Impact of Encouraging Vehicles to Refuel at Night on Ozone and Non-Methane Hydrocarbons (NMHCs): A Case Study in Ji’nan, China

: Gasoline evaporation is a potential source of ambient non-methane hydrocarbons (NMHCs) during summer, and thus the policy of encouraging vehicles to refuel at night has been implemented to control ground-level ozone (O 3 ) and NMHCs. In this study, NMHCs and trace gases were observed online at an urban site of Ji’nan during May–July in 2019 and 2020 to assess the impact of this policy. After the implementation of this policy, the average concentration of daily maximum 8 h moving average O 3 decreased from 198 µ g/m 3 to 181 µ g/m 3 . Meanwhile, the average mixing ratio of NMHCs decreased from 19.89 ppbv to 18.02 ppbv. Sources of NMHCs were then apportioned using the positive matrix factorization model. Four factors were resolved and identiﬁed, including vehicle exhaust, paint and solvents usage, gasoline evaporation, and biogenic emission. Relative contributions of these four sources were 52.5%, 20.6%, 18.3%, and 8.6%, respectively. After the implementation of this policy, relative contributions of gasoline evaporation in 1:00–4:00 increased from 20.2–22.7% to 25.4–28.2%, while those for 16:00–18:00 decreased from 16.8–18.7% to 13.9–15.7%. The non-linear relationship of O 3 with NMHCs and NO x was investigated using a box model based on observations. Results suggest that O 3 production was mainly controlled by NMHCs. Aromatics and alkenes were the key NMHC species in O 3 formation. Furthermore, two scenarios of encouraging vehicles to refuel at night were designed to evaluate their impact on O 3 . The relative decreases of O 3 peak concentrations were lower than 1%, indicating that this policy had a limited impact on O 3 during the observation period.


Introduction
Ground-level ozone (O 3 ) is a typical secondary pollutant that is formed by a series of photochemical reactions of non-methane hydrocarbons (NMHCs) and nitrogen oxides (NO x ) under solar radiation [1,2]. Therefore, O 3 usually displays higher concentrations at noon and early afternoon and lower values at night and in the morning [3,4]. A lot of studies reported that O 3 production shows a highly non-linear relationship with NMHCs and NO x [5][6][7]. Due to the abundant NO x emission from vehicular exhaust, O 3 formation tends to be mainly controlled by NMHCs (i.e., in the NMHCs-limited regime) in urban regions, which means that reduction of NMHCs emissions is more effective to control O 3 pollution [5][6][7]. Ambient NMHCs were pre-concentrated by two capillary columns filled with absorbents (enrichment trap and focus trap) at room temperature (about 25 °C), and then were vaporized and injected into a gas chromatography (GC) system equipped with a flame ionization detector (FID) and a mass spectrometer detector (MSD) (GCMS-QP2020, Shimadzu Cooperation, Kyoto City, Japan) for separation and detection. Except for the three species with two carbon atoms, the other target NMHCs were separated on a DB-1 column (0.25 mm × 60 m × 1 μm, Agilent Inc., Santa Clara, CA, USA). The oven temperatures for the GC-MS/FID system initiated at 35 °C which was held for 10 min, then increased from 35 °C to 100 °C at 10 °C/min and was held for 5 min at 100 °C, then increased to 220 °C at 10 °C/min and was then held for 3 min at 220 °C, followed by an increase to 240 °C at 10 °C/min, which was then held for 3.5 min. Fifty-seven NMHC species were quantified by using this GC-MS/FID system, including 29 alkanes, 11 alkenes, acetylene, and 16 aromatics. The mixed 1-ppm standard gas for 57 NMHCs (Spectra gasses Inc., Branchburg, NJ, USA) was diluted to 6 calibration standards with mixing ratios of 0.5 ppbv, 2 ppbv, 3 ppbv, 4 ppbv, 8 ppbv, and 10 ppbv. Correlation coefficient (r) values of calibration curves ranged from 0.99 to 1.00. The method detection limits (MDL) for these NMHC species were in the range of 0.009-0.161 ppbv.

Observation-Based Model (OBM)
In this study, the relationship of O3 formation with its precursors was analyzed using a 0-dimension observations-based chemical transport box model that was developed by Cardelino and Chameides (1995) [16]. The 1 h average measurement data for NMHCs, NOx, O3, CO, temperature, and relative humidity were inputted into the OBM to simulate the net production potential of O3  during the daytime (07:00-19:00). The following Equation (1) was used to calculate relative incremental reactivity (RIR) values for individual precursors (X): Ambient NMHCs were pre-concentrated by two capillary columns filled with absorbents (enrichment trap and focus trap) at room temperature (about 25 • C), and then were vaporized and injected into a gas chromatography (GC) system equipped with a flame ionization detector (FID) and a mass spectrometer detector (MSD) (GCMS-QP2020, Shimadzu Cooperation, Kyoto City, Japan) for separation and detection. Except for the three species with two carbon atoms, the other target NMHCs were separated on a DB-1 column (0.25 mm × 60 m × 1 µm, Agilent Inc., Santa Clara, CA, USA). The oven temperatures for the GC-MS/FID system initiated at 35 • C which was held for 10 min, then increased from 35 • C to 100 • C at 10 • C/min and was held for 5 min at 100 • C, then increased to 220 • C at 10 • C/min and was then held for 3 min at 220 • C, followed by an increase to 240 • C at 10 • C/min, which was then held for 3.5 min. Fifty-seven NMHC species were quantified by using this GC-MS/FID system, including 29 alkanes, 11 alkenes, acetylene, and 16 aromatics. The mixed 1-ppm standard gas for 57 NMHCs (Spectra gasses Inc., Branchburg, NJ, USA) was diluted to 6 calibration standards with mixing ratios of 0.5 ppbv, 2 ppbv, 3 ppbv, 4 ppbv, 8 ppbv, and 10 ppbv. Correlation coefficient (r) values of calibration curves ranged from 0.99 to 1.00. The method detection limits (MDL) for these NMHC species were in the range of 0.009-0.161 ppbv.
Ambient concentrations of O 3 , NO x , and CO were online measured using an ultraviolet (UV) absorption spectrometer (Model T400, Teledyne API Inc., San Diego, CA, USA), a chemiluminescence NO-NO 2 -NO x analyzer (Model T200, Teledyne API Inc., USA), and an infrared absorption spectrometer (Model T300, Teledyne API Inc., USA), respectively. Meteorological parameters were measured using a weather station (Met One Instruments Inc., Grants Pass, OR, USA), including wind speed and direction, temperature, and relative humidity.

Observation-Based Model (OBM)
In this study, the relationship of O 3 formation with its precursors was analyzed using a 0-dimension observations-based chemical transport box model that was developed by Cardelino and Chameides (1995) [16]. The 1 h average measurement data for NMHCs, NO x , O 3 , CO, temperature, and relative humidity were inputted into the OBM to simulate the net production potential of O 3 (P O3−NO ) during the daytime (07:00-19:00). The following where ∆X represents the change of concentration for X. S(X) and ∆S(X) represent the source function of X and its change attributed to ∆X, respectively. ∆S(X)/S(X) was set as 10% in this study.

Positive Matrix Factorization (PMF) Model
The positive matrix factorization (PMF) model version 5.0 (PMF 5.0) that was developed by the U.S. Environmental Protection Agency (EPA) was applied in this study for NMHCs source apportionment. More details about the principles of PMF 5.0 were introduced in the user guide [17]. Briefly, the speciated NMHCs measurement data matrix (x) can be decomposed into two matrices, i.e., factor contributions (g) and factor profiles (f ) (Equation (2)).
where x ij means the concentration of j species measured in i sample. g ik and f kj represent the contribution of the k source to i sample and percentage of j species in the k source, respectively. e ij represents residual for j species in i sample; p represents the number of factors.

Meteorological Conditions during May-July of 2019 and 2020
Meteorological parameters can influence the formation and removal processes of O 3 [18]. Figure 2 shows wind roses during 1 May-15 June and 16 June-31 July of 2019 and 2020. The dominant wind direction was southeast through the whole two months of 2020. Similarly, the wind was also mainly from the southeast during 16 June-31 July, 2019. However, wind directions during 1 May-15 June 2019 showed a significant difference from the other three periods, which were mainly from southeast, northwest, and southwest. The average wind speeds during 16 June-31 July of 2019 and 2020 were both 2.0 m/s, slightly lower than the results for 1 May-15 June of 2019 and 2020 (2.3-2.5 m/s). This means that wind direction and speed during 1 May-15 June 2019 were different from results for the other three periods, which could result in a discrepancy of concentrations for O 3 and its precursors.
To avoid meteorological influences on ambient levels of O 3 and its precursors, those observation data under some meteorological conditions that are not beneficial for O 3 local production were eliminated, mainly including those days with precipitation or daily average wind speed ≥3.4 m/s. There were 24, 30, 28, and 25 days left during 1 May-15 June and 16 June-31 July of 2019 and 2020, respectively. Measurement data for meteorological parameters and mixing ratios of O 3 and its precursors on these days were used to further discuss differences of meteorological conditions among these periods and evaluate the impact of this price incentives policy for nighttime refueling on O 3 and NMHCs.
As shown in Figure 3, the average daily temperature during 1 May-15 June and 16 June-31 July, 2020 was 25 • C and 26 • C, respectively. There were 16 and 18 days with the daily maximum temperature (DMT) exceeding 30 • C during these two periods. The cumulative hours of temperature exceeding 30 • C were 129 and 122 during these two periods, respectively. In 2019, the average daily temperature during 1 May-15 June and 16 June-31 July was 24 • C and 29 • C, respectively. There were 11 days (90 h) and 27 days (281 h) with DMT exceeding 30 • C during these two periods, respectively. Overall, the meteorological conditions in the two periods of 2020 were similar but showed significant differences to 2019. The average temperature and days and hours with temperature exceeding 30 • C during 16 June-31 July, 2019 were significantly higher than the results for the other three periods, which could be beneficial for O 3 formation. However, days and cumulative hours with temperature exceeding 30 • C for 1 May-15 June 2019 were lower than the other three periods. Based on the above results, it can be found that meteorological conditions, including wind direction and speed, and temperature, were similar between the two periods in 2020, but showed differences to those for 2019. To avoid meteorological influences on ambient levels of O3 and its precursors, those observation data under some meteorological conditions that are not beneficial for O3 local production were eliminated, mainly including those days with precipitation or daily average wind speed ≥3.4 m/s. There were 24, 30, 28, and 25 days left during 1 May-15 June and 16 June-31 July of 2019 and 2020, respectively. Measurement data for meteorological parameters and mixing ratios of O3 and its precursors on these days were used to further discuss differences of meteorological conditions among these periods and evaluate the impact of this price incentives policy for nighttime refueling on O3 and NMHCs.
As shown in Figure 3, the average daily temperature during 1 May-15 June and 16 June-31 July, 2020 was 25 °C and 26 °C, respectively. There were 16 and 18 days with the daily maximum temperature (DMT) exceeding 30 °C during these two periods. The cumulative hours of temperature exceeding 30 °C were 129 and 122 during these two periods, respectively. In 2019, the average daily temperature during 1 May-15 June and 16 June-31 July was 24 °C and 29 °C, respectively. There were 11 days (90 h) and 27 days (281 h) with DMT exceeding 30 °C during these two periods, respectively. Overall, the meteorological conditions in the two periods of 2020 were similar but showed significant differences to 2019. The average temperature and days and hours with temperature exceeding 30 °C during 16 June-31 July, 2019 were significantly higher than the results for the other three periods, which could be beneficial for O3 formation. However, days and cumulative hours with temperature exceeding 30 °C for 1 May-15 June 2019 were lower than the other three periods. Based on the above results, it can be found that meteorological conditions, including wind direction and speed, and temperature, were similar between the two pe-

Ambient Levels of O 3 and Non-Methane Hydrocarbons (NMHCs) during 2019 and 2020 3.2.1. Comparison of O 3 Concentrations
As shown in Figure 4, the average value of DMA-8h O 3 concentrations at the Ji'nan site during 1 May-15 June and 16 June-31 July 2020 was 198 µg/m 3 and 181 µg/m 3 , respectively. There were 21 days and 18 days with DMA-8h O 3 concentrations exceeding 160 µg/m 3 during these two periods, respectively. It seems that the average DMA-8h O 3 concentration decreased by 8.6% and non-attainment days were reduced by 3 after the implementation of the price incentives policy for nighttime refueling from 16 June 2020.

Comparison of O3 Concentrations
As shown in Figure 4, the average value of DMA-8h O3 concentrations at the Ji'nan site during 1 May-15 June and 16 June-31 July 2020 was 198 μg/m 3 and 181 μg/m 3 , respectively. There were 21 days and 18 days with DMA-8h O3 concentrations exceeding 160 μg/m 3 during these two periods, respectively. It seems that the average DMA-8h O3 concentration decreased by 8.6% and non-attainment days were reduced by 3 after the implementation of the price incentives policy for nighttime refueling from 16 June, 2020.
The average DMA-8h O3 concentration during 1 May-15 June 2020 was 27 μg/m 3 higher than that for the same period in 2019, while the value for 16 June-31 July 2020 was 20 μg/m 3 lower. The differences in meteorological conditions were possible causes for the discrepancy of O3 concentrations between 2019 and 2020. The average temperature showed the lowest value during 1 May-15 June 2019 but the highest value for 16 June-31 July 2019 ( Figure 3). To avoid influences from meteorological conditions, further analysis of O3 and NMHCs mainly focused on the two periods before and after the implementation of this policy in 2020.

Variations of NMHCs Mixing Ratios
Average NMHCs mixing ratios at the Ji'nan site during 1 May-15 June and 16 June-31 July in 2019 and 2020 were compared in Figure 5. In general, the average mixing ratio of NMHCs during May-July 2020 was 18.95 ppbv, which was 17.8% higher than the result (16.09 ppbv) for 2019. For the two periods in 2020, the average NMHCs mixing ratio during 16 June-31 July was 18.02 ppbv, which was slightly lower than the result (19.89 ppbv) during 1 May-15 June.

Diurnal Variations of O3 and NMHCs Levels
To further analyze the impact of this price incentives policy for nighttime refueling, the average diurnal variation patterns of O3 and NMHCs levels during 1 May-15 June and 16 June-31 July 2020 were compared in Figure 6. Mixing ratios of NMHCs in two periods both showed the maximum value in the morning rush hour (7:00-9:00), and then decreased gradually to the minimum value in the afternoon (13:00-15:00). This was probably due to the expansion of the boundary layer and the strong photochemical removal during noon and early afternoon [19]. Contrary to NMHCs mixing ratios, O3 concentrations showed a maximum value around 15:00 and then decreased gradually to low values during nighttime. This indicates that O3 at this site was mainly from secondary production.
During 1 May-15 June 2020, the average mixing ratios of NMHCs during daytime (06:00-20:00) and nighttime (20:00-06:00) were 18.25 ppbv and 22.27 ppbv, respectively. The ratio of NMHCs mixing ratios during nighttime versus daytime of 1.22. After the implementation of this policy, the average mixing ratios of NMHCs during daytime and

Diurnal Variations of O 3 and NMHCs Levels
To further analyze the impact of this price incentives policy for nighttime refueling, the average diurnal variation patterns of O 3 and NMHCs levels during 1 May-15 June and 16 June-31 July 2020 were compared in Figure 6. Mixing ratios of NMHCs in two periods both showed the maximum value in the morning rush hour (7:00-9:00), and then decreased gradually to the minimum value in the afternoon (13:00-15:00). This was probably due to the expansion of the boundary layer and the strong photochemical removal during noon and early afternoon [19]. Contrary to NMHCs mixing ratios, O 3 concentrations showed a maximum value around 15:00 and then decreased gradually to low values during nighttime. This indicates that O 3 at this site was mainly from secondary production.

Variations of Pentanes Mixing Ratios
Pentanes (i.e., the sum of i-pentane and n-pentane) are important components of gasoline and, therefore, they are often used as tracers for gasoline evaporation [10,12]. Figure 7 compares average mixing ratios of pentanes and their percentages in total NMHCs during the daytime (06:00-20:00) and nighttime (20:00-06:00) of 1 May-15 June and 16 June-31 July 2020. Before the implementation of this policy, the average mixing ratios of pentanes during nighttime and daytime were 2.29 ppbv and 1.66 ppbv, respectively. The ratio of daytime average level for pentanes versus nighttime was 1.38. During 16 June-31 July nighttime decreased to 16.75 ppbv and 20.28 ppbv, respectively. The ratio of NMHCs mixing ratios during nighttime versus daytime was 1.21, close to the value for 1 May-15 June 2020. It should be noted that the relative decrease of NMHCs levels during 13:00-17:00 between these two periods was 15%, larger than that for nighttime (9%). For O 3 , its hourly average concentrations during 13:00-17:00 decreased from 191-208 µg/m 3 during 1 May-15 June 2020 to 181-186 µg/m 3 during 16 June-31 July 2020.

Variations of Pentanes Mixing Ratios
Pentanes (i.e., the sum of i-pentane and n-pentane) are important components of gasoline and, therefore, they are often used as tracers for gasoline evaporation [10,12]. Figure 7 compares average mixing ratios of pentanes and their percentages in total NMHCs during the daytime (06:00-20:00) and nighttime (20:00-06:00) of 1 May-15 June and 16 June-31 July 2020. Before the implementation of this policy, the average mixing ratios of pentanes during nighttime and daytime were 2.29 ppbv and 1.66 ppbv, respectively. The ratio of daytime average level for pentanes versus nighttime was 1.38. During 16 June-31 July 2020, average mixing ratios of pentanes during nighttime and daytime decreased to 2.11 ppbv and 1.56 ppbv, respectively. The ratio of daytime average level for pentanes versus nighttime was 1.35, close to the value for 1 May-15 June 2020. From the perspective of percentages of pentanes in total NMHCs, its values during nighttime and daytime of 16 June-31 July, 2020 were 10.7% and 9.6%, higher than results for 1 May-15 June 2020 (9.8% for the daytime; 8.9% for the nighttime), respectively. The ratio of average percentage for pentanes in total NMHCs during nighttime versus daytime increased from 1.12 to 1.15 after the implementation of this policy. This finding indicates that this policy possibly resulted in a slight increase of contributions from gasoline evaporation on ambient NMHCs during nighttime.

Identification of PMF-Resolved Factors
Values of signal versus noise (i.e., S/N > 5), indications for sources, and high ambient mixing ratios were considered when selecting these NMHC species as inputs of the PMF model [17]. In this study, measurement data for 19 NMHC species were selected and inputted into the PMF model for source apportionment, including 8 alkanes, 6 aromatics, 4 alkenes, and acetylene ( Figure 8). The PMF solutions with 3-8 factors were resolved and compared with each other. When factors were more than 4, the PMF-resolved solutions were hardly explained by individual sources. Therefore, the 4-factor PMF solution was used for source apportionment. Chemical profiles of these factors are shown in Figure 8a-d.

NMHCs Sources Apportionment during May-July of 2020 3.3.1. Identification of PMF-Resolved Factors
Values of signal versus noise (i.e., S/N > 5), indications for sources, and high ambient mixing ratios were considered when selecting these NMHC species as inputs of the PMF model [17]. In this study, measurement data for 19 NMHC species were selected and inputted into the PMF model for source apportionment, including 8 alkanes, 6 aromatics, 4 alkenes, and acetylene ( Figure 8). The PMF solutions with 3-8 factors were resolved and compared with each other. When factors were more than 4, the PMF-resolved solutions were hardly explained by individual sources. Therefore, the 4-factor PMF solu-

Relative Contributions of Individual Sources to NMHCs Levels and Ozone Formation Potential (OFP) during May-July of 2020
The average relative contributions of 4 sources to the summed mass concentrations of 19 NMHC species during May-July 2020 are shown in Figure 9a. Vehicle exhaust was the dominant source of NMHCs, with a relative contribution of 52.5%, followed by paint and solvent usage (20.6%), gasoline evaporation (18.3%), and biogenic emission (8.6%). Figure 9b,c compare the sources of NMHCs before and after the implementation of this policy. Relative contributions of vehicle exhaust were close between these two periods. Gasoline evaporation contributed 18.7% and 17.9% of NMHCs before and after the implementation of this policy. Meanwhile, relative contributions of biogenic emission increased from 7.5% during 1 May-15 June to 10.1% during 16 June-31 July, while relative contributions of paint and solvent usage decreased from 22.2% to 18.6%.  Factor 1 was characterized by the high abundances of C2-C4 alkanes, C6-C7 alkanes, ethylene, acetylene, and benzene, with relative contributions of 52.5-90.1%, 53.8-54.9%, 78.1%, 72.3%, and 74.6%, respectively. These species were reported to be important components of NMHCs emitted from vehicle exhaust [11,[20][21][22], and thus this factor was considered vehicle exhaust. For factor 2, its relative contributions to C7-C8 aromatics, including toluene, ethylbenzene, xylenes, and styrene, showed high values of 32.6-73.9%. Aromatics are important components of paint and solvents, which are widely used in industrial processes, painting, coating, and household products, etc. [23,24]. Therefore, this factor was identified as paint and solvent usage. Factor 3 was characterized by its high relative contribution (50.7%) to i-pentane, which was an important component of gasoline [13,21]. Therefore, this factor was identified as gasoline evaporation. Factor 4 was characterized by its high relative contribution (91.9%) to isoprene, which is the most important NMHC species from biogenic emissions [25]. In the daytime of summer, the biogenic source was usually considered as the largest contributor to isoprene [26], and thus factor 4 was identified as biogenic emission.

Relative Contributions of Individual Sources to NMHCs Levels and Ozone Formation Potential (OFP) during May-July of 2020
The average relative contributions of 4 sources to the summed mass concentrations of 19 NMHC species during May-July 2020 are shown in Figure 9a. Vehicle exhaust was the dominant source of NMHCs, with a relative contribution of 52.5%, followed by paint and solvent usage (20.6%), gasoline evaporation (18.3%), and biogenic emission (8.6%). Figure 9b,c compare the sources of NMHCs before and after the implementation of this policy. Relative contributions of vehicle exhaust were close between these two periods. Gasoline evaporation contributed 18.7% and 17.9% of NMHCs before and after the implementation of this policy. Meanwhile, relative contributions of biogenic emission increased from 7.5% during 1 May-15 June to 10.1% during 16 June-31 July, while relative contributions of paint and solvent usage decreased from 22.2% to 18.6%. and solvent usage (20.6%), gasoline evaporation (18.3%), and biogenic emission (8.6%). Figure 9b,c compare the sources of NMHCs before and after the implementation of this policy. Relative contributions of vehicle exhaust were close between these two periods. Gasoline evaporation contributed 18.7% and 17.9% of NMHCs before and after the implementation of this policy. Meanwhile, relative contributions of biogenic emission increased from 7.5% during 1 May-15 June to 10.1% during 16 June-31 July, while relative contributions of paint and solvent usage decreased from 22.2% to 18.6%.  To find out the reason why the relative contribution of gasoline evaporation increased after the implementation of this policy, changes of mass concentrations for 4 PMF-resolved factors were further compared. The total mass concentration of 19 NMHCs decreased from 37.08 µg/m 3 to 35.64 µg/m 3 after the implementation of this policy. However, mass concentrations from gasoline evaporation were close during these two periods, with respective values of 6.63 µg/m 3 and 6.67 µg/m 3 . Meanwhile, NMHCs concentrations contributed by vehicle exhaust decreased from 19.43 µg/m 3 to 18.71 µg/m 3 , paint and solvent usage decreased from 8.23 µg/m 3 to 6.64 µg/m 3 , and biogenic emission increased from 2.79 µg/m 3 to 3.62 µg/m 3 , respectively. These results suggest that NMHCs concentrations from gasoline evaporation did not significantly decrease after the implementation of this policy, whereas relative contributions of gasoline evaporation showed an increase due to the decrease of NMHCs concentrations from other sources.
Ozone formation potential (OFP) of individual NMHC species was calculated using the following equation OFP = NMHC i × MIR i , where NMHC i was the mass concentration of individual 19 NMHC species, MIR i was the maximum incremental reactivity coefficient of individual NMHC species simulated by Venecek et al. (2018) [27]. Relative contributions of 4 individual PMF-resolved sources to the total OFP of 19 NMHCs were shown in Figure 9d. Biogenic emission has a higher relative contribution to OFP (16.4%) than that for mass concentration of NMHCs, due to the high reactivity of isoprene. For the three anthropogenic sources, vehicle exhaust was the largest contributor to OFP (39.0%), followed by paint and solvent usage (27.4%) and gasoline evaporation (17.3%). Figure 9e,f compare the relative contributions of individual sources to OFP before and after the implementation of this policy. Relative contributions of vehicle exhaust to OFP were close between these two periods, with respective values of 38.7% and 39.0%. Relative contributions of gasoline evaporation to OFP were 17.0% and 17.4% during these two periods, respectively. Relative contributions of biogenic emission increased from 14.5% during 1 May-15 June 2020 to 19.1% during 16 June-31 July 2020, while relative contributions of paint and solvent usage decreased from 29.8% to 24.4%.
To further investigate the impact of this price incentive policy for nighttime refueling, the average diurnal variation patterns of relative contributions from gasoline evaporation to NMHCs mass concentrations were compared before and after the implementation of this policy (Figure 10). Gasoline evaporation showed higher relative contributions during the nighttime and morning and then decreased gradually to the minimum value in the afternoon (around 16:00-18:00) during both periods. After the implementation of this policy, relative contributions of gasoline evaporation were in the range of 25.4-28.2% in the early morning (1:00-4:00), while the values for 1 May-15 June were 20.2-22.7%. Meanwhile, relative contributions of gasoline evaporation in the afternoon (16:00-18:00) were in the range of 13.9-15.7%, lower than results for 1 May-15 June, with values of 16.8-18.7%. These findings indicate that this policy possibly influenced the diurnal distribution of NMHCs emissions from gasoline evaporation.

Sensitivity of O3 Production of NMHCs and NOx
The O3 concentrations measured in this study showed a negative correlation (r = −0.508 **) with NOx concentrations but a positive correlation (r = 0.202 **) with the ratio of NO2 versus NOx (NO2/NOx) which was often used to indicate aging of air masses [28]. Furthermore, the non-linear relationship of O3 formation with NMHCs and NOx during May-July 2020 was analyzed using two methods, including the RIR values (Equation (1)) and the empirical kinetic modeling approach (EKMA) based on the OBM. The EKMA was modeled based on 20 × 20 combinations of anthropogenic NMHCs and NOx reduction scenarios. Each scenario represented a 5% reduction of anthropogenic NMHCs and NOx from 0% to 100%. Figure 11a shows the isopleth diagram of DMA-8h O3 with relative reductions of anthropogenic NMHCs and NOx (i.e., the EKMA plot). This plot can be divided into 2 parts by the ridgeline (the dashed black line in Figure 11a). The upper-left part represents O3 formation is in the NMHCs-limited regime, while the lower-right part represents O3 formation in the NOx-limited regime. Based on the EKMA plot, the O3 formation in Ji'nan during May-July 2020 was in the NMHCs-limited regime. This means that O3 concentrations would decrease by reducing anthropogenic NMHCs (AHC), while a small reduction of NOx would lead to an increase in O3 concentration. This price incentive policy encouraged the public to refuel vehicles at night and could reduce NMHCs levels in the daytime to some extent, and therefore this policy could inhibit the formation of O3 during the daytime. Figure 11b shows the RIR values for NOx, NHC (i.e., isoprene), AHC, and CO that were calculated by the OBM. The RIR value of NOx was negative (−0.103%/%), indicating that reducing NOx possibly resulted in O3 increase. Contrary to NOx, the RIR values of NHC, AHC, and CO were all positive, which suggests that the reduction of NMHCs and CO would decrease O3 formation. Although the RIR value for CO was also positive, its value was only 0.076%/%, significantly lower than the RIR values for AHC (0.354%/%) and NHC (0.434%/%). This implies that ambient NMHCs played a more important role in O3 formation than CO. Considering it was hard to reduce biogenic emission, reduction of AHC would be the relatively effective way to control O3 pollution.  The O 3 concentrations measured in this study showed a negative correlation (r = −0.508 **) with NO x concentrations but a positive correlation (r = 0.202 **) with the ratio of NO 2 versus NO x (NO 2 /NO x ) which was often used to indicate aging of air masses [28]. Furthermore, the non-linear relationship of O 3 formation with NMHCs and NO x during May-July 2020 was analyzed using two methods, including the RIR values (Equation (1)) and the empirical kinetic modeling approach (EKMA) based on the OBM. The EKMA was modeled based on 20 × 20 combinations of anthropogenic NMHCs and NO x reduction scenarios. Each scenario represented a 5% reduction of anthropogenic NMHCs and NO x from 0% to 100%. Figure 11a shows the isopleth diagram of DMA-8h O 3 with relative reductions of anthropogenic NMHCs and NO x (i.e., the EKMA plot). This plot can be divided into 2 parts by the ridgeline (the dashed black line in Figure 11a). The upper-left part represents O 3 formation is in the NMHCs-limited regime, while the lower-right part represents O 3 formation in the NO x -limited regime. Based on the EKMA plot, the O 3 formation in Ji'nan during May-July 2020 was in the NMHCs-limited regime. This means that O 3 concentrations would decrease by reducing anthropogenic NMHCs (AHC), while a small reduction of NO x would lead to an increase in O 3 concentration. This price incentive policy encouraged the public to refuel vehicles at night and could reduce NMHCs levels in the daytime to some extent, and therefore this policy could inhibit the formation of O 3 during the daytime.

The Impact of This Policy on Simulated O3 Concentrations
To quantitatively evaluate the impact of this policy on O3 concentrations, two scenarios for gasoline evaporation control were designed. The simulated O3 concentrations by the OBM were compared between the baseline scenario and two control scenarios. The measurement data of O3 and its precursors during 1 May-15 June 2020 (i.e., before the implementation of this policy) were used as inputs for the baseline scenario. According to the above results in Section 3.2.4, the ratio of pentanes percentages in NMHCs during the nighttime versus daytime increased by about 3% after the implementation of this policy. Therefore, scenario 1 assumed that the relative contribution of gasoline evaporation to NMHCs decreased by 1.5% during the daytime and increased by 1.5% during the nighttime. Scenario 2 assumed that stricter control measures or preferential policies were implemented to encourage the public to refuel vehicles at night, and then the relative contribution of gasoline evaporation to NMHCs decreased by 30% during the daytime and increased by 30% during the nighttime. It should be pointed out that the concentrations of NMHCs from gasoline evaporation changed in scenario 1 and scenario 2, while concentrations of NMHCs from other sources, as well as levels of NOx, CO, and meteorological parameters, did not change among the baseline scenario and two control scenarios. Figure 12 shows the average diurnal variation patterns of decreases of O3 concentrations simulated by the OBM between the baseline scenario and two control scenarios for all days and non-attainment days during 1 May-15 June 2020. For scenario 1, the average peak concentration of O3 decreased by 0.16 ± 0.15 μg/m 3 at 15:00 and 0.19 ± 0.15 μg/m 3 at 18:00 for all days (Figure 12a) and non-attainment days (Figure 12b), respectively. In scenario 2, the average peak concentration of O3 at 17:00 decreased by 2.25 ± 1.40 μg/m 3 and 2.76 ± 1.19 μg/m 3 for all days and non-attainment days, respectively. These findings suggest that this price incentives policy for nighttime refueling would decrease O3 peak concentrations in the early afternoon. However, the relative decreases of O3 peak concentrations were lower than 1% for scenario 1. This means that the current policy had a limited impact on O3 concentrations at the Ji'nan site during the observation period of 2020.  Figure 11b shows the RIR values for NO x , NHC (i.e., isoprene), AHC, and CO that were calculated by the OBM. The RIR value of NO x was negative (−0.103%/%), indicating that reducing NO x possibly resulted in O 3 increase. Contrary to NO x , the RIR values of NHC, AHC, and CO were all positive, which suggests that the reduction of NMHCs and CO would decrease O 3 formation. Although the RIR value for CO was also positive, its value was only 0.076%/%, significantly lower than the RIR values for AHC (0.354%/%) and NHC (0.434%/%). This implies that ambient NMHCs played a more important role in O 3 formation than CO. Considering it was hard to reduce biogenic emission, reduction of AHC would be the relatively effective way to control O 3 pollution.

The Impact of This Policy on Simulated O 3 Concentrations
To quantitatively evaluate the impact of this policy on O 3 concentrations, two scenarios for gasoline evaporation control were designed. The simulated O 3 concentrations by the OBM were compared between the baseline scenario and two control scenarios. The measurement data of O 3 and its precursors during 1 May-15 June 2020 (i.e., before the implementation of this policy) were used as inputs for the baseline scenario. According to the above results in Section 3.2.4, the ratio of pentanes percentages in NMHCs during the nighttime versus daytime increased by about 3% after the implementation of this policy. Therefore, scenario 1 assumed that the relative contribution of gasoline evaporation to NMHCs decreased by 1.5% during the daytime and increased by 1.5% during the nighttime. Scenario 2 assumed that stricter control measures or preferential policies were implemented to encourage the public to refuel vehicles at night, and then the relative contribution of gasoline evaporation to NMHCs decreased by 30% during the daytime and increased by 30% during the nighttime. It should be pointed out that the concentrations of NMHCs from gasoline evaporation changed in scenario 1 and scenario 2, while concentrations of NMHCs from other sources, as well as levels of NO x , CO, and meteorological parameters, did not change among the baseline scenario and two control scenarios. Figure 12 shows the average diurnal variation patterns of decreases of O 3 concentrations simulated by the OBM between the baseline scenario and two control scenarios for all days and non-attainment days during 1 May-15 June 2020. For scenario 1, the average peak concentration of O 3 decreased by 0.16 ± 0.15 µg/m 3 at 15:00 and 0.19 ± 0.15 µg/m 3 at 18:00 for all days (Figure 12a) and non-attainment days (Figure 12b), respectively. In scenario 2, the average peak concentration of O 3 at 17:00 decreased by 2.25 ± 1.40 µg/m 3 and 2.76 ± 1.19 µg/m 3 for all days and non-attainment days, respectively. These findings suggest that this price incentives policy for nighttime refueling would decrease O 3 peak concentrations in the early afternoon. However, the relative decreases of O 3 peak concentrations were lower than 1% for scenario 1. This means that the current policy had a limited impact on O 3 concentrations at the Ji'nan site during the observation period of 2020.

Conclusions
The price incentives policy for nighttime refueling was implemented from 16 June, 2020 in Ji'nan to control O3 pollution. The impact of this policy on O3 and NMHCs was evaluated based on online observation data of O3, NMHCs, NOx, and CO at an urban site in Ji'nan during May-July of 2019 and 2020.
Changes in O3 concentrations, NMHCs levels, and sources: The average DMA-8 O3 concentration during 16 June-31 July 2020 decreased by 8.6% to 181 μg/m 3 . Meanwhile, the average mixing ratios of NMHCs decreased by 9.5% to 18.02 ppbv. The PMF model was applied for NMHCs source apportionment. Four source categories were identified, including vehicle exhaust, paint and solvent use, gasoline evaporation, and biogenic emission. Vehicle exhaust was the dominant source to NMHCs, with a relative contribution of 52.5%, followed by paint and solvent use (20.6%), gasoline evaporation (18.3%), and biogenic emission (8.6%). After the implementation of this policy, relative contributions of gasoline evaporation in 1:00-4:00 increased from 20.2-22.7% to 25.4-28.2%, while those in 16:00-18:00 decreased from 16.8-18.7% to 13.9-15.7%. This indicates that this policy possibly influenced the diurnal distribution of gasoline evaporation.
O3 sensitivity to its precursors: The RIR values for AHC, NHC, and CO were positive,