Characterizing Traffic-Related Ultrafine Particles in Roadside Microenvironments: Spatiotemporal Insights from Industrial Parks

Ultrafine particles (UFPs; PM 0.1 ) and black carbon (BC) were measured at different roadside microenvironments in the vicinity of the urban industrial park area. Simultaneous measurement campaigns were conducted at industrial roadside (IN), residential roadside (RS)


INTRODUCTION
Current studies have highlighted the impacts of ultrafine particles (UFPs) and black carbon (BC) on human health (HEI, 2013;Silva et al., 2022;Zhang et al., 2022;Zhu et al., 2023).Chronic exposure to these pollutants has been linked with cardiovascular disease (CVD), chronic obstructive pulmonary disease (COPD), pre-term birth, and congestive health failures.Typically, the emitted particles which considered as UFPs are particulates with diameter size below 100 nm (Abdillah and Wang, 2023).UFPs may originate from various anthropogenic emission sources including industrial

Overview of Study Sites
In this study, measurements were conducted at different types of roadside environments in industrial park area located in Zhongli district (24°57′25′′N, 121°13′25′′E), Taoyuan City, Taiwan (Fig. 1(a)).It has a total area of 76.52 km 2 with total population around 397,083 people per 2017 (TYCG, 2021).This district is included in one of the designated sites for Taoyuan Industrial Park developmental plan (Fig. 1(b)).Taoyuan houses a diverse range of industry clusters from food manufacturing, textile, chemical, automotive, logistics, aviation, optoelectronics, and biotechnology (TYCG, 2021).The industrial park vicinity covers two district profiles of industrial and residential areas where both are connected by different roads (Figs.1(c-d)).Between both areas, there is a community park which is designated as urban background area in this study.The average daily traffic volume on the industrial roadside can reach up to 12,000 vehicles day -1 and 3,500 vehicles day -1 on the residential roadside.The main difference between these two roadside profiles is their emission sources profiles.Industrial activities and heavy-duty vehicles mainly contributed to the pollutant concentrations in the industrial roadside.On the other hand, the primary emission source in the residential roadside come from different vehicle composition such as small trucks, pick-ups, and motorcycles.

Measurement Campaigns
Seasonal data collection campaigns which consisted of 6 weeks of measurement activities were conducted at three corresponding sampling sites (Fig. 1).The measurement activities were rolled out on 4-24 July 2022 for the warm period and 31 October-20 November 2022 for the cold period.Daily measurement for UFPs PNC, BC mass concentration, meteorological conditions, and vehicle activities were conducted continuously at each of the respective sampling sites by two pairs of personnels (Table S1).The daily measurement durations were 10-hours (08.00 AM-18.00PM) and pollutant concentrations were collected at 1-minute measurement interval.In total, 3-weeks of daily measurements data were obtained for each seasonal measurement period, with 1-week data collected for urban background (UB) and 2 weeks of simultaneous measurement data for the industrial (IN) and residential roadside (RS).
Measurements for IN was conducted at Jilin N. Road, Zhongli District, Taoyuan City, 320, Taiwan.This roadside was chosen as it exists adjacent to several industrial plant facilities which become the main part of Zhongli Industrial Park.As for RS, the measurement was done in Jili 11 th Road, Zhongli District, located approximately 1 km from the outer part of the industrial park.UB measurement was carried out in Zhongshan community park located at the center of IN and RS (Fig. 2).No measurement activities were conducted during rain to preserve the mobile sampling instruments used in this study.

Measurement Instruments and Calibrations
Several high-time-resolution measurement instruments for UFPs and BCs were put inside a housing platform which has been fixed at each roadside microenvironment prior to daily monitoring activities.The distance between the housing platform (measurement point) and the roads was set to 1.25 meters.The measurement inlet was 150 cm above ground (breathing height of average Asian adults).In UB, instrument housing was set up on the gazebo platform in the middle of the community park.Portable external power sources were used to maintain the portable instruments lifetime.All instruments have been calibrated according to their factory standards, and their inter-comparability has been ensured prior to the measurement activities.The list of measurement instruments and their associated parameters are summarized in Table S2.

UFPs particle number concentrations (UFPs PNC)
A pair of portable condensation particle counters (TSI CPC 3007) was used to measure the number concentrations of ultrafine particles (# cm -3 ) during the study.This particle counter instrument could quantify UFPs PNC with sizes ranging from 10 nm-1000 nm.It operates in the concentration range between 1-100,000 # cm -3 .The instruments principle works by relying on the surface condensation process of ultrafine particles.Afterward, condensed ultrafine particles will be counted by laser optical counter at downstream part of the instrument.

Black carbon (BC)
The BC mass concentration (ng m -3 ) was measured by utilizing portable Aethalometer (MicroAeth AE51).This instrument estimates the BC concentration through the light absorption principle.Briefly, BC was estimated by dividing the value of attenuated light intensity (I) with the amount of light intensity which is passing through the clean filter paper (I0), as written in Eq. (1).(1) The estimated equivalent BC mass concentration value was then calculated by MicroAeth AE51 instrument by the infrared light attenuation at 880 nm wavelength by Eq. ( 2): ( ) where, σATN represents mass attenuation coefficient of BC, Q represents inlet flow rate (m 3 s -1 ) of MicroAeth AE51, A represents filter paper area, and Δt represents the measurement time.
Acquired BC concentration data were processed prior to further analysis to compensate optical noise generated from loading effect on filter paper that led to a decreased value of ΔATN (negative measurement results).Therefore, an R package which could process the BC dataset by using optical noise-reduction averaging algorithm (ONA) was applied to correct the data (Hagler et al., 2011;Sanjeev, 2020).In short, ONA recognize the ΔATN between consecutive readings and removes data spikes associated with ΔATN < 0.05.ONA then averages BC mass concentration data throughout time intervals associated with ΔATN = 0.05.

Meteorological data
Hourly meteorological conditions data including relative humidity (RH), wind direction (WD), wind speed (WS), and temperature were obtained from Taiwan Central Weather Bureau (TCWB) observation data platform (TCWB, 2022).The acquired weather data was selected from the online monitoring results of the local weather station which was located near to the study area.

Traffic activities and vehicle compositions
Throughout the whole measurement campaign, two pairs of personnel manually recorded traffic compositions using digital tally at each roadside microenvironment.Recorded traffic activities data were categorized into heavy vehicles, gasoline cars, and motorcycles.It should be noted that the profile of heavy vehicles in industrial roadside was completely different than in residential roadside area.Heavy vehicles at IN consisted of heavy-duty containers, city buses, industrial vehicles (excavator, forklifts, tractors), and trucks.On the other hand, heavy vehicles at RS consisted of only small trucks and minibuses which occasionally drove across the road.

Data Analysis
Descriptive statistic method was used to organize all collected data and visualize the spatiotemporal variability of UFPs PNC and BCs.Obtained data were dynamically averaged at 1-h time resolution by conducting t-test of variance analysis.Outliers (outside 95% CI) were also removed prior to analysis.Pearson's r correlation was used to analyze the correlations between potential confounding factors (traffic activities and meteorological data) and observed pollutants concentrations.In addition, interquartile range/median ratio (IQR/M) ratio analysis was deployed to investigate the existence of intra-urban spatial variability of the roadside ultrafine particles (Rakowska et al., 2014;Wu et al., 2015).Briefly, intra-urban spatial variability between each measurement site could be identified if the difference of IQR/M ratio are ≥ 1.5 (Wu et al., 2015).Figures and statistical analysis results were generated from OriginPro software (OriginLab, v2022).

Seasonal Variability of Roadside UFPs and BCs
Initially, the overall concentrations of UFPs PNC and BC throughout the whole measurement campaigns were varied between three corresponding sampling sites.IN exhibited the highest average of UFPs PNC and BC concentration (38,000 ± 9,000 # cm -3 and 25,00 ± 600 × 10 3 ng m -3 ) and then it was followed by RS with (25,000 ± 8,000 # cm -3 and 1,900 ± 300 ng m -3 ), and UB with (23,000 ± 9,000 # cm -3 and 1,400 ± 300 ng m -3 ).However, throughout different seasons UFPs PNC and BC concentrations were obviously higher during cold period compared to warm period.The average meteorological conditions observed in this study are presented in Table 1.The average UFPs PNC and BC in IN during cold period could reach up to 41,200 ± 10,400 # cm -3 and 2,720 ± 610 ng m -3 , which was significantly higher compared to warm period with 35,500 ± 8,600 # cm -3 and  2,260 ± 460 ng m -3 (Table S3).In overall, 11.26%-16.06%and 20.35%-24.32% increase of average UFPs PNC and BC mass concentrations were identified during cold period at all measurement sites.Furthermore, the measurement results have at three sites have also shown a substantial increase in the distribution of average hourly concentrations for UFPs PNC and BC (Figs.S1(a-d)).The underlying reasons on why the pollutants concentrations were higher during cold period might be related to the characteristics of atmospheric behavor in a lower temperature condition.
During typical winter or cold conditions, lower levels of atmosphere exhibits higher condensation sink and lower solar intensity than in other seasons (Young et al., 2012;Pikridas et al., 2015).Therefore, the conditions could inhibit the rate of particle growth in the upper atmosphere, leading to a higher concentrations of pollutants (Argyropoulos et al., 2016;Cheung et al., 2016).In addition, the temperature inversion phenomena which often occur during the cold seasons may suppress vertical transport and vertical mixing process of the particles via stable stratification process (Li et al., 2019;Rose et al., 2021).On the other hand, temperatures during warm conditions vary more widely within a day, leading to an intensified vertical mixing effects compared to cold season (Agudelo-Castañeda et al., 2019).In summary, lower temperature during cold period may promote the accumulation of pollutants.Hence, it may explain the seasonality of UFPs PNC and BC observed in this study.
Furthermore, similar seasonality characteristics were also identified in several urban settings.For instance, Cheung et al. (2016) reported a higher level of UFPs PNC during winter (17,400 ± 4,700 # cm -3 ) compared to summer (16,700 ± 7,600 # cm -3 ) in ambient air of Taipei, Taiwan.On the other hand, Kurppa et al. (2015) have also reported a substantial increase of UFPs PNC during winter in Helsinki, Finland.These findings could further highlight the effects of seasonality on the concentrations of roadside UFPs PNC and BC.As have been previously mentioned, the occurrence of seasonal variations on the UFPs PNC is highly related to the solar intensity and condensation sink conditions of the atmosphere (Pikridas et al., 2015;Cheung et al., 2016;Chen et al., 2017).On the other hand, solar intensity may also vary throughout diurnal cycle, when highest solar intensity can be observed at noon and lowest at midnight (Lee et al., 2021;Young et al., 2023).In the cold period, all sampling sites (IN, RS, UB) demonstrated strong peaks for UFPs PNC and BC as the result of lower atmospheric ventilation (Figs.3(a-f)).Interestingly, the semi-diurnal profiles of observed UFPs PNC and BC during winter were not significantly different compared to warm period apart from their concentrations level (Figs.3(a-f)).However, it is important to notice that IN and RS exhibited distinct semi-diurnal concentrations trends compared to UB.
IN and RS possessed similar semi-diurnal patterns where strong peaks were identified during typical morning and evening rush hours at the period of 08.00-11.00and 15.00-18.00(Figs.3(a-d)).However, the peak concentration at UB was only identified during the noon period and it gradually decreased until evening (Figs.3(e-f)).The average semi-diurnal profiles of UFPs PNC and BC at IN and RS indicated the combined effects of traffic activities and atmospheric ventilation patterns (Gani et al., 2021).On the other hand, the UFPs PNC trend in urban background might follow daily variation of solar intensity, indicating that majority of the observed PNC were from secondary emission process (Nie et al., 2022).This finding is notably different from other typical roadside  UFPs studies where they only identified similar emission profile in all measurement sites (Tran et al., 2022;Rajagopal et al., 2023).

Peak Periods Concentrations
Figs. 4(a-b) highlighted the weekly variations of average UFPs PNC and BC concentrations during warm and cold periods.The daily hourly average of UFPs PNC and BC concentrations gradually increased from Monday to Friday in the weekdays period.Afterwards, the concentrations were significantly dropped during the weekend period for both pollutants.For instances, the UFPs PNC and BC in warm period at IN during weekdays could reach up to 37,700 ± 7,400 # cm -3 and 2,350 ± 500 ng m -3 , whereas during weekend the average was only 29,900 ± 8,900 # cm -3 and 2,050 ± 280 ng m -3 (Table 2).It was also observed that the weekly variations of UFPs PNC and BC at IN were similar with RS.However, UB exhibited distinct trend that might be attributed to the difference of emission profile.
The seasonality of UFPs PNC and BC could also be identified in the weekly concentration variations of these pollutants (Figs.4(a-b)).For examples, the average weekdays UFPs PNC and BC concentrations at IN during winter could reach up to 45,800 ± 8,700 # cm -3 and 2,850 ± 650 ng m -3 , which was obviously higher than the weekdays concentration during warm period with 37,700 ± 7,400 # cm -3 and 2,350 ± 500 ng m -3 .The same instances were also identified during weekend where the average concentrations in residential roadside reached up to 20,400 ± 4,400 # cm -3 and 1,810 ± 270 ng m -3 during winter and 17,200 ± 4,900 # cm -3 and 1,480 ± 260 ng m -3 .These findings indicated specific emission patterns of UFPs PNC and BC during certain peak periods throughout the week.
Furthermore, observed roadside UFPs PNC and BC concentrations in this study exhibited distinct concentration patterns during morning (08.00-11.00)and afternoon (15.00-18.00)peak/rush periods (Figs.5(a-d)).In overall, measured UFPs PNC and BC concentrations were considerably higher during peaks period compared to non-peak period.Average UFPs PNC during morning peaks period at IN were 14.73% and 23.65% higher compared to non-peak periods in warm and cold periods.Additionally, morning peak BC concentrations at IN were also 11.39% and 19.73%  higher.This result is relatively similar to previous on-road study in the urban road of Taipei, Taiwan.It was reported that during rush hour period (08.00-10.30and 17.00-19.30),the amount of UFPs PNC and BC concentrations significantly increased up to 34.80% and 31.80%(Lin et al., 2022).Furthermore, ~95% increase of traffic-related air pollutants (TRAPs) concentrations was also identified in the roadside area of urban highway in Oakland, USA during morning rush hour period (Gani et al., 2021).The substantial increase during this peak period might reflected the dependence of UFPs PNC and BC concentrations towards the instance of traffic composition and flux during that specific period (Wang et al., 2008;Bergmann et al., 2022).
Apart from that, it can be inferred that the average concentrations of UFPs PNC and BC during morning peaks were considerably higher compared to afternoon peak at both IN and RS.Similar studies suggested that the increased concentrations during morning peak was due to the combination effects of peak vehicle flux and intensified solar irradiation rate which promote new particle formation (NPF) event in the roadside (Cheung et al., 2016;Gani et al., 2021;Nie et al., 2022).On the other hand, the peak hours variation of BC concentrations might be solely attributed to the higher number of vehicle flux observed in the morning period compared to the afternoon period (Lin et al., 2022).These findings further suggest the evidence of peaks period variations of UFPs PNC and BC concentrations that were successfully observed in this study.

Intra-Urban Spatial Variability of Roadside UFPs and BCs
Fig. 6 portrayed the IQR/M ratio distributions between UFPs PNC and BC mass concentrations at three corresponding measurement sites.Intra-urban spatial variability of UFPs PNC between roadside sites (IN and RS) and UB were identified proven by the IQR/M ratio difference of > 1.5.Furthermore, BC in UB also exhibited intra-spatial variability with IQR/M ratio of 1.5 times higher than IN.This identified characteristic was retained through the cold period for UFPs PNC while for BC, it was disappeared during the season change.It is important to note that the approximate  distance between each sampling site is approximately less than 1 km (Fig. 1).These findings highlighted the potential occurrence of intra-urban spatial variability in this study.
Interestingly, the result of this analysis could be used to explain the difference between concentrations level identified at roadsides (IN and RS) and UB.The distinct concentration variations among these three sites could be described by the intra-urban spatial variability characteristic that was possessed by the UFPs PNC.This result is consistent with Wu et al. (2015) where they reported intra-spatial variability between roadside environment with IQR/M ratio difference of up to 3 times higher compared to the other sites.This finding may imply that roadside UFPs PNC in specific roadside microenvironments such as in urban industrial park areas might possess different types of spatial heterogeneity compared to other traditional types of pollutant (e.g., PM10, PM2.5, NOx, SOx).

Potential Influences of Meteorological Conditions and Traffic Activities
In typical urban microenvironments, the behaviors of UFPs PNC and BC are heavily dependent on various factors, specifically traffic activities and meteorological conditions such as relative humidity, wind direction, wind speed.Similarly, it was hypothesized from seasonality and peak periods analysis that UFPs PNC and BC at IN and RS were originated from traffic activities, while the concentrations at UB was linked to secondary particle emission.The secondary particle emissions process of UFPs PNC are highly influenced by the concentrations of gaseous precursors and specific meteorological conditions (Abdillah and Wang, 2023).Hence, these two factors were considered as the main confounding factors that might influence observed pollutant concentrations in this study.
The summary of observed meteorological conditions and semi-diurnal traffic profiles in this study are presented in Fig. S2, Fig. S3, and Fig. S4.The effect of meteorological conditions to UFPs PNC was more visible at UB, particularly.The semi-diurnal profile of RH was inversely proportional to UFPs concentration at UB (Fig. S2 and Fig. 3(e)).Typically, the variation of RH is also inversely correlated with solar radiation.RH and solar radiation normally played a significant role during NPF process of secondary UFPs (Lee et al., 2021;Zilli Vieira and Koutrakis, 2021).In addition, the scatter plot correlation between UFPs PNC and BC at UB (Fig. S5) indicated that there was no significant correlation between UFPs PNC and BC (r = -0.04).These findings have further highlighted the distinct source variation and meteorological effects which occurred at UB during the study period.The observed UFPs PNC at UB would likely consist of secondary UFPs that were influenced by the conditions of RH and solar radiation during NPF process.
On the other hand, influences from traffic related activities on UFPs PNC and BC were obviously identified at IN and RS.Significant correlations were found between UFPs PNC and BC at both IN (r = 0.80, ρ < 0.005) and RS (r = 0.87, ρ < 0.05) (Fig. S5).BC is the main tracer pollutant of vehicle exhaust emission from traffic activities (Tang et al., 2022).Furthermore, the peaks in vehicles fluxes at both roadside microenvironments were also simultaneously identified during the daily peak periods of UFPs PNC and BC at morning (08.00-11.00)and evening rush (15.00-18.00)periods (Fig. S4 and Figs. 3(a-d)).The highest average flux and UFPs PNC at both roadsides were identified during peak morning rush period (09.00-10.00)with 1,288-1,409 vehicle/hour and 59,500 # cm -3 and 221-262 vehicle/hour and 41,100 # cm -3 for IN and RS during cold season.Furthermore, the average UFPs PNC and BC concentration were considerably higher during weekdays at IN and RS (Table 2) in conjunction with the weekly total vehicle flux per day during weekdays (Fig. S6).These findings further highlight the potential influence of traffic activities to the pollutants.
One of the main strengths of this study is related to the different traffic composition observed at IN and RS.In order to assess the extent of contributions for each type of vehicle towards observed UFPs PNC and BC, Pearson's r statistical method was deployed throughout the measurement datasets (Fig. 7).In overall, UFPs PNC and BC possessed positive significant correlation with number of vehicles observed throughout the whole study (ρ ≤ 0.01).For instance, UFPs PNC and BC at IN and RS were significantly correlated with total hourly vehicle flux (IN: r = 0.64 and 0.56, ρ ≤ 0.01) and (RS: r = 0.80 and 0.70, ρ ≤ 0.01).More specifically, UFPs PNC and BC at IN exhibited significant correlations with different types of traffic profiles in the strength order of heavy vehicles (r = 0.69 and 0.55, ρ ≤ 0.01) > motorcycles (r = 0.63 and 0.61, ρ ≤ 0.01) > gasoline vehicles (r = 0.40 and 0.29).Whereas on the other hand, UFPs PNC and BC at RS were positively correlated with the From this result, it can be noted that emission from heavy industrial vehicles might have dominated the UFPs PNC and BC at IN, while the UFPs PNC and BC at RS was possibly dominated by gasoline vehicles (Fig. S7).These findings imply that UFPs PNC and BC concentrations are heavily dependent on the source emission profiles and might not be constrained to the spatial scale of the urban environment.The spatial variability of UFPs PNC could exist even in the small and specific urban microenvironment such as in the different roadside profiles of urban industrial park area (Cassee et al., 2019).Furthermore, modest correlations between WS and RH and the pollutants were also identified at IN and RS.This result could highlight the possibility of dispersion/dilution phenomena that occurred due to the wind turbulences generated either from traffic activities or wind speed and direction (Kumar et al., 2008).The significantly inversed proportional correlations between RH and UFPs PNC at UB (Fig. 7) could also possibly explain the influences of RH to the secondary emission process that were discussed earlier.These findings have further highlighted the potential confounding effects of traffic activities and meteorological conditions on the roadside UFPs and BC concentrations.

STUDY LIMITATIONS AND IMPLICATIONS FOR FUTURE RESEARCH
Similar to other studies which used portable measuring instruments, there are several limitations in this study.First, due to instrument constraints, this study only focuses on quantifying UFPs PNC for particles above 10 nm and does not include any discussion about nanoparticles with diameter below 5 nm.Hence, the UFPs PNC measured at UB might be underestimated as the dominant size of aged UFPs from secondary emission are in the range of 3-15 nm.In addition, traffic activities might also emit very small particles in the range of 2 to < 10 nm as results of aerosol and gas-to-particle phase emissions process of fuel burning.Hence, more adequate instruments and analysis methods such as SMPS and chemical content analysis of the UFPs are needed in the future.Correction algorithm was used to minimize the uncertainty generated from data noise for BC measurement.Furthermore, as various factors may affect the spatiotemporal variability of UFPs PNC, IQR/M ratio analysis method at three different sites might not be strong enough to confirm the characteristic.Further studies with more sampling points, longer time resolutions (i.e., 24-hour and multi-year sampling), and chemical characterizations are essential in the future.
Regardless of the limitations, this study has successfully reported the seasonal variations of UFPs PNC and BC concentrations at different roadside microenvironments in the vicinity of urban industrial park area.Observed results from this study have highlighted elevated risk of exposure in urban industrial park areas compared to other roadside microenvironments studies (Table 3).Based on the findings from this study and recent roadside studies, it can be noted that the average UFPs PNC concentrations in the roadside environments have exceeded the "high level" (> 20,000 # cm -3 at 1-h) according to the best practice statement of UFPs in WHO AQG 2021(WHO, 2021).In addition, annual BC level above 1.15 µg m -3 were stated to be the one of the major factors that is associated with adverse health outcomes (WHO, 2021), and it is far below the observed roadside BC level from this study (2.2 µg m -3 ) and other studies.This would imply that roadside environments possess intensified exposure risks toward pedestrians and residents in a typically high-density building in urban environments.Future traffic-related UFPs exposure and health impact assessment (cohort) studies in specific roadside environments are essential to address this potential issue.
Furthermore, it is identified from this study that UFPs possessed intra-urban spatial variability characteristic which is different from traditional PMs.In other words, it emphasized the importance for different policies/approaches for managing resident and pedestrian exposure to these pollutants of concern.Currently, there are no specific guidelines for UFPs and BC.Reflecting on the unique properties of UFPs, they might require more specific approaches which account for their spatial heterogeneity to be completely abated and controlled.Results from correlation analysis have showed major influences of different types of vehicles on the UFPs PNC level at roadside microenvironments.Hence, well-planned control policies for these particular emission sources could be beneficial to address the issue.Several stakeholders have supported the transition into non-combustion vehicle (EV cars) to pursue both net-zero goals and cleaner air quality (Li et al., 2016).However, UFPs might also be generated from non-exhaust process through brake and tyre wears (Piscitello et al., 2021;Abdillah and Wang, 2023).There are currently limited studies which comprehensively characterized the contribution of non-exhaust UFPs in the total UFPs PNC at roadside microenvironments.Hence, future roadside UFPs studies should also focus on this issue to provide substantial insight for stakeholders in developing suitable policies.Future UFPs study may also utilize AI-driven methods to assist in integrating real-time data features such as air quality or traffic vehicles image data that can be used to further analyze intra-urban spatial variability of UFPs or carbonaceous aerosol (Lin et al., 2023).

CONCLUSIONS
This study investigated the spatiotemporal variability of UFPs PNC and BC concentrations at different roadside microenvironments and urban background in the vicinity of urban industrial park area.Higher UFPs PNC and BC concentrations were identified at all sampling sites during cold period due to lower mixing and temperature inversion.In addition, peaks average concentrations of UFPs PNC and BC were identified at IN and RS during morning peaks and weekdays periods, following the diurnal profiles of traffic flux compositions and total vehicle numbers per day.Intraurban spatial variability was identified between roadsides UFPs and urban background UFPs, highlighting the heterogeneity characteristic of UFPs.Furthermore, the findings in this study suggested that the concentrations of UFPs PNC and BC at IN and RS were significantly correlated with traffic activities, while at UB were associated with meteorological conditions and secondary emission.Observed UFPs PNC concentrations at IN were heavily influenced by heavy duty vehicles and motorcycles, while RS was heavily influenced by gasoline cars.
Elevated levels of roadside UFPs and BC identified in this study have exceeded the typical "high-level" of UFPs concentration according to WHO AQG 2021, indicating an elevated exposure risk for pedestrians and residents.Moreover, intra-urban spatial variability characteristic of UFPs and BC identified from this study have further highlighted their difference compared to the currently regulated PMs.Hence, future exposure estimation models in UFPs and BC health effect study should also account for this characteristic.This study may serves as reference for future epidemiological study related to roadside UFPs in specific urban microenvironments such as industrial park areas.

Fig. 6 .
Fig. 6.IQR/M ratio of UFPs PNC and BC concentration in warm and cold period.

Fig. 7 .
Fig. 7. Pearson's r correlation heatmap between UFPs PNC and BC and potential confounding factors.Note: Motor = Number of motorcycles; Heavy = Number of heavy vehicles; Gasoline = Number of gasoline cars; TotalVec = Total vehicle numbers.

Table 1 .
Summary of average meteorological conditions observed during study.

Table 3 .
Comparison with previous roadside UFPs and BC measurement studies.