Quantifying vehicle restriction related PM2.5 reduction using field observations in an isolated urban basin

Vehicle (related particulate matter) emissions, including primary vehicle (related particulate matter) emissions, secondary nitrate, and road dust, have become an important source of fine particulate matter (PM2.5) in many cities across the world. The relationship between vehicle emissions and PM2.5 during vehicle restrictions has not yet been revealed using field observational data. To address this issue, a three-month field campaign on physical and chemical characteristics of PM2.5 at hourly resolution was conducted in Lanzhou, an urban basin with a semi-arid climate. The Lanzhou municipal government implemented more strict vehicle restriction measure during the latter part of field campaign period. The concentration of nitrogen oxides (NO x ) and PM2.5 decreased by 15.6% and 10.6%, respectively during the strict vehicle restriction period. The daily traffic fluxes decreased by 11.8% due to the vehicle restriction measure. The vehicle emission reduction led to a decrease of 2.43 μg·m−3 in PM2.5, including the decrease of primary vehicle emissions, secondary nitrate, and road dust. The contribution of vehicle emissions to PM2.5 decreased by 9.0% based on the results derived from a positive matrix factorization model. The sources other than vehicle emissions increased by 0.2 μg·m−3. Combining all evidence from the observations, the reduction of vehicle emissions is almost equal to the observed reduction in PM2.5. A further extrapolation that 9.0% reduction in vehicle emissions led to the observed reduction in PM2.5 (2.32 μg·m−3). This study clearly quantifies the vehicle restriction related PM2.5 reduction using field observations. The results provide scientific support for the implementation of effective vehicle emission reduction measures.


Introduction
Vehicle (related particulate matter) emissions, including primary vehicle (related particulate matter) emissions, secondary nitrate, and road dust, are a major source of fine particulate matter (PM 2.5 ) that can reduce visibility and damage human health [1][2][3].Source apportionment studies using receptor model showed that the contribution of vehicle emissions to the total carcinogenic risk has surpassed 50% during low pollution days and has increased from 34% to 63% in recent years [4,5].The sources of PM 2.5 , and their contributions, have changed significantly with the development of society and economy.Industrial emissions have declined over the past few years, whereas vehicle emissions have increased [6][7][8][9][10].Because of the increasing number of vehicles [11], vehicle emissions have become an important source of PM 2.5 in many cities, especially in the developed countries in recent years [12][13][14][15][16].Therefore, quantifying vehicle restriction related PM 2.5 reduction is crucial.
There has been an increasing interest in studying the impact of vehicle emission reduction on PM 2. 5 .Some studies have directly or indirectly assessed the impact of vehicle emission reduction on PM 2. 5 .Many studies have used model simulations and prediction models directly to assess the impact of vehicle emission reduction on PM 2.5 [17][18][19][20][21].However, model simulations or prediction models may suffer from uncertainties and limitations in accurately reflecting the actual situation.Real-time observations are obtained from the atmosphere.Therefore, realtime observation data can yield insights that may not have been drawn by model simulations or prediction models.Some studies investigated the influence of vehicle emission reduction on PM 2.5 using observational data during several special events such as the 2006 Sino-African Summit [22], the 2008 Beijing Olympics [23,24], the 2019 Coronavirus Disease (COVID-19) [25,26], Beijing 2022 Olympic Games [20].Not only were vehicle emissions restricted, but emissions from other sources, such as factories with high emissions were temporarily shut down during these events.It interfered with assessing the impact of vehicle emission reduction on PM 2.5 .Previous studies have demonstrated that policies on vehicle emission reduction have not been successful in reducing PM 2.5 , potentially owing to their relatively clean background, as well as the effects of short-term meteorology, the implemented way of policies [27,28].Some studies evaluated the changes of the conventional air pollutants (PM 10 , PM 2.5 , and NO x ), watersoluble inorganic ions and volatile organic compounds (VOCs) during vehicle emission reduction [29][30][31][32][33][34][35].Nevertheless, it is difficult to clearly distinguish between the changes in PM 2.5 due to changes in vehicle emissions or other emission sources using the conventional air pollutants, water-soluble inorganic ions or VOCs.Researcher estimated the contributions of vehicular emissions to PM 2.5 using continuous measurement of OC-EC data, which the result indicated the effectiveness of vehicle control measures [36].A critical review of the literature revealed that quantifying vehicle restriction related PM 2.5 reduction has not yet been revealed using observational data during the control of vehicle emission.There are several main reasons.Firstly, model simulations or prediction models may suffer from uncertainties and limitations in accurately reflecting the actual situation.Additionally, the data on detailed chemical composition from in observational data are lacking during the control of vehicle emission.Thirdly, the selected emission reduction events are inappropriate.Not only were vehicle emissions restricted, but emissions from other sources.Moreover, the changes in PM 2.5 is also influenced by meteorological conditions and regional transport.Therefore, more efforts needed to be made to quantify vehicle restriction related PM 2.5 reduction.
Lanzhou is an isolated valley city, which is less affected by regional transport from other areas than Lanzhou.The number of vehicles registered in Lanzhou has rapidly increased in recent years, from 614 800 at the end of 2014 to 1208 800 by the end of 2021.Vehicle emissions were a major source in Lanzhou, accounting for the largest fraction of PM 2.5 in all seasons, except spring, when vehicle emissions ranked second to mineral dust emitted from natural sources [37,38].To alleviate traffic congestion and to improve the air quality, the Lanzhou municipal government to optimize vehicle restriction measure on 13 July 2021, which provides a unique opportunity to investigate the relationship between vehicle emissions and PM 2.5 .A three-month field campaign on physical and chemical characteristics of PM 2.5 at hourly resolution was conducted in Lanzhou, an urban basin with semi-arid climate, to reveal the relationship between vehicle emissions and PM 2.5 during the control of vehicle emission.The observation data in situ was used to study vehicle emission reduction on PM 2.5 from the perspective of the source of PM 2.5 , assess the changes in physical and chemical properties, clearly distinguish the changes in the source of PM 2.5 during the control of vehicle emission.To our best knowledge, this is the first study that clearly quantifies vehicle restriction related PM 2.5 reduction using field observations in an isolated urban basin.The results of this study will provide a more comprehensive understanding of vehicle emissions control in semi-arid regions.This paper reveals that the vehicle restriction measure in a semi-arid city could play a positive role in improving air quality.

Field sampling
The field campaign was performed at the Lanzhou Atmospheric Components Monitoring Superstation (LACMS; 36.05 • N, 103.87 • E) from 1 June 2021, to 18 August 2021.To have a better representation on city scale, the instruments were positioned on the roof of a 10-story building on the campus of Lanzhou University.The LACMS site is surrounded by residential and commercial areas and is far from major traffic arteries.The West Donggang and Tianshui highways are located to the north and west of the LACMS site, respectively.

Data collection
A total of 27 advanced environmental monitoring instruments were installed in the LACMS, including photochemical pollution observation, an ambient air quality monitoring, and ground-based remote sensing observation systems.Multiple online instruments were deployed to continuously measure.A detailed introduction of the instruments and data are in table S1 and can be found in the literature [11,[39][40][41].The study employed a dataset of the surface PM 2.5 mass concentration, the V5.GL.02, with a spatial resolution of 0.01 • × 0.01 • , in July and August 2021.The travel intensity index from the Baidu Migration Map in this study.Baidu leverages artificial-intelligence powered mapping systems to identify the daily travel intensity index.The daily traffic flux is a key determinant of the traffic speed.Increasing traffic flux leads to travel time delay which further results in road traffic congestion [42].Thus, to quantify the vehicle restriction measure, the daily traffic fluxes were calculated in this study according to the research method [25].

Source apportionment
PMF, a multivariate factor receptor model provided by the US Environmental Protection Agency, has been widely used for source apportionment.Here, the observed PM 2.5 and chemical composition data were pooled together and input into the PMF model.The details are available in the supplementary materials.The results were evaluated using error estimation methods to ensure the stability of the solution, including bootstrap (BS), displacement of factor elements, and BS with displacement (BS-DISP).

Deweathering using the random forest (RF) model
A meteorological normalization technique using the RF model was applied to normalize impact of meteorological conditions.In this study, the prediction features used in establishing the model include meteorological and time variables.The meteorological normalized concentration at a particular hour was calculated by averaging 1000 predictions from the meteorological variables (excluding all time variables) randomly resampled from the observation period.In this way, the impact of weather variations on air pollutants can be normalized.More detailed description about this technique can be found in in the literature [43,44].

Variations in air pollutants and optical properties during control period and strict control period
The study period was divided into two phases based on the dates of the implementation of the vehicle restriction measure: (1) the control period (CP), from June 1 to July 7, and (2) the strict control period (SCP), from July 13 to August 18.The period from July 8-12 (i.e. the Lanzhou Investment and Trade Fair) was skipped to avoid the impact of different vehicle restriction measure (figure S1).The spatial distribution of PM 2.5 and the comparison of NO x , aerosol scattering coefficients (σ sp ), aerosol absorption coefficients (σ ap ), PM 2.5 , and visibility during the CP and SCP are presented in figure 1.Two different vehicle restriction measures were implemented during the CP and SCP.Therefore, in order to estimate the impacts of optimizing vehicle restriction measure on the air quality, we compared the atmospheric pollutants during the CP and SCP.The mass concentration of PM 2.5 during the SCP was 19.59 ± 10.81 µg•m −3 , which is lower by 10.6% compared with the CP (21.91 ± 12.83 µg•m −3 ) (figure 1(c)).The values after the '±' symbol indicate the standard deviation of the measurement.According to the data by the Atmospheric Composition Analysis Group of Washington University, the mass concentration of PM 2.5 show a decreasing trend over most of the region from CP to SCP (figures 1(a) and (b)).The percentage decrease in PM 2.5 is comparable to the result of Zhengzhou and Beijing-Tianjin-Hebei (13.5%-17.5%)[18,[45][46][47], but is lower than that in the Yangtze River Delta region (28%-46%) [48].Compared the changes of PM 2.5 between CP and SCP periods in Lanzhou with previous events when other cities that not only were vehicle emissions restricted, but emissions from other sources, which our aim highlights the effectiveness of the vehicle restriction measure in Lanzhou.
Vehicle emissions are the major contributor to NO x [49].The NO x concentration during the SCP was 45.50 ± 37.06 µg•m −3 , which is lower by 15.6% compared with the CP (figure 1(c)).The changes of NO x concentration were observed to be different from the result that the NO x concentration increased during Hangzhou G20 summit, which is mainly due to long-range transport during Hangzhou G20 summit [50].The implementation of vehicle emission measure resulted in a reduction of 15.6% and 10.6% in NO x and PM 2.5 , respectively.It strongly indicates the effectiveness of vehicle emission reduction in reducing NO x and PM 2.5 .
Aerosol light extinction (b ext ), which includes aerosol absorption (b ap ) and scattering (b sp ), is a key determinant of the ambient visibility.The changes of aerosol absorption and scattering was observed to be the same as the result in Wuhan (figure 1(c)), where aerosol scattering decreased more than aerosol absorption [51].The decrease in aerosol scattering was faster than the aerosol absorption, resulting in a reduction in SSA from 0.81 to 0.79 at 520 nm (figure 1(c)).The visibility increased by 11.2% from 26.1 km in CP to 29.7 km in SCP (figure 1(c)).This is different from the result that the low-visibility frequency (<10 km) barely changed despite the reduction in PM 2.5 was more than 30% in Pearl River Delta regions [52], which indicates the uniqueness in reducing PM 2.5 in Lanzhou.Importantly, it also indicated that strict vehicle emission measure is effective in improving ambient visibility while reducing PM 2.5 .

Characteristics and sources of PM 2.5 during CP and SCP
Variations in chemical species and sources of PM 2.5 during CP and SCP were investigated.The variations in chemical species and sources of PM 2.5 are illustrated in figure 2. The contribution of organic matter (OM) to PM 2.5 was the largest during CP and SCP (figure 2(b)).No significant change was found in the proportion of OM in recent years compared  with the previous study conducted in Lanzhou [53].Industrial sources were the major contributors to SO 2 [10].The mixed industrial emission exhibited strong correlations with the SO 2 (R 2 = 0.58), which means that mixed industrial emission was the major contributors to SO 2 (figure S2).From CP to SCP, mixed industrial emission increased, which corresponded to the increase in SO 2 .High concentrations of SO 2 may provide enough gas precursors to form large amounts of sulfate.The contribution of SO 4 2-increased from 12.9% in of CP to 16.22% in SCP, which was consistent to the increase of its precursor SO 2 in SCP.No significant difference was recorded in the sulfur conversion ratio, although SO 4 2-exhibited the increase (12.8%) during SCP, indicating the degree of secondary formation varied slightly (figure S3), which is the same as results in Beijing [54].The contribution of NO 3 − to PM 2.5 decreased compared with the CP (figure 2(b)), which was consistent with the decrease in its precursor NO x from CP to SCP.Cu and Zn are derived from brake and tire wear in vehicle emissions, while Ca 2+ is associated with road dust [55][56][57].The Ca 2+ concentration decreased compared with the CP (figure 2(a)), corresponding to a decrease in road dust.The reductions in Cu, Zn, and Ca 2+ concentrations indicated that vehicle emission measure was effective in reducing road dust.
The sources of PM 2.5 were apportioned using the PMF model.The results of the source apportionment during different sampling periods are shown in figure 2. The CP samples yielded six PM 2.5 sources, including primary vehicle emissions, secondary sulfate, secondary nitrate, mixed industrial emission, mineral dust, and road dust (figure 2(c)).The correlation coefficients between these sources and their tracers are above 0.83 (figure S4), indicating that these sources are independent.Primary vehicle emissions were characterized by high OC and EC loadings.A strong correlation was observed between this source and the EC (R = 0.83) (figure S4).Mixed industrial emission was characterized by high fractions of Fe, Mn, Pb, and Zn [58].A significant correlation was found between Mn and mixed industrial emission (R = 0.96) (figure S4).High loadings of crustal elements (e.g.Fe, Ca, Si, and Ti) were found in the mineral dust, exhibiting a good correlation with the constructed fine soils (R = 0.97) (figure S4).The extremely high fractions of NH 4 + , SO 4 2-, and NO 3 -, NH 4 + defined the sources of secondary sulfate and secondary nitrate, respectively.Significant correlations were also observed between the sources and their tracers, indicating that secondary sulfate and nitrate were independent (R ⩾ 0.87) (figure S4).The secondary formation was further divided into secondary sulfate and secondary nitrate.Road dust was characterized by a high proportion of Ca 2+ , and the source demonstrated a good correlation with Ca 2+ (R = 0.94) (figure S4).
The SCP samples yielded seven PM 2.5 sources, including primary vehicle emissions, secondary sulfate, secondary nitrate, coal combustion, mixed industrial emissions, mineral dust, and road dust (figure 2(b)).Coal combustion was characterized by a high proportion of C1 -.A significant correlation was observed between Cl − and the coal combustion (R = 0.99) (figure S5).
To better understand the sources of PM 2.5, the contribution of the PM 2.5 sources was further compared during CP and SCP.Primary vehicle emissions, secondary formation, and mineral dust were the main contributors to PM 2.5 .The contribution of primary sources such as primary vehicle emissions, and mineral dust in Lanzhou was higher than the megacities such as Beijing [54], and Qingdao [8], whereas the secondary formation was lower than in these cities.The result shows that secondary formation is predominant in megacities, while the primary vehicle emissions are predominant in semi-arid regions.Thus, the control of vehicle emissions is an effective way to reduce the concentration of PM 2.5 in Lanzhou.Primary vehicle emissions were the largest contributors to PM 2.5 during the CP and SCP.The primary vehicle emissions contribution to PM 2.5 in CP and SCP were 7.28 µg•m −3 and 5.84 µg•m −3 , with a contribution percentage of 38% and 35%, respectively.Compared to the CP, the contribution of primary vehicle emissions, secondary nitrate, road dust, and mineral dust to PM 2.5 correspondingly decreased during SCP (figure 2(c)).However, the contributions of secondary sulfate, coal combustion, and mixed industrial emission increased (figure 2(c)).

Variations in concentration of air pollutants in 2021 compared with the same time in 2020
To assess the impact of vehicle emission reduction on PM 2.5 , the primary vehicle emissions were compared (figure 3).In 2020, there was a slight increase in primary vehicle emissions from CP to SCP.It can be attributed to no change in vehicle emission measures in 2020.However, strict vehicle emission measure was implemented in 2021, leading to a 19.8% and 15.6% reduction in primary vehicle emissions and NO x from CP to SCP, respectively (figure 3(a)).By comparing the data for in 2020 at the same periods as in 2021, we find that primary vehicle emissions increased in 2020 at the same periods as in 2021, further demonstrating the effectiveness of the vehicle restrictions in 2021.
Although the number of vehicles increased by 5.9% in 2021 compared with 2020, primary vehicle emissions in the same period decreased by 9.6% in 2021 compared with the SCP in 2020 (figure 3(b)), demonstrating the effectiveness of vehicle emission control.Vehicle emissions are considered the most important source of EC emissions [37].The EC during the SCP in 2021 reached 0.56 ± 0.51 µg•m −3 (figure 3(b)), which decreased by 30.0%compared to the SCP in 2020.This decline could be attributed to the implemented vehicle restriction measures.Vehicle emission reduction is generally highly reflected in NO x concentration reduction.The average NO x concentrations reached 57.69 ± 32.38 µg•m −3 in 2021, which decreased by 21.1% compared to the same period in 2020 (figure 3(b)).The daily traffic fluxes decreased from 1780 during CP to 1602 during SCP, influenced by the vehicle restriction measure (figure S6).The daily traffic fluxes decreased by 11.8% influenced by the vehicle restriction measure.

Quantifying vehicle restriction related PM 2.5 reduction
Vehicle emissions directly emit organic carbon (OC), elemental carbon (EC), NO x , and NH 3 and cause road dust, leading to secondary formation [59,60].According to previous studies, NO x emitted by vehicle emissions is converted into secondary nitrate through photochemical reactions in the atmosphere [61].Road dust is generated by tire and brake wear, which is also part of vehicle emissions [62].Considering the indirect effects of vehicle emissions on road dust and secondary nitrate, the impact of vehicle emissions on PM 2.5 includes primary vehicle emissions, secondary nitrate, and road dust [59][60][61][62].
To study the relationship between vehicle emissions on PM 2.5 , in-situ observation data was used to reveal the relationship between vehicle emissions and PM 2.5 during the control of vehicle emission.To better understand the impact of vehicle emissions on PM 2.5 , the relationship between vehicle emissions and PM 2.5 was further studied and assess the impact of the vehicle restriction measure on chemical characteristics and sources of PM 2.5 .
The primary vehicle emissions, secondary nitrate and road dust are 7. Considering the impact of vehicle emissions on PM 2.5 , including primary vehicle emissions, secondary nitrate, and road dust [59][60][61][62].The contribution of vehicle emissions to PM 2.5 decreased by 9.0%.The vehicle emission reduction led to a decrease of 2.43 µg•m −3 in PM 2.5 , which is different from the result that the PM 2.5 concentrations could only be reduced by 1.2 µg•m −3 even without vehicle emissions in urban and southern rural areas of Beijing [21].
To evaluate the effectiveness of the vehicle emission reduction, the influence of meteorological conditions and emissions on PM 2.5 need to be quantified.Utilizing the positive matrix factorization model, we clearly distinguish the change in the source of PM 2.5 during CP and SCP.Simultaneously, a meteorological normalization technique was employed to quantify the influence of meteorological conditions on PM 2. 5 .The details about the meteorological normalization model performance evaluation are available in the supplementary materials (figure S7).
As variations in the deweathered concentrations could represent emission-driven changes [40,41].Combining all evidence from the observations, the reduction of vehicle emissions is almost equal to the observed reduction in PM 2.5 .A further extrapolation that 9.0% reduction in vehicle emissions led to the observed reduction in PM 2.5 (2.32 µg•m −3 ).

Conclusion
By utilizing the opportunity of the Lanzhou municipal government to optimize vehicle restriction measure on 13 July 2021, a three-month field campaign on physical and chemical characteristics of PM 2.5 at hourly resolution was conducted in Lanzhou, an urban basin with semi-arid climate, to reveal the relationship between vehicle emissions and PM 2.5 during the control of vehicle emission and assess the impact of the vehicle restriction measure on chemical characteristics and sources of PM This study demonstrated that controlling vehicle emissions is an effective approach for reducing PM 2.5 .This study provides new insight for the formulation of effective policies to improve aerosol pollution in semi-arid regions.

Figure 2 .
Figure 2. (a) The average concentrations chemical species during CP and SCP (b) the average contributions of chemical species to PM2.5 during CP and SCP (c) the average contribution of PM2.5 sources during CP and SCP.

Figure 3 .
Figure 3. (a) The comparison of primary vehicle emissions and NOx during the same time in 2020 and 2021 (b) variations in the EC, primary vehicle emissions and NOx during SCP in 2020 and 2021.

Figure 4 .
Figure 4. (a) The concentration and contribution of primary vehicle emissions, secondary nitrate and road dust during CP and SCP.(b) Emission and meteorology contribute to the changes of PM2.5.(c) The PM2.5 changes in observed and deweathered.

Figure 5 .
Figure 5.The changes in PM2.5 sources in SCP relative to CP in 2021.

2 . 5 .
The mass concentration of PM 2.5 during the SCP was 19.59 ± 10.81 µg•m −3 , which is lower by 10.6% compared with the CP period (21.91 ± 12.83 µg•m −3 ).The concentrations of NO x decreased by 15.6%, with an 11.2% increase in visibility.The daily traffic fluxes decreased by 11.8% influenced by the vehicle restriction measure.Although the number of motor vehicles increased by 5.9% in 2021 compared to 2020, the primary vehicle emissions in the same period in 2020 decreased by 9.6% compared to the SCP in 2021.The EC during the SCP in 2020 reached 0.80 ± 0.53 µg•m −3 , demonstrating a 30.0%decrease compared to the SCP in 2021.The average NO x concentrations of the SCP in 2021 reached 57.69 ± 32.38 µg•m −3 , demonstrating a 21.1% decrease compared to the same period in 2020.The impact of vehicle emissions on PM 2.5 , includes primary vehicle emissions, secondary nitrate, and road dust.The results demonstrated that the vehicle emission reduction led to a decrease of 2.43 µg•m −3 in PM 2.5 and the contribution of vehicle emissions decreased by 9.0% based on the positive matrix factorization model.Importantly, the sources other than vehicle emissions increased by 0.2 µg•m −3 to PM 2.5 concentration.Combining all evidence from the observations, the reduction of vehicle emissions is almost equal to the observed reduction in PM 2.5 .A further extrapolation that 9.0% reduction in vehicle emissions led to the observed reduction in PM 2.5 (2.32 µg•m −3 ).