Sheng Xiang1, Shaojun Zhang1,2,3, Yu Ting Yu1, Hui Wang1, Ye Deng4, Qinwen Tan4, Zihang Zhou4, Ye Wu This email address is being protected from spambots. You need JavaScript enabled to view it.1,2,3

1 School of Environment, State Key Joint Laboratory of Environment Simulation and Pollution Control, Tsinghua University, Beijing 100084, China
2 State Environmental Protection Key Laboratory of Sources and Control of Air Pollution Complex, Beijing 100084, China
3 Beijing Laboratory of Environmental Frontier Technologies, School of Environment, Tsinghua University, Beijing 100084, China
4 Chengdu Academy of Environmental Sciences, Chengdu 610072, China

Received: February 25, 2023
Revised: June 30, 2023
Accepted: July 2, 2023

 Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Download Citation: ||  

Cite this article:

Xiang, S., Zhang, S., Yu, Y.T., Wang, H., Deng, Y., Tan, Q., Zhou, Z., Wu, Y. (2023). Evaluating Ultrafine Particles and PM2.5 in Microenvironments with Health Perspectives: Variability in Concentrations and Pollutant Interrelationships. Aerosol Air Qual. Res. 23, 230046.


  • PM concentrations were measured by a mobile platform in microenvironments.
  • Traffic emissions contribute ~30% particle numbers regardless of the seasons.
  • Solid fuel burning significantly contributes to rural PM2.5 mass during winter.
  • No strong correlation was found between the source-discerned PNC and PM2.5.


Regulation has been applied to the fine particles (PM2.5) but not to particle number concentrations (PNC). We use a mobile platform to measure PNC and PM2.5 in four microenvironments (diesel plume, urban freeway, urban street, and rural freeway). A total of 38661 pairs of measurements in two years (winter 2018 and autumn 2020) are used to evaluate variability in the pollutant concentrations and their interrelationships. Source-discerned total PNC (PNCtot) and temporal-adjusted PM2.5 (∆PM2.5) are calculated and evaluated. Results showed that the average PNCtot in winter (4.8 × 104 pt cm–3) were over two times higher than autumn (0.36 × 104–0.56 × 104 pt cm–3). Moreover, the traffic emissions (PNCd,tr) contribute 30% of the PNC throughout the study while solid fuel burning (PNCd,sfb) could be a major contributor only in winter (29%). Seasonal variability in PNCd,tr and PNCd,sfb was found, with 2–3 times higher median PNCd,tr and 7 times higher median PNCd,sfb in winter compared to autumn. Similarly, PM2.5 in winter (109 µg m–3) was 3–5 times higher than autumn, while ∆PM2.5 (40 µg m–3) was 3–6 times higher. In winter, the PM2.5 and ∆PM2.5 showed higher concentrations in urban street and rural freeway similar to PNCd,sfb but opposite to the trend of PNCtot and PNCd,tr. The correlation coefficient (R2) is investigated as three combinations (i.e., PNCtot vs. PM2.5, PNCd,tr vs. ∆PM2.5, PNCd,sfb vs. ∆PM2.5). Here, the R2 showed a comparable seasonal trend (winter lower than autumn) and similar magnitude as the literature, but no strong correlation (R2 < 0.15) was found. This stresses the fact that mitigation measures of PM2.5 do not necessarily reduce PNC and monitoring networks evaluate PM2.5 exposure are unlikely to represent PNC exposure. The concentration ratios in the three combinations are found to vary with microenvironments and seasons. This variability implies that control policies should be diversified with pollutant types and energy usage of the city.

Keywords: Mobile monitoring, Particulate matter, Urban microenvironments, Air quality, Exposure


A large number of epidemiological studies have found connections between fine particulate matter (particles with an aerodynamic diameter (Dp) < 2.5 µm, PM2.5) and premature mortality (Pope III et al., 2009; Apte et al., 2018). It is well known that exposure to PM2.5 could lead to adverse health effects, therefore, the ambient concentrations of PM2.5 have been regulated by many countries. Many efforts have been put into controlling PM2.5 pollution, however, PM2.5 is not the only metric that characterizes particulate matter in terms of air pollution (HEI, 2013). Recent toxicological studies provide evidence that ultrafine particles (particles with Dp < 0.1 µm, UFPs) can penetrate deeper into the respiratory system and may translocate directly into the blood and the brain because of their smaller size (Peters et al., 2006; Chen et al., 2017). Thus, not only the penetration pathways but also the adverse health effects related to UFPs may be different from PM2.5. PM2.5 is characterized by mass concentrations (PM2.5 mass, µg m–3). However, the contribution of UFPs to particle mass is considered negligible (Morawska et al., 1998; Mejía et al., 2008), hence, the particle number concentration (PNC, particle cm–3 or pt cm–3) is often used as a proxy measure (Hinds, 1999).

The World Health Organization (WHO) issued a new air quality guideline in 2021 that suggests a more stringent requirement of ambient PM2.5 mass concentration but not for PNC (WHO, 2021). Field measurement studies that used stationary monitoring sites examined the variations of PNC and PM2.5 mass concentrations worldwide and found poor to moderate Pearson correlation (R2 ranged from 0.005 to 0.36) between the two metrics (PNC vs. PM2.5 mass) (Marconi et al., 2007; Rodríguez et al., 2007; De Jesus et al., 2019). However, Saha et al. (2020) found a relatively strong Pearson correlation (R2 = 0.4) between the two metrics by defining different types of Pearson correlation (i.e., temporal and spatial correlations). Here, the temporal correlation is the Pearson correlation between the two metrics on a time basis while the spatial correlation represents the Pearson correlation between the two metrics at well-defined domain. Depending on the experimental design, the extent of the domain could be as large as a city or as small as a road segment (Levy et al., 2014; Saha et al., 2020). They applied their definition to literature (Hoek et al., 2011; Eeftens et al., 2015; Cattani et al., 2017; Wolf et al., 2017) and found that spatial correlations generally showed a range of 0.17 to 0.46. Mobile measurements were also utilized to investigate the variability of the two metrics and their interrelationship in urban areas. Levy et al. (2014) applied a car-based mobile platform to collect multiple air pollutants in Montreal, Canada and found the correlation between the two metrics was 0.10 and 0.14 during the winter and autumn, respectively. Hankey and Marshall (2015) used a bicycle-based mobile platform to evaluate the variability of particulate air pollution in Minneapolis, USA, and found no correlation (R2 from 0.001 to 0.02) between the two metrics. In addition to R2, the ratios of PNC to PM2.5 mass concentrations (PNC/PM2.5 ratios, unit: 109 pt µg–1) were also investigated to provide a qualitative assessment of the source mixture. Based on results from different continental, De Jesus et al. (2019) found that the PNC/PM2.5 ratios were between 0.14 × 109 pt µg–1 and 2.2 × 109 pt µg–1 with higher values (> 1) shown for roadside monitoring stations. Saha et al. (2020) subtracted the ambient background of PNC and PM2.5 mass concentrations and found the ratios of the incremental concentrations (∆PNC/∆PM2.5) ranged from 0.59 × 109 pt µg–1 to 5.8 × 109 pt µg–1 with lowest and highest ratio shown in Asia and European cities, respectively.

The variability in R2 and ratios of the pollutants indicates that the relationship between the two metrics could be related to the different dominant sources among the cities and intracity variability (Apte et al., 2017; Saha et al., 2020; Yu et al., 2022a, 2022b). In this study, we applied a car-based mobile platform to investigate the variability in the PNC and PM2.5 mass concentrations and evaluate the interrelationship between them in different microenvironments. The mobile platform provides opportunities to conduct measurements in microenvironments that stationary measurements cannot address and hence advance our understanding of PNC and PM2.5 mass concentrations. To the authors’ knowledge, a study applying mobile monitoring in the assessment of the source-discerned PNC variability and its interrelationship with PM2.5 over a large study domain (3800 km2) has not been conducted previously. The main objective of the present study is to quantify the PNC that attribute to different source emissions (i.e., traffic and solid fuel burning) and evaluate the correlation of PNC and PM2.5 mass concentrations in a variety of microenvironments. We used a mobile platform to characterize the variations of PNC in the four microenvironments (i.e., diesel plume, urban freeway, urban street, rural freeway) in 2018 and 2020. We applied an approach proposed by Casquero-Vera et al. (2021) to quantify the PNC contributed by traffic and solid fuel burning emissions. This proposed theory is based upon the work of Rodríguez and Cuevas (2007) and Sandradewi et al. (2008a, 2008b) and is demonstrated in the Methods Section. The Pearson correlation of source-discerned PNC and PM2.5 mass concentration is characterized and intercompared across the four microenvironments. The implication of the pollutant concentration variability and their interrelationship is discussed.


2.1 Air Pollutant Measurements

Field measurements were conducted in Chengdu, an inland megacity with more than 20 million inhabitants and 5.2 million registered vehicles in 2020 (Chengdu Bureau of Statistics, 2019) in southwest China. This study was part of the Chengdu Air Pollution Evaluation (CAPE) project (Xiang et al., 2021, 2022). Three sampling campaigns were conducted in 2018 and 2020: 2018 non-holiday (Jan. 12–18, Winter), 2020 non-holiday (Sep. 21–30, Oct. 9–10, Autumn), and 2020 holiday (Oct. 1–8, Autumn). One of the most important holidays in China, the National Holiday (Oct. 1–8 in 2020), was covered by the field measurements. Categorizing data into the three sampling campaigns enables variability of pollutant concentrations and their interrelationship to be evaluated by: (1) the highly variable vehicle fleet composition (lower presence of heavy-duty diesel vehicles (HDDVs) during holidays), and (2) the usage of solid fuel-burning during winter (2018 non-holiday). In this study, we focus on the characteristic of PNC and PM2.5 mass and used measurements of eBC for our analysis. For mobile measurements, high resolution (1-Hz) measurements of PNC, PM2.5, and equivalent black carbon (eBC) concentrations were conducted with a mobile platform (gasoline-powered SUV) equipped with a Global Positioning System (GPS) to record speed and geolocation at 1-s resolution. A condensation particle counter (CPC3007, size range: 10 nm–1000 nm) was used to measure the total particle number concentrations (PNCtot). The PM2.5 mass concentrations was measured by a desktop light-scattering aerosol monitor, TSI DustTrak (model 8533) (Wang et al., 2009). eBC concentrations were measured by an aethalometer (AE33) that provided real-time loading compensation using dual-spot technology that eliminates the need for correction for filter loading. The AE33 provided eBC mass concentrations for 7-wavelength (370, 470, 520, 590, 660, 880, and 950 nm) (Drinovec et al., 2015) and was connected to a 2.5 µm sharp-cut cyclone. The default mass absorption efficiency values given by the manufacturer were used in this study.

For all sampling days, the sampling time window was between 9:00 and 18:00. Detailed information on the sampling domain and experimental setup has been provided in our previous studies (Xiang et al., 2021). Briefly, the mobile platform has been driven on 400 km of roadways with a 5,500 km driving distance. The mobile measurements were divided into four microenvironments depending on the characteristics of the roadway where the measurements were conducted. Diesel plume microenvironment was defined as measurements from plume-chasing a heavy-duty diesel vehicle (HDDV) within 30 m distance (equivalent to 1.5 s residence time) on urban and rural freeways (Tong et al., 2022; Xiang et al., 2022, 2023). The urban freeway and rural freeway microenvironments were defined based on the location of the roadways and the numbers of operating HDDVs on the roadways (see the following section for detail). HDDVs were prohibited from entering the city center between 9:00 and 17:00 (overlapping our sampling time). Therefore, the local streets and arterial roadways in the city center were defined as urban street microenvironments. This division of data allows the relative importance of different sources to be examined. The pollutant concentrations in diesel plume and urban freeway microenvironments are more likely to be related to HDDV emissions while those in the urban street and rural freeway microenvironments are more liked affected by emissions of light-duty gasoline vehicles and solid fuel burning. Detailed information regarding the sampling routes of mobile measurements is provided in Fig. S1 in Supporting Information (SI).

2.2 Traffic and Meteorological Information

It is expected that the HDDVs flow rate would reduce during the National Holiday. To quantify the reduction of HDDV flow rate due to the holiday effect, traffic fleet information (e.g., HDDV flow rate) was derived from traffic monitors in urban freeway and rural freeway microenvironments (see Section S1 for traffic analysis). Generally, the total vehicle flow rate was similar during non-holiday and holiday campaigns for both microenvironments. The average vehicle flow rate of HDDV during non-holiday (~680 veh h–1) was 4.4 times higher than during holiday (~154 veh h–1) in the urban freeway microenvironment. There is no noticeable reduction of HDDVs flow rate in rural freeway microenvironment (24 veh h–1 and 22 veh h–1 during non-holiday and holiday, respectively). Since the traffic monitors were not available in 2018, we assume the traffic conditions in 2018 nonholiday were similar to those during 2020 nonholiday. This is considered reasonable because the COVID-19 lockdown control policies (Level 1) in Chengdu was ended on Feb 26th and no stringent control policy were implemented during the 2020 sampling campaigns. Thus, in this study, the effect of traffic condition on the variations of pollutants were mainly based on the comparison of 2020 nonholiday and 2020 holiday sampling campaigns.

Meteorological information was taken from an airport weather tower located within the study domain. The 2018 sampling campaign showed lower average temperature (12°C) and relative humidity (48%) than the 2020 sampling campaigns (21°C and 80%, respectively). All campaigns had similar average wind speeds (~2.0 m s1) with northerly wind direction as prevailing wind. The major industrial areas are located in the south and southeast (downwind side of the prevailing wind) relative to the urban area of Chengdu, based on the city municipal planning (Chengdu Municipal Bureau of Planning and Natural Resources, 2021). The distance between the industrial areas and the nearest sampling route is about 10 km. Considering the location and the distance between the industrial areas and our sampling route, the industrial activities may not a major source of particulate matter and eBC in the study domain. Measurements were conducted only when atmospheric stability was classified as neutral (class D) or stable (class E). The distribution of wind speed and wind direction are described in detail in our previous study (Xiang et al., 2021).

2.3 Data Processing and Analysis

For mobile measurements, pollutant concentrations were synchronized to GPS timestamps and averaged at a resolution of 5-s. Each 5-s measurement was a separate measurement. Daily in-field calibration checks for zero and flow checks were performed before and after each sampling day. The 5-s measurement were first divided into the three sampling campaigns then subdivided into the four microenvironments. To minimize bias in the measurements due to self-pollution, 5-s averaged measurements (i.e., PNCtot, PM2.5 mass, and eBC concentrations) were removed if the speed of the mobile platform were less than 5 km h–1. The 5-s averaged measurement was removed if the GPS information (i.e., longitude, latitude) of the mobile platform was missing. The pairs of 5-s averaged measurements were also removed if one of the following conditions was met: (1) PNCtot were lower than 500 pt cm–3; (2) PM2.5 mass concentrations were lower than 2 µg m–3; (3) eBC concentrations were lower than 100 ng m–3. 85% of the mobile measurements (38661 pairs of 5-s averages) passed the quality control protocol of this study. These 5-s averaged measurements were used for further data analysis.

2.4 PNC Source Apportionment

A recent household energy use survey (Duan et al., 2014) shows that in the urban area of Chengdu, solid fuel (mainly coal combustion and biomass burning) is used for nearly 30% and 8% for cooking and heating activities, respectively. However, in rural areas, solid fuel is responsible for cooking and heating activities with higher percentages usage (Duan et al., 2014; Tao et al., 2018; Meng et al., 2019). To estimate the PNC contributed by different sources (i.e., traffic and solid fuel burning), we first discerned PNC into two components (i.e., direct and indirect emissions) following the methodology introduced by Rodríguez and Cuevas (2007). A scaling factor (Sd) was determined based on the relationship between PNCtot and total eBC (ng m–3) concentrations. The Sd is interpreted as the minimum number of UFPs from direct emissions per each nanogram of eBC (106 pt ng–1 eBC). After determination of the Sd, Eq. (1) was used to estimate the PNCtot (pt cm–3) from direct emissions (PNCd), see Eq. (1) as follow:


To further discern the PNCd contributed by traffic and solid fuel burning, we followed a new approach proposed by Casquero-Vera et al. (2021) that determined contribution from different sources to PNCtot based on Aethalometer Model (Sandradewi et al., 2008a, 2008b). The Aethalometer Model is a method that apportions eBC based on real-time measurement of the aerosol light absorption. In this study, the light absorption at λ1 = 370 nm and λ2 = 880 nm were used for the eBC source apportionment analysis. This is because the plume of traffic emission and solid fuel burning absorb light that falls in this range of wavelength (Bond et al., 2002; Zotter et al., 2017). Details regarding the calculation of eBC source apportionment is provided in SI (Section S2). We first applied the Aethalometer Model to estimate the eBCtr which is the eBC concentrations from traffic emissions. After that, we evaluate the relationship between the PNCtot and the calculated eBCtr (ng m–3) concentrations to determine a new scaling factor (Sd,tr, 106 pt ng–1 eBCtr). In this study, the Sd is calculated by the quantile linear regression for the 5th percentile as suggested by Casquero-Vera et al. (2021). Finally, the PNCtot that contributed by direct emission of traffic (PNCd,tr, pt cm–3) can be estimated by Eq. (2):


The PNCd,tr are not only the particles directly emitted in vehicle exhaust plume but also the particles nucleating immediately after the emissions (e.g., emissions from passing vehicles) (Rodríguez and Cuevas, 2007; Rönkkö et al., 2017). Assuming the significant sources of eBC are traffic and solid fuel burning emissions, the PNCtot from direct emission of solid fuel burning (PNCd,sfb, pt cm–3) can be calculated as shown in Eq. (3):


After determination of PNCtot from direct emissions (i.e., PNCd,tr, PNCd,sfb), the PNCtot that contributed by non-direct emission (PNCnd, pt cm–3) can be determined, see Eq. (4):


Here, the PNCnd is considered to account for three components: (1) secondary particles formed by nucleation or chemical reactions from gaseous precursors in the atmosphere, (2) ultrafine particles from long-range transport due to emissions different to traffic and solid fuel burning in other locations at a time, and (3) pre-existed UFPs in the ambient air due to anthropogenic emissions (e.g., aged particle due to emissions from the neighborhood). After the calculation of the different components of PNC, the relative contribution of each component was evaluated. This evaluation allows the relative importance of the PN sources to be identified in different microenvironments.

2.5 Evaluation of PNC and PM2.5 Mass Concentrations

The PM2.5 mass concentrations were collected by a light scattering device (DustTrak 8533) in this study. We collocated the device with a Federal Equivalent Method (FEM) instrument (TEOM 1405, USA) to evaluate the performance of the DustTrak. The device collocation was conducted before and after every sampling day for at least two hours (one hour before, one hour after) in the three sampling campaigns. Fig. S3 shows the comparison between the two instruments. There were 56 one-hour averaged data used for this comparison. The comparison of PM2.5 concentrations showed a slope near 1 with a Pearson correlation coefficient (R2) equal to 0.98 (p < 0.01) which suggests a strong linear relationship between the two instruments. Other statistical parameters that were used to evaluate the agreement among the two instruments include normalized mean square error (NMSE), mean absolute error (MAE), and the root-mean-square error (NRMSE). A value of 0 for the three parameters would indicate an ideal agreement between the two measurements. The definition of these quantitative measures was provided in Section S3. Overall, a good agreement was found between the two instruments (NMSE = 0.02, MAE = 8.5 µg m3, NRMSE = 17%) which indicates that deploying DustTrak 8533 for mobile monitoring can adequately characterize time-averaged PM2.5 mass concentrations at specific locations.

To illustrate the relationship between PNC and PM2.5 mass concentrations in the near-source field, temporal adjustments were applied to measure PM2.5 mass concentrations. This quantifies the incremental PM2.5 concentrations (∆PM2.5) owing to changes in local emissions between different sample days. The measured PM2.5 mass concentrations were adjusted for between-day temporal trends by subtracting the daily minimum value in each microenvironment as:


where PM2.5,i is the measurement of PM2.5 mass concentrations in microenvironment i (µg m–3). ∆PM2.5,i is the temporal adjusted PM2.5 mass concentrations that accounted for local emissions in microenvironment i (µg m–3). The adjustment is considered to be less likely to alter the spatial distribution of the pollutant concentrations emitted by local emissions in a given microenvironment (Brantley et al., 2014). Since the calculated PNCd,tr and PNCd,sfb were considered PNCs from direction emissions, no temporal adjustment was applied to these species.

The linear relationship between PNC and PM2.5 mass concentrations was explored through the Pearson correlation coefficients (R2) and the PNC to PM2.5 mass concentration ratios (PNC/PM2.5 ratios, 109 pt µg–1) across the four microenvironments. Unlike the direct measurements of the emission sources (e.g., tailpipe measurements), the measurements in the four microenvironments enable the linear relationship to be evaluated under real-world conditions with the relative importance of different sources considered. Therefore, the linear relationship evaluates the correspondence between the PNC and PM2.5 mass concentrations that would aid in air pollution control strategy (e.g., end-of-pipe control policies) from a public health perspective. There were three combinations of PNC/PM2.5 ratios being evaluated in this study which were (1) PNCtot vs. PM2.5, (2) PNCd,tr vs. ∆PM2.5, and (3) PNCd,sfb vs. ∆PM2.5. The linear relationship was evaluated on daily basis. Within each combination, if the linear relationship showed no statistical significance (i.e., p > 0.05), the data of the combination on that sampling day will be voided for further analysis of the PNC/PM2.5 ratio. By plotting the PNC against PM2.5 mass concentrations with a scatter plot, the slope of the line provides an estimate of the PNC/PM2.5 ratio (109 pt µg–1). Removing the data of the combination with no statistical significance ensure that the slope of the line is not equal to zero. Variations of the slope are related to the mixture of emission sources. The larger value of the PNC/PM2.5 ratios indicates the larger increase in the PNC that is associated with a unit increase in the PM2.5 mass concentrations. In other words, emissions in microenvironments with a large value of the PNC/PM2.5 ratios were possibly dominated by primary PN emission sources (e.g., vehicle emissions). The variability in the three PNC/PM2.5 ratio combinations provides a qualitative assessment of the emission strength in different microenvironments.


3.1 PNCs in Different Microenvironments

In this study, PNC was direct measurement while the PNCd,tr and PNCd,sfb were calculated following the method recently proposed by Casquero-Vera et al. (2021) (see Fig. S4 in Section S4). Table 1 summarized the scaling factors that were applied in this study. Detail information regarding Sd for each microenvironment during different sampling campaigns in this study was provided in Table S3. The average value of Sd from this study (1.6 × 106 pt ng-1 eBC) was closed to the lower limit reported by the previous studies (from 2 × 106 to 10 × 106 pt ng-1 eBC). The lower value of Sd indicates fewer particles were emitted for a unit of eBC emission. The lower value of Sd in this study could be related to the larger instrument cut-off size (10 nm) since the instruments used by the previous studies have a cut-off size starting from 2.5 nm. For a given emission source, the instruments with a larger cut-off size are more likely to report lower PNC, in our case, resulting in a lower value of Sd. Nevertheless, the calculated Sd and Sd,tr (1.9 × 106 pt ng–1 eBC) in this study were comparable to those values reported by Casquero-Vera et al. (2021) that shared the same instrument cut-off size. Since the result in this study were based on mobile measurement in a variety of microenvironments, no further comparison of Sd,tr can be conducted due to the newness of the data.

Table 1. Summary of average scaling factors obtained from the relationship between PNC and eBC (or eBCtr) concentrations that were reported in previous studies and this study.

The boxplot of PNC, PNCd,tr and PNCd,sfb obtained from the four microenvironments during the 2018 and 2020 sampling campaigns are shown in Fig. 1. On general average, the median PNCtot during 2018 sampling campaign (4.8 × 104 pt cm–3) was higher than 2020 non-holiday (2.2 × 104 pt cm–3) and holiday (1.5 × 104 pt cm–3) campaigns. Similarly, the median PNCd,tr during 2018 non-holiday campaign (1.3 × 104 pt cm–3) was also higher than 2020 non-holiday (0.56 × 104 pt cm–3) and holiday (0.36 × 104 pt cm–3) campaigns (see Figs. 1(A) and 1(B)). Unlike the PNCtot and PNCd,tr that showed 2–3 times higher concentrations during the 2018 non-holiday campaign, the median PNCd,sfb during 2018 campaign was higher than the 2020 campaigns by 7 times, on average. Especially for rural freeway microenvironment, the median PNCd,sfb during the 2018 non-holiday campaign was 16–21 times higher than during the 2020 campaigns. Aside from the averages, the variations of PNCtot and PNCd,tr shared a similar trend, both showed higher median concentrations in diesel plume and urban freeway microenvironments but lower median concentrations in the urban street and rural freeway microenvironments (see Figs. 1(A) and 1(B)). Except for the 2020 holiday campaign, which showed a comparable median concentration of PNCd,tr in urban freeway (0.33 × 104 pt cm–3) and urban street (0.39 × 104 pt cm–3) microenvironments. As demonstrated in the Methods Section, this could be because a large number of HDDVs were not operating during the holidays which significantly reduced the PN emissions (Xiang et al., 2019b, 2019a). The PNCd,sfb showed a trend opposite to the PNCtot and PNCd,tr during 2018 non-holiday campaign. It showed the median concentrations were higher in urban street and rural freeway microenvironments but lower in diesel plume and urban freeway microenvironments (see Fig. 1(C)). However, the there was no clear trend of PNCd,sfb shown during 2020 sampling campaigns. During 2018 campaigns, the relatively higher concentrations of PNCtot and PNCd,tr could be attributed to the higher particle number emission factor due to the increased condensation/nucleation during winter (Saha et al., 2018; Wang et al., 2018) while the significantly higher concentration of PNCd,sfb were related to residential heating activities, especially for rural areas (up to 21 times higher) (Wang et al., 2020).

Fig. 1. Variations of (A) total particle number concentration (PNCtot), (B) PNC from direct traffic emissions (PNCd,tr) and (C) PNC from direct solid fuel-burning emissions (PNCd,sfb) in the four microenvironments for the 2018 and 2020 sampling campaigns. The numbers in bold represent the value of median concentrations in the four microenvironments with a unit of 104 pt cm–3. The lower and upper box boundary indicate the 25th percentile (Q1) and 75th percentile (Q3), respectively. The interquartile range (IQR) represents the distance between Q3 and Q1. The lower and upper levels of the whisker represent the distance from Q3 or Q1 to 1.5IQR.Fig. 1. Variations of (A) total particle number concentration (PNCtot), (B) PNC from direct traffic emissions (PNCd,tr) and (C) PNC from direct solid fuel-burning emissions (PNCd,sfb) in the four microenvironments for the 2018 and 2020 sampling campaigns. The numbers in bold represent the value of median concentrations in the four microenvironments with a unit of 104 pt cm–3. The lower and upper box boundary indicate the 25th percentile (Q1) and 75th percentile (Q3), respectively. The interquartile range (IQR) represents the distance between Q3 and Q1. The lower and upper levels of the whisker represent the distance from Q3 or Q1 to 1.5IQR.

Fig. 2 shows the contribution to PNC from different components (PNCd,tr, PNCd,sfb, PNCnd) in the four microenvironments. The measured PNCs were ranked from the highest (100th) to lowest (0th) values with contributions from different components quantified. Table S4 summarized the detailed statistical results of the contributions for each microenvironment. The average contribution of PNCd,tr was ~30% (± 2, one standard deviation and hereinafter) throughout the sampling campaigns.

Fig. 2. Five second-average PNCtot ranked from the highest (100th) to the lowest (0th) value in (A–C) diesel plume, (D–F) urban freeway, (G–I) urban street, and (J–L) rural freeway microenvironments. The relative contributions (%) of PNCd,tr (red) and PNCd,sfb (green) and PNCnd (grey) to PNC are identified, respectively. The dashed lines represent the median concentrations (50th percentile) of the PNC in Fig. 1.Fig. 2. Five second-average PNCtot ranked from the highest (100th) to the lowest (0th) value in (A–C) diesel plume, (D–F) urban freeway, (G–I) urban street, and (J–L) rural freeway microenvironments. The relative contributions (%) of PNCd,tr (red) and PNCd,sfb (green) and PNCnd (grey) to PNC are identified, respectively. The dashed lines represent the median concentrations (50th percentile) of the PNC in Fig. 1.

The relative steady contribution of PNCd,tr indicates that the traffic emissions are major sources of PNC, regardless of the type of microenvironments. Although one would expect larger contribution from the PNCd,tr, the similar contribution of PNCd,tr is a result from direct emission, nearby emissions (e.g., passing vehicles) and ambient atmosphere (Larson et al., 2017), even for the diesel plume microenvironment. Across the four microenvironments, it can be observed that the contribution of PNCd,sfb were much higher during the 2018 sampling campaign than during the 2020 sampling campaigns, especially in the urban street and rural freeway microenvironments (see Figs. 2(G–L)). In this study, the contributions of PNCd,sfb was ~9 (± 6)%. When compared to PNCd,tr, the larger variations in the contribution of PNCd,sfb are probably related to the residential heating activities during the 2018 sampling campaign of which solid fuel burning contributed 21% and 29% of the PNC in the urban street and rural freeway microenvironments, respectively (see Table S4). Similar to the value reported by Casquero-Vera et al. (2021), the contribution of PNCnd was 52% and 66% during the 2018 and 2020 sampling campaigns, respectively.

As shown in Fig. 2, when the PNC decreased from 100th percentile (highest) to 0th percentile (lowest), the contribution of PNCd,tr and PNCd,sfb increased. For high PNCs (i.e., higher than the median), the average contribution of PNCd,tr during 2018 campaign (24%) was similar to the 2020 sampling campaigns (22%). During the 2018 campaign, the contribution of PNCd,sfb was not major (< 7%) for high PNC levels except for urban streets (15%) and rural freeways (22%). However, for the 2020 campaigns, the contribution of PNCd,sfb was negligible (~3.3%). This indicates that the policymakers aim at air quality compliance through reducing direct emissions need to implement diversified policies based on the local energy use to ensure the effectiveness since the emission sources could not only show significant season variability but also showed spatial variability (e.g., microenvironment-dependent).

3.2 The Relationship between PNC and PM2.5 Concentrations

Fig. 3 shows the measured PM2.5 and temporal adjusted PM2.5 (∆PM2.5) mass concentration in the four microenvironments. On general average, the median PM2.5 mass concentration during the 2018 sampling campaign was 109 (± 32) µg m–3 with higher concentrations occurring in urban streets (135 µg m–3) and rural freeway (138 µg m–3) microenvironments. For the 2020 sampling campaigns, there was no peak concentration shown in a particular microenvironment as in the 2018 sampling campaign (see Fig. 3(A)). The averages of the median PM2.5 mass concentrations were 37 (± 1.8) µg m–3 and 23 (± 5.8) µg m–3 during the 2020 non-holiday and holiday campaigns. Similarly, the median ∆PM2.5 mass concentrations during the 2018 campaign (40 ± 19 µg m–3) were higher than 2020 non-holiday (15 ± 2.9 µg m–3) and holiday (7.1 ± 2.2 µg m–3) campaigns (see Fig. 3(B)). The median ∆PM2.5 mass concentrations in urban streets (41 µg m–3) and rural freeways (66 µg m–3) were higher than in other microenvironments during the 2018 campaign, whereas no similar trend was found during the 2020 sampling campaigns.

Fig. 3. Variations of (A) fine particle mass concentration (Dp ≤ 2.5 µm, PM2.5) and (B) temporal-adjusted PM2.5 concentrations (∆PM2.5) in the four microenvironments for the 2018 and 2020 sampling campaigns. The lower and upper box boundary indicate the 25th percentile (Q1) and 75th percentile (Q3), respectively. The interquartile range (IQR) represents the distance between Q3 and Q1. The lower and upper levels of the whisker represent the distance from Q3 or Q1 to 1.5IQR.Fig. 3. Variations of (A) fine particle mass concentration (Dp ≤ 2.5 µm, PM2.5) and (B) temporal-adjusted PM2.5 concentrations (∆PM2.5) in the four microenvironments for the 2018 and 2020 sampling campaigns. The lower and upper box boundary indicate the 25th percentile (Q1) and 75th percentile (Q3), respectively. The interquartile range (IQR) represents the distance between Q3 and Q1. The lower and upper levels of the whisker represent the distance from Q3 or Q1 to 1.5IQR.

The median PM2.5 mass concentrations were higher in the urban street and rural freeway microenvironments during the 2018 sampling campaign and no clear trend was found during the 2020 sampling campaign. This is different from the PNC which showed consistently higher median concentration in diesel plume and urban freeway microenvironments across all the sampling campaigns (see Fig. 1(A)). Besides the trend of pollutant concentrations, the variations in the contribution of ∆PM2.5 also showed different pattern as those shown in PNCd,tr and PNCd,sfb (see Fig. S5). The discrepancy in the pattern of variations hints that the measurements of PM2.5 mass concentration may not be a suitable indicator of PNC measurements. Moreover, the median ∆PM2.5 mass concentrations showed a similar trend as PNCd,sfb instead of PNCd,tr. The median ∆PM2.5 mass concentrations were higher in urban street and rural freeway microenvironments during the 2018 sampling campaign and were lower during 2020 sampling campaigns with small variations. This indicates that the variations in ∆PM2.5 mass concentrations may more likely be related to solid fuel burning emissions instead of traffic emissions in Chengdu.

Fig. 4 shows the coefficient of determination (R2) of measured (PNCtot vs. PM2.5) and calculated pollutant concentrations (PNCd,tr or PNCd,sfb vs. ∆PM2.5) in the four microenvironments. There were ~20% pairs of the data removed since the linear relationship of the two variables was not significant (see Section S7 for detail). As shown in Fig. 4(A), the results from this study were largely comparable to results reported by Levy et al. (2014) who also evaluated the interrelationship between the PNCtot and PM2.5 mass concentrations based on mobile measurements. The R2 of PNCtot vs. PM2.5 mass concentrations from Levy et al. (2014) were 0.1 and 0.14 during Winter and Autumn, respectively. In this study, the average of the median values of R2 in 2018 (Winter), 2020 (Autumn) and whole sampling campaigns was 0.06 (± 0.03), 0.12 (± 0.04), and 0.1 (± 0.04) respectively. Despite of the differences in sampling years and locations, the R2 from this study showed the same order of magnitude as those derived from Levy et al. (2014) and showed a similar seasonal trend (Winter lower than Autumn). The results from this study were further compared to the median value of literature R2 to provide a qualitative assessment. Studies that evaluate the correlation of PNC and PM2.5 were summarized in Table S8 in Section S8. The temporal correlation of PNCtot and PM2.5 mass concentrations from literature showed a range from 0.005 to 0.36 (median = 0.09). While the spatial correlation of PNCtot and PM2.5 mass concentrations showed a range from 0.17 to 0.46 (median = 0.38). The poor temporal correlation depicts that the variations in PM2.5 mass concentrations may not be representative for PNC. This means that monitoring measures of PM2.5 do not necessarily reflect on PNC from the epidemiological point of view. The large variability in the spatial correlation indicates that the correlation (PNCtot and PM2.5) found in one city may not necessarily reflects in other cities. When compared to the literature, the average of the median values of R2 in this study fell in the range of the temporal correlation but lower than the reported spatial correlation. The results in this study are broadly consistent with previous measurements.

Fig. 4. Variations of Pearson correlation coefficient of (A) measured PNCtot and PM2.5 mass concentrations, (B) calculated PNCd,tr and ∆PM2.5 concentrations and (C) calculated PNCd,sfb and ∆PM2.5 mass concentrations in the four microenvironments for the 2018 and 2020 sampling campaigns. The dashed lines in Fig. 4(A) represent results from Levy et al. (2014). The lower and upper box boundary indicates the 25th percentile (Q1) and 75th percentile (Q3), respectively. The interquartile range (IQR) represents the distance between Q3 and Q1. The lower and upper levels of the whisker represent the distance from Q3 or Q1 to 1.5IQR.Fig. 4. Variations of Pearson correlation coefficient of (A) measured PNCtot and PM2.5 mass concentrations, (B) calculated PNCd,tr and ∆PM2.5 concentrations and (C) calculated PNCd,sfb and ∆PM2.5 mass concentrations in the four microenvironments for the 2018 and 2020 sampling campaigns. The dashed lines in Fig. 4(A) represent results from Levy et al. (2014). The lower and upper box boundary indicates the 25th percentile (Q1) and 75th percentile (Q3), respectively. The interquartile range (IQR) represents the distance between Q3 and Q1. The lower and upper levels of the whisker represent the distance from Q3 or Q1 to 1.5IQR.

The R2 of PNCd,tr vs. ∆PM2.5 and PNCd,sfb vs. ∆PM2.5 were shown in Figs. 4(A) and 4(C), respectively. The average of the median values of R2 for PNCd,tr vs. ∆PM2.5 across the whole study was 0.15 (± 0.08) with a range from 0.04 to 0.34. Large variations were found in the R2 of PNCd,sfb vs. ∆PM2.5. The average of the median values of R2 for PNCd,sfb vs. ∆PM2.5 in the 2018 sampling campaign (0.18 ± 0.06, ranging from 0.1 to 0.23) was comparable to R2 for PNCd,tr vs. ∆PM2.5 across the whole study. This R2 in the 2018 sampling campaign was also nearly a factor of 6 higher than in the 2020 sampling campaigns (0.03 ± 0.016). The extremely low R2 in the 2020 sampling campaigns could be a result from lacking significant emission sources of PNCd,sfb (e.g., residential heating activities). Combining the median values of R2 of PNCd,sfb vs. ∆PM2.5 (only 2018 sampling campaign) and R2 of PNCd,tr vs. ∆PM2.5 (all sampling campaigns) yields the average R2 as 0.15 (± 0.07). This combined average R2 was 50% higher than the average R2 of PNCtot vs. PM2.5 (0.1) and was close to the lower limit of the spatial correlation reported by the literature (0.1–0.46). Even though removing the pollutant concentrations from non-direct emissions raise the combined average R2, the data are still poorly correlated. This is evidence that the source of PN and PM2.5 may not be the same or co-located in the city (Chengdu). The different types of microenvironments represent distinct emission characteristics. Since the measurements in this study were conducted in microenvironments under real-world condition (field-based) instead of well-controlled condition (lab-based). The results could provide transformative information for epidemiological studies in the urban environment. The poor correlation across the three combinations (PNCtot vs. PM2.5, PNCd,tr vs. ∆PM2.5, PNCd,sfb vs. ∆PM2.5) persisted when examining the results from different seasons (i.e., 2018 and 2020 sampling campaigns) and the four microenvironments. This demonstrates that emission control strategies apply on one metric (e.g., PM2.5) may not reflect on the other, regardless of the season and the types of microenvironments.

The PNC/PM2.5 ratios for the three combinations were shown in Fig. 5. The cumulative distribution of the ratio of PNCtot vs. PM2.5 (unit: 109 pt µg–1) was shown in Fig. 5(A). Wide ranges of the PNCtot vs. PM2.5 ratios were shown in all campaigns and the range of the ratios within a campaign was larger than the range among campaigns. In other words, in this study, the PNCtot vs. PM2.5 ratios were more related to the location of measurements instead of the time of the measurements. When compare the PNCtot vs. PM2.5 ratios in microenvironments categories, diesel plume, and urban freeway microenvironments showed a wider range (from 0.15 × 109 pt µg–1 to 8.0×109 pt µg–1) with a higher median value of ratios (0.72 × 109 pt µg–1 and 1.0 × 109 pt µg–1, respectively) compared to urban street and rural freeway microenvironments (range: 0.15 × 109 pt µg–1 ~5.4 × 109 pt µg–1, median: 0.23 × 109 pt µg–1 and 0.48 × 109 pt µg–1, respectively). The literature that simultaneously measured PNC and PM2.5 mass concentrations was also summarized in Table S8. The literature reported a range of ratios from 0.34 × 109 pt µg–1 to 2.2 × 109 pt µg–1 (shown as dashed lines in Fig. 5(A)) with a median value of 0.83. Generally, ~70% of the data in this study fell into the range reported by the literature. In Table S8, most of the literature were conducted in European cities.

 Fig. 5. The cumulative distribution of (A) PNCtot/PM2.5 mass ratio, (B) PNCd,tr/∆PM2.5 ratio and (C) PNCd,sfb/∆PM2.5 ratio in the four microenvironments during the 2018 and 2020 sampling campaigns. The unit of the ratios (x-axis) is 109 pt µg–1. Median values of the ratios of microenvironments were shown in bold numbers. The dashed lines represent the range of PNCtot/PM2.5 mass ratio reported by literature that summarized in Table S8.Fig. 5. The cumulative distribution of (A) PNCtot/PM2.5 mass ratio, (B) PNCd,tr/∆PM2.5 ratio and (C) PNCd,sfb/∆PM2.5 ratio in the four microenvironments during the 2018 and 2020 sampling campaigns. The unit of the ratios (x-axis) is 109 pt µg–1. Median values of the ratios of microenvironments were shown in bold numbers. The dashed lines represent the range of PNCtot/PM2.5 mass ratio reported by literature that summarized in Table S8.

The median value of the ratios reported in the literature (0.83 × 109 pt µg–1) was reasonably close to the average median value of diesel plume (~0.95 × 109 pt µg–1) and urban freeway (~0.73 × 109 pt µg–1) microenvironments but higher than those in the urban street (0.42 × 109 pt µg–1) and rural freeway (0.34 × 109 pt µg–1) microenvironments. One of the reasons that the literature reported a similar value of ratios to the vehicle emission intensive microenvironments (i.e., diesel plume and urban freeway) is because the traffic fleet in European cities generally had a higher percentage of diesel fuel usage than the cities in Asia (De Jesus et al., 2019). Besides the fuel usage, lack of standardization of measuring instruments (e.g., limit of size range) could also leads to variations in the PNCtot vs. PM2.5 ratios. For those studies conducted PNC measurements with cut-off size lower than 10 nm (for CPC), the median value of the ratios was 0.9 × 109 pt µg–1. On the other hand, the median value of the ratios was 0.48 ×109 pt µg–1 for those studies that using CPCs with cut-off size larger than 10 nm.

The cumulative distribution of the ratio of PNCd,tr vs. ∆PM2.5 and PNCd,sfb vs. ∆PM2.5 were shown in Figs. 5(B) and 5(C), respectively. It is hard to directly compare the results to the literature since most of the studies that measured particulate matters were not evaluating the variability of PNC/PM2.5 ratios in various microenvironments with discerned PNC from different sources. For the ratios of PNCd,tr vs. ∆PM2.5, different sampling campaigns showed similar average median values in diesel plume (0.53 × 109 pt µg–1), urban freeway (0.46 × 109 pt µg–1) and rural freeway (0.26 × 109 pt µg–1) microenvironments, except in the urban street microenvironment. In the urban street microenvironment, the median value of PNCd,tr vs. ∆PM2.5 ratios were 0.8 ×109 pt µg–1 in 2020 holiday campaigns, nearly 2–3 times higher than ratios in 2018 and 2020 non-holiday campaigns. Similarly, the median value of PNCd,sfb vs. ∆PM2.5 ratios in the 2018 campaign were higher than the ratios in the 2020 campaigns in all the microenvironments but not in the urban street microenvironment. In the 2020 holiday campaign, the median value of the PNCd,sfb vs. ∆PM2.5 ratios in the urban street microenvironment (0.2 × 109 pt µg–1) were more than 2 times higher than those in the 2018 and 2020 non-holiday campaigns. Such increases in PNC/PM2.5 ratios (i.e., PNCd,tr vs. ∆PM2.5 and PNCd,sfb vs. ∆PM2.5) could be related to the intensive activities of tourists in the city center during the National Holiday. During the holiday, large numbers of tourists visit the city center with gasoline cars. The fact that the gasoline cars emit particles contribute more to PNC instead of mass concentrations raises the value of PNCd,tr vs. ∆PM2.5 ratios in urban street microenvironments (Ban-Weiss et al., 2009). Moreover, the increases in PNCd,sfb vs. ∆PM2.5 in urban street could probably related to the enhanced restaurant cooking emissions that contribute to particles mostly in the size of 30–50 nm, thus, contribute more to the number concentration than the mass concentration (Zhang et al., 2010; Saha et al., 2019).


There is a large bulk of studies on the characteristics and health effects of PNC and PM2.5 mass concentrations. However, the interrelationship between PNC and PM2.5 across different microenvironments has not been well characterized. The PNC that often used as a proxy measure of UFPs that are substantially affected by combustion processes (e.g., traffic and solid fuel emissions). However, the variations in PM2.5 mass concentration are mostly related to air mass from regional transport (e.g., atmospheric oxidation reactions formed secondary aerosols). This study provides an evaluation on the variability of the PNC and PM2.5 mass concentrations as well as the interrelationship between them in different microenvironments. On average, the PNCtot during 2018 non-holiday (4.8 × 104 pt cm–3) were more than two times higher than 2020 sampling campaigns (0.36 × 104–0.56 × 104 pt cm–3). Significant variability in source-discerned PNC concentrations (i.e., PNCd,tr and PNCd,sfb) were found and can be related to types of microenvironments and seasons. The median PNCd,tr during the 2018 non-holiday exhibited 2-3 times higher concentrations while the median PNCd,sfb during 2018 non-holiday were 7 times higher than the 2020 campaigns. The diesel plumes and urban freeways microenvironments reported higher PNCd,tr across our campaigns, while the urban streets and rural freeways reported higher PNCd,sfb during winter. Traffic emissions were found to contribute about 30% of PNCtot throughout the whole study while emissions of solid fuel burning were found as an important contributor only during winter (21%–29% contribution).

Both of the PM2.5 and ∆PM2.5 mass concentrations were higher during 2018 non-holiday than 2020 campaigns. The average median PM2.5 mass concentration was 109 µg m–3 during 2018 non-holiday which was 3–5 times higher than those during 2020 campaigns. Similarly, the average median ∆PM2.5 mass concentration during 2018 non-holiday (40 µg m–3) was found 3 times higher than 2020 non-holiday and 6 times higher than 2020 holiday campaigns. In addition, the PM2.5 and ∆PM2.5 mass concentrations showed a clear trend during winter (2018 non-holiday) but not during Autumn (2020 campaigns). During 2018 non-holiday, tin median PM2.5 mass concentrations urban street and rural freeway microenvironments (~137 µg m–3) were nearly two times higher than the other microenvironments (~82 µg m–3). Moreover, higher median ∆PM2.5 mass concentrations (~54 µg m–3) were found in urban street and rural freeway microenvironments compared to the diesel plume and urban freeway microenvironments (~27 µg m–3). However, the PM2.5 and ∆PM2.5 mass concentrations in the 2020 campaign (autumn) showed no variability that could be related to microenvironments.

The differences in the variability of PNC and PM2.5 mass concentrations are also reflected in their low correlation coefficient (R2). The average R2 of PNCtot vs. PM2.5 was 0.1 which was 50% lower than the temporal adjusted R2 (i.e., PNCd,tr vs. ∆PM2.5 and PNCd,sfb vs. ∆PM2.5). Seasonal variations in the R2 were found when compared results during winter (2018 non holiday) to autumn (2020 campaigns). The value of R2 during autumn was 0.12, two times higher than the R2 during winter. The raise in R2 due to temporal adjustment implies that it is important to consider source inventory when evaluating the PNC and PM2.5 mass concentrations since they have different transport mechanism (Pierce and Adams, 2007) and possibly distinct emission sources considering the persisted low R2 (0.01–0.34) throughout the campaigns. There were three combinations of PNC/PM2.5 ratios being evaluated. The PNCtot vs. PM2.5 ratios in this study were largely comparable to literature data. No comparison was conducted with PNCd,tr vs. ∆PM2.5 ratios and PNCd,sfb vs. ∆PM2.5 ratios due to the newness of the data. In urban street microenvironments, the two combinations of PNC/PM2.5 ratios in 2020 holiday campaign were 2–3 times higher than 2018 and 2020 non-holiday campaigns due to the impact of holiday events (e.g., intensive tourist activities).

This study was conducted with real-world field measurements in different microenvironments. Thus, results are considered suitable for the evaluation of human exposure in the urban environment. The discrepancy in the trend of PNC and PM2.5 concentrations and their poor correlation underscores how the control policies (e.g., end-of-pipe control policies) for reducing direct emissions of one metric do not necessarily apply to the other. Therefore, achieving more stringent targets to control exposure to PM2.5 mass concentration does not reflect in the mitigation of PNC exposure. Since the PNC is not regulated, relying on the current stationary monitoring networks that measure PM2.5 mass concentrations for exposure assessment is unlikely to fully represent the level of exposure to PNC across the city. Besides, the large seasonal variability in the pollutant concentrations implies that not only the control policies should be diversified with pollutant types, but also should be adjusted with the energy usage of the specific city.

This study does not consider a number of variables that can affect the variations in the pollutant concentrations. First of all, the traffic information was roughly collected in this study. This could bring uncertainties in our results since the effect of traffic pattern on the variability of pollutant concentrations were mainly based on the traffic data during 2020 sampling campaigns. Besides the traffic pattern, there was limited information available regarding the aftertreatment status of the vehicles that operating in the four microenvironments. It known that the application of the diesel oxidation catalyst and the diesel particulate filter could enhance the ultrafine particle (PNC) formation during oxidation reactions (Maricq et al., 2002; Cédric et al., 2016) even with the absence of soot particles (eBC). This could affect the scaling factor (Sd) and the PNC/PM2.5 ratio, especially in the vehicle emission intensive microenvironments (i.e., diesel plume and urban freeway). Lastly, the category of the microenvironments in this study were mainly based on characterization of the sampling routes. Therefore, it is hard to locate the exact location of the major sources based on mobile measurements along. Future studies that combine mobile measurement and the characteristics of local emissions sources (e.g., spatial distribution and emission types) would aid in addressing the bias in emission categorization and provide precise suggestions for control policy decisions.


Author Contributions

S.X., S.Z., and Y.W. conceived the research idea; S.X., H.W., and Y.Y. collected mobile measurement data; Y.D., Q.T., and Z.Z. provided resources in Chengdu. S.X., Y.Y., S.Z., and Y.W. analyzed the data; S.X. and Y.W. wrote the paper with contributions from all the authors.

Acknowledgments and Disclaimer

We are grateful to the National Key Research and Development Program (2022YFC3701800), the National Natural Science Foundation of China (Nos. 52170111 and 41977180) and the China Environmental Protection Foundation’s automobile environmental protection innovation leading plan. The authors acknowledge support from the National Engineering Laboratory for Mobile Source Emission Control Technology (NELMS2019A15). The contents of this paper are solely the responsibility of the authors and do not necessarily represent the official views of the sponsors.

Supporting Information

Demonstration of sampling route and auxiliary data (Section S1), calculation of source-discerned eBC (Section S2), evaluation of DustTrak against TEOM (Section S3), evaluation of scaling factors (Section S4), contribution from difference sources of PNC (Section S5), variations in PM2.5 and ∆PM2.5 (Section S6), Pearson correlation coefficient between PNC and PM2.5 mass concentrations (Section S7) and PNC/PM2.5 ratio (Section S8). Supporting figures (Figs. S1−S6) and tables (Tables S1−S8).


  1. Apte, J.S., Messier, K.P., Gani, S., Brauer, M., Kirchstetter, T.W., Lunden, M.M., Marshall, J.D., Portier, C.J., Vermeulen, R.C., Hamburg, S.P. (2017). High-resolution air pollution mapping with Google street view cars: exploiting big data. Environ. Sci. Technol. 51, 6999–7008.

  2. Apte, J.S., Brauer, M., Cohen, A.J., Ezzati, M., Pope III, C.A. (2018). Ambient PM2.5 reduces global and regional life expectancy. Environ. Sci. Technol. Lett. 5, 546–551.​acs.estlett.8b00360

  3. Ban-Weiss, G.A., Lunden, M.M., Kirchstetter, T.W., Harley, R.A. (2009). Measurement of black carbon and particle number emission factors from individual heavy-duty trucks. Environ. Sci. Technol. 43, 1419–1424.

  4. Bond, T.C., Covert, D.S., Kramlich, J.C., Larson, T.V., Charlson, R.J. (2002). Primary particle emissions from residential coal burning: Optical properties and size distributions. J. Geophys. Res. 107, ICC 9-1–ICC 9-14.

  5. Brantley, H., Hagler, G., Kimbrough, E., Williams, R., Mukerjee, S., Neas, L. (2014). Mobile air monitoring data-processing strategies and effects on spatial air pollution trends. Atmos. Meas. Tech. 7, 2169–2183.

  6. Casquero-Vera, J.A., Lyamani, H., Titos, G., Minguillón, M.C., Dada, L., Alastuey, A., Querol, X., Petäjä, T., Olmo, F., Alados-Arboledas, L. (2021). Quantifying traffic, biomass burning and secondary source contributions to atmospheric particle number concentrations at urban and suburban sites. Sci. Total Environ. 768, 145282.

  7. Cattani, G., Gaeta, A., di Bucchianico, A.D.M., De Santis, A., Gaddi, R., Cusano, M., Ancona, C., Badaloni, C., Forastiere, F., Gariazzo, C. (2017). Development of land-use regression models for exposure assessment to ultrafine particles in Rome, Italy. Atmos. Environ. 156, 52–60.

  8. Cédric, L., Goriaux, M., Tassel, P., Perret, P., André, M., Liu, Y. (2016). Impact of aftertreatment device and driving conditions on black carbon, ultrafine particle and NOx emissions for Euro 5 diesel and gasoline vehicles. Transp. Res. Procedia 14, 3079–3088.​trpro.2016.05.454

  9. Chen, H., Kwong, J.C., Copes, R., Tu, K., Villeneuve, P.J., Van Donkelaar, A., Hystad, P., Martin, R.V., Murray, B.J., Jessiman, B. (2017). Living near major roads and the incidence of dementia, Parkinson's disease, and multiple sclerosis: a population-based cohort study. Lancet 389, 718–726.

  10. Chengdu Municipal Bureau of Planning and Natural Resources (2021). Municipal Planning of Chengdu 2011-2020. Chengdu Municipal Bureau of Planning and Natural Resources, China.

  11. Chengdu Bureau of Statistics (2019). Chengdu Statistical Yearbook. Chengdu Bureau of Statistics, China.

  12. De Jesus, A.L., Rahman, M.M., Mazaheri, M., Thompson, H., Knibbs, L.D., Jeong, C., Evans, G., Nei, W., Ding, A., Qiao, L. (2019). Ultrafine particles and PM2.5 in the air of cities around the world: Are they representative of each other? Environ Int. 129, 118–135.​envint.2019.05.021

  13. Drinovec, L., Močnik, G., Zotter, P., Prévôt, A., Ruckstuhl, C., Coz, E., Rupakheti, M., Sciare, J., Müller, T., Wiedensohler, A. (2015). The" dual-spot" Aethalometer: an improved measurement of aerosol black carbon with real-time loading compensation. Atmos. Meas. Tech. 8, 1965–1979.

  14. Duan, X., Jiang, Y., Wang, B., Zhao, X., Shen, G., Cao, S., Huang, N., Qian, Y., Chen, Y., Wang, L. (2014). Household fuel use for cooking and heating in China: results from the first Chinese Environmental Exposure-Related Human Activity Patterns Survey (CEERHAPS). Appl. Energy 136, 692–703.

  15. Eeftens, M., Phuleria, H.C., Meier, R., Aguilera, I., Corradi, E., Davey, M., Ducret-Stich, R., Fierz, M., Gehrig, R., Ineichen, A. (2015). Spatial and temporal variability of ultrafine particles, NO2, PM2.5, PM2.5 absorbance, PM10 and PMcoarse in Swiss study areas. Atmos. Environ. 111, 60-–70.

  16. Fernández-Camacho, R., Rodríguez, S., De La Rosa, J., Sánchez De La Campa, A.M., Viana, M., Alastuey, A., Querol, X. (2010). Ultrafine particle formation in the inland sea breeze airflow in Southwest Europe. Atmos. Chem. Phys. 10, 9615–9630.

  17. González, Y., Rodríguez, S., García, J.C.G., Trujillo, J.L., García, R. (2011). Ultrafine particles pollution in urban coastal air due to ship emissions. Atmos. Environ. 45, 4907–4914.​10.1016/j.atmosenv.2011.06.002

  18. Health Effects Institute (HEI) (2013). Understanding the health effects of ambient ultrafine particles HEI Perspectives 3. Health Effects Institute, Boston, MA, 122.

  19. Hinds, W.C. (1999). Aerosol technology: properties, behavior, and measurement of airborne particles. John Wiley & Sons.

  20. Hoek, G., Beelen, R., Kos, G., Dijkema, M., Zee, S.C.v.d., Fischer, P.H., Brunekreef, B. (2011). Land use regression model for ultrafine particles in Amsterdam. Environ. Sci. Technol.45, 622–628.

  21. Larson, T., Gould, T., Riley, E.A., Austin, E., Fintzi, J., Sheppard, L., Yost, M., Simpson, C. (2017). Ambient air quality measurements from a continuously moving mobile platform: Estimation of area-wide, fuel-based, mobile source emission factors using absolute principal component scores. Atmos. Environ. 152, 201–211.

  22. Levy, I., Mihele, C., Lu, G., Narayan, J., Brook, J.R. (2014). Evaluating multipollutant exposure and urban air quality: pollutant interrelationships, neighborhood variability, and nitrogen dioxide as a proxy pollutant. Environ. Health Perspect. 122, 65–72.​1306518

  23. Marconi, A., Cattani, G., Cusano, M., Ferdinandi, M., Inglessis, M., Viviano, G., Settimo, G., Forastiere, F. (2007). Two-years of fine and ultrafine particles measurements in Rome, Italy. J. Toxicol. Environ. Health Part A 70, 213–221.

  24. Maricq, M.M., Chase, R.E., Xu, N., Laing, P.M. (2002). The effects of the catalytic converter and fuel sulfur level on motor vehicle particulate matter emissions: Light duty diesel vehicles. Environ. Sci. Technol.36, 283–289.

  25. Mejía, J.F., Morawska, L., Mengersen, K. (2008). Spatial variation in particle number size distributions in a large metropolitan area. Atmos. Chem. Phys. 8, 1127–1138.​10.5194/acp-8-1127-2008

  26. Meng, W., Zhong, Q., Chen, Y., Shen, H., Yun, X., Smith, K.R., Li, B., Liu, J., Wang, X., Ma, J. (2019). Energy and air pollution benefits of household fuel policies in northern China. Proc. Natl. Acad. Sci. U.S.A. 116, 16773–16780.

  27. Morawska, L., Thomas, S., Bofinger, N., Wainwright, D., Neale, D. (1998). Comprehensive characterization of aerosols in a subtropical urban atmosphere: particle size distribution and correlation with gaseous pollutants. Atmos. Environ. 32, 2467–2478.​S1352-2310(98)00023-5

  28. Peters, A., Veronesi, B., Calderón-Garcidueñas, L., Gehr, P., Chen, L.C., Geiser, M., Reed, W., Rothen-Rutishauser, B., Schürch, S., Schulz, H. (2006). Translocation and potential neurological effects of fine and ultrafine particles a critical update. Part. Fibre Toxicol. 3, 1–13.

  29. Pierce, J., Adams, P. (2007). Efficiency of cloud condensation nuclei formation from ultrafine particles. Atmos. Chem. Phys. 7, 1367–1379.

  30. Pope III, C.A., Ezzati, M., Dockery, D.W. (2009). Fine-particulate air pollution and life expectancy in the United States. N. Engl. J. Med. 360, 376–386.

  31. Reche, C., Querol, X., Alastuey, A., Viana, M., Pey, J., Moreno, T., Rodríguez, S., González, Y., Fernández-Camacho, R., De la Rosa, J., Dall'Osto, M. (2011). New considerations for PM, Black Carbon and particle number concentration for air quality monitoring across different European cities. Atmos. Chem. Phys. 11, 6207–6227.

  32. Rodríguez, S., Cuevas, E. (2007). The contributions of “minimum primary emissions” and “new particle formation enhancements” to the particle number concentration in urban air. J. Aerosol Sci. 38, 1207–1219.

  33. Rodríguez, S., Van Dingenen, R., Putaud, J.P., Dell'Acqua, A., Pey, J., Querol, X., Alastuey, A., Chenery, S., Ho, K.F., Harrison, R. (2007). A study on the relationship between mass concentrations, chemistry and number size distribution of urban fine aerosols in Milan, Barcelona and London. Atmos. Chem. Phys. 7, 2217–2232.

  34. Rönkkö, T., Kuuluvainen, H., Karjalainen, P., Keskinen, J., Hillamo, R., Niemi, J.V., Pirjola, L., Timonen, H.J., Saarikoski, S., Saukko, E. (2017). Traffic is a major source of atmospheric nanocluster aerosol. Proc. Natl. Acad. Sci. U.S.A. 114, 7549–7554.

  35. Saha, P.K., Khlystov, A., Snyder, M.G., Grieshop, A.P. (2018). Characterization of air pollutant concentrations, fleet emission factors, and dispersion near a North Carolina interstate freeway across two seasons. Atmos. Environ. 177, 143–153.​01.019

  36. Saha, P.K., Zimmerman, N., Malings, C., Hauryliuk, A., Li, Z., Snell, L., Subramanian, R., Lipsky, E., Apte, J.S., Robinson, A.L. (2019). Quantifying high-resolution spatial variations and local source impacts of urban ultrafine particle concentrations. Sci. Total Environ. 655, 473–481.

  37. Saha, P.K., Sengupta, S., Adams, P., Robinson, A.L., Presto, A.A. (2020). Spatial correlation of ultrafine particle number and fine particle mass at urban scales: implications for health assessment. Environ. Sci. Technol. 54, 9295–9304.

  38. Sandradewi, J., Prévôt, A., Weingartner, E., Schmidhauser, R., Gysel, M., Baltensperger, U. (2008a). A study of wood burning and traffic aerosols in an Alpine valley using a multi-wavelength Aethalometer. Atmos. Environ. 42, 101–112.

  39. Sandradewi, J., Prévôt, A.S., Szidat, S., Perron, N., Alfarra, M.R., Lanz, V.A., Weingartner, E., Baltensperger, U. (2008b). Using aerosol light absorption measurements for the quantitative determination of wood burning and traffic emission contributions to particulate matter. Environ. Sci. Technol.42, 3316–3323.

  40. Tao, S., Ru, M., Du, W., Zhu, X., Zhong, Q., Li, B., Shen, G., Pan, X., Meng, W., Chen, Y. (2018). Quantifying the rural residential energy transition in China from 1992 to 2012 through a representative national survey. Nat. Energy 3, 567–573.

  41. Tong, Z., Li, Y., Lin, Q., Wang, H., Zhang, S., Wu, Y., Zhang, K.M. (2022). Uncertainty investigation of plume-chasing method for measuring on-road NOx emission factors of heavy-duty diesel vehicles. J. Hazard. Mater. 424, 127372.

  42. Wang, F., Cernuschi, S., Ozgen, S., Ripamonti, G., Vecchi, R., Valli, G., Lonati, G. (2016). UFP and BC at a mid-sized city in Po valley, Italy: Size-resolved partitioning between primary and newly formed particles. Atmos. Environ. 142, 120–131.​030

  43. Wang, D., Li, Q., Shen, G., Deng, J., Zhou, W., Hao, J., Jiang, J. (2020). Significant ultrafine particle emissions from residential solid fuel combustion. Sci. Total Environ. 715, 136992.

  44. Wang, J.M., Jeong, C.H., Hilker, N., Shairsingh, K.K., Healy, R.M., Sofowote, U., Debosz, J., Su, Y., McGaughey, M., Doerksen, G. (2018). Near-road air pollutant measurements: accounting for inter-site variability using emission factors. Environ. Sci. Technol. 52, 9495–9504.​10.1021/acs.est.8b01914

  45. Wang, X., Chancellor, G., Evenstad, J., Farnsworth, J.E., Hase, A., Olson, G.M., Sreenath, A., Agarwal, J.K. (2009). A novel optical instrument for estimating size segregated aerosol mass concentration in real time. Aerosol Sci. Technol. 43, 939–950.​02786820903045141

  46. World Health Organisation (WHO) (2021). WHO global air quality guidelines: particulate matter (PM2.5 and PM10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide. World Health Organization.

  47. Wolf, K., Cyrys, J., Harciníková, T., Gu, J., Kusch, T., Hampel, R., Schneider, A., Peters, A. (2017). Land use regression modeling of ultrafine particles, ozone, nitrogen oxides and markers of particulate matter pollution in Augsburg, Germany. Sci. Total Environ. 579, 1531–1540.

  48. Xiang, S., Yu, Y.T., Hu, Z., Noll, K.E. (2019a). Characterization of dispersion and ultrafine-particle emission factors based on near-roadway monitoring Part I: Light duty vehicles. Aerosol Air Qual. Res. 19, 2410–2420.

  49. Xiang, S., Yu, Y.T., Hu, Z., Noll, K.E. (2019b). Characterization of dispersion and ultrafine-particle emission factors based on near-roadway monitoring Part II: Heavy duty vehicles. Aerosol Air Qual. Res. 19, 2421–2431.

  50. Xiang, S., Zhang, S., Wang, H., Wen, Y., Yu, Y.T., Li, Z., Wallington, T.J., Shen, W., Deng, Y., Tan, Q., Zhou, Z., Wu, Y. (2021). Mobile measurements of carbonaceous aerosol in microenvironments to discern contributions from traffic and solid fuel burning. Environ. Sci. Technol. Lett. 8, 867–872.

  51. Xiang, S., Zhang, S., Wang, H., Yu, Y.T., Wallington, T.J., Shen, W., Kirchner, U., Deng, Y., Tan, Q., Zhou, Z., Wu, Y. (2022). Variability of NO2/NOx ratios in multiple microenvironments from on-road and near-roadway measurements. ACS EST Eng. 2, 1599–1610.​acsestengg.2c00027

  52. Xiang, S., Zhang, S., Yu, Y.T., Wang, H., Shen, Y., Zhang, Q., Wang, Z., Wang, D., Tian, M., Wang, J., Yin, H., Jiang, J., Wu, Y. (2023). Evaluation of the relationship between meteorological variables and NOx emission factors based on plume-chasing measurements. ACS EST Eng. 3, 417–426.

  53. Yu, Y.T., Xiang, S., Li, R., Zhang, S., Zhang, K.M., Si, S., Wu, X., Wu, Y. (2022a). Characterizing spatial variations of city-wide elevated PM10 and PM2.5 concentrations using taxi-based mobile monitoring. Sci. Total Environ. 154478.

  54. Yu, Y.T., Xiang, S., Zhang, T., You, Y., Si, S.C., Zhang, S.J., Wu, Y. (2022b). Evaluation of city-scale disparities in PM2.5 exposure using hyper-localized taxi-based mobile monitoring. Environ. Sci. Technol. 56, 13584–13594.

  55. Zhang, Q., Gangupomu, R.H., Ramirez, D., Zhu, Y. (2010). Measurement of ultrafine particles and other air pollutants emitted by cooking activities. Int. J. Environ. Res. Public Health 7, 1744–1759.

  56. Zotter, P., Herich, H., Gysel, M., El-Haddad, I., Zhang, Y., Močnik, G., Hüglin, C., Baltensperger, U., Szidat, S., Prévôt, A.S. (2017). Evaluation of the absorption Ångström exponents for traffic and wood burning in the Aethalometer-based source apportionment using radiocarbon measurements of ambient aerosol. Atmos. Chem. Phys. 17, 4229–4249.​10.5194/acp-17-4229-2017 

Share this article with your colleagues 


Subscribe to our Newsletter 

Aerosol and Air Quality Research has published over 2,000 peer-reviewed articles. Enter your email address to receive latest updates and research articles to your inbox every second week.

77st percentile
Powered by
   SCImago Journal & Country Rank

2022 Impact Factor: 4.0
5-Year Impact Factor: 3.4

Aerosol and Air Quality Research partners with Publons

CLOCKSS system has permission to ingest, preserve, and serve this Archival Unit
CLOCKSS system has permission to ingest, preserve, and serve this Archival Unit

Aerosol and Air Quality Research (AAQR) is an independently-run non-profit journal that promotes submissions of high-quality research and strives to be one of the leading aerosol and air quality open-access journals in the world. We use cookies on this website to personalize content to improve your user experience and analyze our traffic. By using this site you agree to its use of cookies.