3 and NO 2 in the linear least square regressions, respectively (Figure 175 S1). Due to the instrumental differences and/or differences in sources and sinks of air pollutants 176 at the two sites, the agreements were worse at low mixing ratios for both O 3 and NO 2 . Therefore, 177 we only used SO 2 and CO monitored at the nearest AQMS in this study, which had lower 178 photochemical reactivity than O 3 and NO 2 , and might be more homogeneous at a larger scale. 179 In addition, the meteorological parameters, including wind speed, wind direction, pressure, 180 temperature and relative humidity were monitored by a widely used weather station (China 181 Huayun group, Model CAWS600). The daily total solar radiation was obtained from the   185 During the real-time measurement of trace gases, offline VOC and oxygenated VOC (OVOC) 186 samples were collected on 9 selective days (i.e., July 20 and 30, August 1, 4-7 and 10-11), 187 referred to as VOC sampling days hereafter. The days were selected to cover the periods with 188 relatively high and low levels of O 3 . The high O 3 days were forecasted prior to sampling based 189 on the numerical simulations of meteorological conditions and air quality. In total, 6 out of 9 those analyzed by UCI and good agreements were achieved ( Figure S2). OVOC samples were 206 eluted with 5 mL acetonitrile, followed by analysis with the high performance liquid 207 chromatography. The DL, accuracy and precision for all OVOCs analysis were within the range 208 of 3-11 pptv, 0.32-0.98% and 0.01-1.03%, respectively. simulate O 3 in this study. WRF v3.6.1 was run to provide the offline meteorological field for 214 CMAQ v5.0.2. A two-nested domain was adopted with the resolution of 36 km (outer domain) 215 and 12 km (inner domain), respectively. As shown in Figure S3    NO and NO 2 were separately measured and input into the model, which experienced different  principles of PMF can be found in Paatero and Tapper (1994). Briefly, the model treats the 295 matrix of input concentrations as the product of two matrixes (i.e., factor contribution and factor 296 profile). Here, hourly concentrations of 31 VOCs, CO, NO and NO 2 in 54 samples were input 297 into the model. The VOCs, which were common tracers of specific sources (e.g., isoprene for 298 biogenic emissions), and had relatively high concentrations (detectable in at least 80% samples), 299 were selected for source apportionment (termed as VOCs* hereafter). On average, VOCs* 300 accounted for 79.5±11.7% of the total quantified VOCs (mean ± 95% confidence interval of the 301 hourly values in the statistical period, same for all the other "a ± b" expressions elsewhere unless for the concentrations lower than and 304 higher than DL, respectively.

305
The model was run for 20 times with a random seed and the best resolution automatically given 306 by the model was accepted. A total of 6 sources of O 3 precursors were resolved by PMF in this   conditions. As aforementioned, the solar radiation on July 30 was much weaker than those 382 during O 3 episodes, which was probably the most critical factor leading to low O 3 on this day. 383 Figure S5 shows the COD retrieved from the terra/MODIS (https://ladsweb.modaps.    (high pressure, low temperature and low solar radiation), as discussed in section 3.1.

412
Nevertheless, we believe that the incoming sea and coastal air at least did not aggravate air  To better understand the relationship between O 3 pollution and the synoptic systems, Table 1 435 summarizes the characteristics of synoptic system, weather condition and air mass origin on 436 individual VOC sampling days. The weather charts for the surface level and 500 hPa on August Ji'nan on August 1, the COD decreased relative to that on July 30 ( Figure S5). Correspondingly, day. In addition, the temperature was relatively low on July 30 and August 1. Though the OH 467 reactivity of O 3 precursors on these days was comparable to or even higher than that on August

496
The water areas are highlighted in blue.  is noteworthy that these mixing ratios were 5 -10 times higher than their averages. Further, most 511 of these VOCs are highly reactive in O 3 photochemistry and may make great contributions to  Table 2. Generally, the lower Diff., RMSE, 538 NMB and NME, but higher IOA indicate better agreement between the simulated and observed    The IPR analysis quantifies the contributions of different processes to the O 3 production rate, as   NO 2 were assigned to the exhausts from diesel and gasoline vehicles, particularly to diesel 624 exhaust which was responsible for more than half of these trace gases. According to the process analysis by WRF-CMAQ, local photochemical formation was an

697
The method was applied to each of the six sources, derived from the PMF analysis, thereby 698 acquiring the contribution to O 3 production rate of each source.

699
On average, the source contributions (in ppbv/hr) to O 3 production rates during O 3 episodes and 700 non-episodes are presented in Table 3. It was found that gasoline exhaust and diesel exhaust 701 were the largest contributors to O 3 production regardless of O 3 episodes or non-episodes.

702
Specifically, the net O 3 production rate was 1.0±0.3 ppbv/hr for both gasoline and diesel exhaust 703 during non-episodes, which however increased to 1.8±0.6 ppbv/hr for gasoline exhaust and   As shown in Figure   production rates in transition area ( Figure 10)  Ji'nan. Figure S12 shows the simulated O 3 production rate as a function of the source emission 811 reduction, which also confirmed the highest efficiencies of O 3 reduction by cutting diesel exhaust 812 (0.58 ppbv· h -1 /10% emission reduction) and gasoline exhaust (0.47 ppbv· h -1 /10% emission 813 reduction). We also found that the reduction of diesel exhaust would lead to the increase of O 3 814 production rate when the reduction percentages were lower than a dividing point (e.g., 60% on 815 August 10), while further reductions would decrease the O 3 production rate. This was due to high