The relative contributions of traffic and non-traffic sources in ultrafine particle formations in Tehran mega city

Emissions of ultrafine particles (UFPs; diameter < 100 nm) are strongly associated with traffic-related emissions and are a growing global concern in urban environments. The aim of this study was to investigate the variations of particle number concentration (PNC) with a diameter > 10 nm at nine stations and understand the major sources of UFPs (primary vs. secondary) in Tehran megacity. The study was carried out in Tehran in 2020. NOx and PNC were reported from a total of nine urban site locations in Tehran and BC concentrations were examined at two monitoring stations. Data from all stations showed diurnal changes with peak morning and evening rush hours. The hourly PNC was correlated with NOx. PNCs in Tehran were higher compared to those of many cities reported in the literature. The highest concentrations were at District 19 station (traffic) and the lowest was at Punak station (residential) such that the average PNC varied from 8.4 × 103 to 5.7 × 104 cm−3. In Ray and Sharif stations, the average contributions of primary and secondary sources of PNC were 67 and 33%, respectively. Overall, we conclude that a decrease in primary emission leads to a decrease in the total concentration of aerosols, despite an increase in the formation of new particles by photo nucleation.


Study area and monitoring sites
With an area of 700 square kilometres, Tehran extends from latitude 35° 35′ N to 35° 48′ N and longitude 51° 17′ E to 51° 33′ E. It is located at more than 1200 m above sea level with a slope of 700 m between the highest and lowest points.Tehran has approximately 13.3 million residents and 10 million commuters 44 .It is located in the foothills of the Alborz Mountains in the north, Jajrood valleys in the east, Karaj valleys in the west, and the south western margin of the central desert from the south.Due to the prevailing meteorological conditions and topography of Tehran, stable meteorological conditions and temperature inversion occur more frequently in winter and autumn, which is one of the main reasons for severe air pollution 45 .However, man-made factors such as rapid population expansion, rapid conversion of agricultural land and natural objects into industrial sites and urban areas, and a relatively old vehicle fleet contribute significantly to the severity of air pollution in Tehran. www.nature.com/scientificreports/

Data collection
Hourly concentrations of NOx in 2020 were obtained from nine air quality monitoring stations (Table 1 and Fig. 1) operated by Tehran's Air Quality Control Company (AQCC) (http:// air.tehran.ir/).Hourly BC data at two air quality monitoring stations (SHU and RAY) were obtained from Tehran Air Quality Control Company.The Environnement S.A-AC32M analyzer was used to monitor NOx concentrations.Criteria air pollutants, including CO, SO 2 , NOx, PM 2.5 , and O 3 were measured in 1-h time resolution.The BC concentrations were measured in 1-min time resolution by using an AE33 BC monitor manufactured by Magee Scientific, USA.PN  Vol:.(1234567890)EN 14,626:2012 for PM and CO analyzers, respectively.QA/QC procedures for Aethalometers were performed following the official user manual of the analyzers published by Magee Scientific 41 .
The particle number concentrations were measured using a NanoTracer, Aerasense (Netherlands).This device is able to determine the average particle size and the particle number concentration in the size range of 10-300 nm up to 10 6 cm −3 .PNC sampling was performed at nine stations such that at each station, NanoTracer was operated for six consecutive days 24 h a day.The sampling intervals were every 10 s; then; the corresponded data converted to 1-h averages.The NanoTracer employed in the experimental studies underwent calibration conducted by the manufacturer.Since the NanoTracer is not a reference instrument, a correction factor has to be applied to its readings.The correction factor was obtained through collocated measurements of a NanoTracer and a reference instrument such as SMPS, FMPS or CPC.The ratio between the concentrations recorded by NanoTracer and a reference instrument is defined as correction factor.The PNC data in this study were adjusted using the correction factor reported in reference 46 .The average correction factor applied for NanoTracer was determined to be 1.9 ± 0.3.

The Segregation of the primary and secondary sources of PNC
Equations ( 1) and ( 2) outline the methodology employed in this study to estimate the contribution of primary and secondary particles to the total PNC 47 .During morning rush hours (primarily 6-9 a.m.), a linear regression correlation is established between PNC and BC, and the estimated PNC derived from this equation is denoted as N 1 .Subsequently, this estimating equation is applied throughout the remainder of the day, utilizing measured BC concentrations to estimate N 1 .N 2 is then calculated by subtracting N 1 from the measured total PNC (N) over the course of the day.In this approach 47 , N 1 represents primary traffic emissions, while N 2 encompasses various scenarios, including newly formed particles in the atmosphere from gas precursors, low BC-bearing primary particles from diverse urban sources excluding traffic, and particles transported by air masses 14,48,49 .S 1 (particles/ ng BC) denotes the slope of the correlation between N and BC during the morning rush hours.N represents the field-measured total number concentration, and BC represents the field-measured black carbon concentration.This methodology has been successfully applied in prior studies conducted in European cities 50,51 , as well as in an Asian megacity and boreal forest site in Finland 52 .
Reference 14 considered the first percentile of the N/BC ratio during the morning rush hour to develop a correlation between N and BC while we used all morning rush hour data since our dataset was not as large as that used by the reference 14 .

Source identification using conditional bivariate polar function (CBPF)
The CBPF method 53 was used to identify potential PNC emission sources.CBPF analysis can identify potential sources around stations and estimate the likelihood that high concentrations will occur there.The CPF method 54 , which incorporates wind speed (or any other parameter) as a third variable.Using the ordinary CPF, we can estimate how likely it is that a pollutant concentration measured in one wind sector will exceed a certain threshold.Unlike wind direction sectors alone, CBPF, defined as Eq. ( 3), takes into account different wind direction and speed ranges: As m �θ ,�u represents the number of samples taken in a given wind sector �θ at wind speeds u , C represents a pollutant concentration, x indicates a high percentile of concentration, such as 75th, and n �θ ,�u indicates the total number of samples taken during the wind direction-speed interval.In the R language (version 4.3.0 55), we performed these analyses utilizing the "OpenAir" package 56 .

Ethical Responsibilities of Authors
All authors have read, understood, and have complied as applicable with the statement on "Ethical responsibilities of Authors" as found in the Instructions for Authors.This study does not involve human subjects or animals.

Spatial and temporal variations of PNC and NOx concentrations
Figure 2 illustrates the overall mean PNC values at nine monitoring stations over the study period.Among these, the lowest PNC value (0.84 × 10 4 particle/cm 3 ) was registered for the residential station PUK, while urban-traffic stations DIS19, followed by SHU, recorded the highest values (5.9 and 4.8 × 10 4 particle/cm 3 , respectively).Utilizing inverse distance-weighted interpolation, the total PNC values demonstrated a spatial pattern with an increase from northern to southern city areas.This pattern suggests the influence of topography and mixed layer height, with rougher and higher conditions in the northern regions aiding in the dispersion and dilution of PNC.Conversely, the central and southern areas, characterized by a higher concentration of pollution sources, including direct emissions from road traffic and industrial zones, experienced elevated PNC values.
(1) Figure 3 shows the diurnal patterns of PNC (left side) and NOx (right side) across various stations throughout the sampling period.Stations PUK, GLB, and AQS consistently maintained PNC values close to the WHO high value threshold for UFP concentration (2 × 10 4 particle/cm 3 for a 1-h period) for most hours of the day 57 .However, other stations consistently exceeded this limit throughout the entire day.DIS19, DIS21, Ray, and SHU exhibited bimodal peaks, with PNC rising from 5 (local time) and peaking between 6 to 8, followed by a second peak in the afternoon from 17 to 20 (local time).Conversely, FSQ, TRBM, and AQS experienced unimodal peak values in the early hours, predominantly between 0 and 5 a.m.These dual increases in PNC are likely associated with morning and evening rush hours, as well as the influence of meteorological condition.In Tehran, traffic regulations impose restrictions on heavy-duty diesel vehicles (HDDVs) during daytime hours.Specifically, heavy-duty trucks are permitted within the city from late night to early morning on workdays (Saturday to Wednesday) and from midnight to early morning on weekends (Thursday and Friday).Other diesel vehicles, including light delivery trucks and public transportation, have the flexibility to operate in Tehran almost continuously 41 .The PNC experienced a marked increase, specifically at DIS19, DIS21, FSQ, SHU, and RAY, and sustained high levels when HDDVs were granted access to the streets, particularly when the MLH was low.As the boundary layer height increased, and heavy trucks were prohibited, the PNC concentration experienced a significant decrease, maintaining a lower level throughout the day until night-time.
Given their proximity to traffic sources, stations DIS21, RAY, SHU, and TRBM exhibited similar trends in both NOx and PNC, displaying simultaneous peaks and troughs.This consistency suggests insufficient time for pollutants to mix within the MLH, indicating that PNCs emitted directly from vehicles started to increase or decrease at these stations almost concurrently with NOx variations.The NOx concentration profile distinctly shows two peaks: one during night-time for HDDVs traffic and another during the daytime for the morning rush hour of light-duty vehicle (LDV) traffic.Notably, the profile underscores the pronounced impact of LDVs on NOx concentration, contrasting with the comparatively lower effect of HDDV traffic during night-time.NOx concentrations ranged from 25.4 to 293 ppb at urban-traffic stations and 17.2-105 ppb at urban-residential stations.Furthermore, NOx concentrations were generally lower in the afternoon than in the morning at most stations, indicating the dominant influence of traffic emissions in the morning.Stations located closer to the center of Tehran, such as SHU, TRBM, and DIS21, exhibited higher NOx concentrations due to increased traffic load and elevated levels of domestic and commercial activities.

Contribution of primary and secondary sources in PNC
Table 2 presents the average percentage of N 1 and N 2 using hourly concentration data for RAY and SHU stations.In urban environments with traffic emissions, an observed association between BC concentration and PNC has been reported 10,58,59 .Scattered plots of BC versus N were analyzed for traffic rush hours in the morning, resulting in estimated values of S 1 (expressed as particles/ng BC) at 14.4 × 10 6 and 15.9 × 10 6 for RAY and SHU stations, respectively.Throughout the day at RAY station, the contribution of N 1 to PNC was generally higher than that of N 2 , except during the noon-afternoon time when the proportions were 34.4% for N 1 and 65.6% for N 2 .This observation underscores the significant contribution of primary particles associated with traffic during those specific hours.At SHU station, the contribution of N 1 was lower from noon till evening, accounting for 42.5% compared to 57.5% for N 2 .N 2 reached peak levels during midday, constituting 65.6% and 51.5% at RAY and SHU, respectively.This peak coincides with the anticipated maximum of photochemical nucleation, attributed to the photo-oxidation of gaseous precursors in the atmosphere during periods of maximum solar radiation.The highest contribution of N 1 was observed at RAY during night-time and morning rush hours due to its location far from the city's core, being influenced by mixed traffic and industrial sources.Additionally, RAY station is situated near Tehran's ring road, where HDDVs are permitted to pass without time restrictions.Consequently, increased heavy diesel vehicle traffic during the night resulted in higher emissions of PNC compared to daytime 41 .Particles below 100 nm, which frequently dominate urban PNC, are directly emitted into the atmosphere from combustion processes associated with industry, traffic, domestic heating, and other sources such as vehicle brakes.Emissions from vehicles can contribute to the presence of both primary and secondary particles in the atmosphere.These pollution episodes may be mitigated through taking a wide variety of implementations, such that the implementation of traffic restrictions, particularly in central areas like The Odd-Even Traffic Rationing zone, The Restricted Traffic Zone, and The Low-Emission Zone, resulted in reduced emissions of CO, NOx, VOCs, and SOx by 4.5%, 2.9%, 5.8%, and 2.7%, respectively 41 .

Correlations between PNC and criteria air pollutants
The relationships between hourly PNC and concentrations of CO, SO 2 , NOx, PM 2.5 , and O 3 were examined through a single-variable regression method, as depicted in Fig. 4. The results revealed moderately low but significant correlations (p-value ≤ 0.05) between PNC and CO for FSQ (R 2 = 0.49) and between PNC and SO 2 for GLB (R 2 = 0.41).For all other stations, the correlations were low, with R 2 values ranging between 0.17 and 0.24 for CO and between 0.01 and 0.17 for SO 2 .Conversely, notable correlations, ranging from relative-high to moderate-low, were observed with NOx at FSQ and RAY (R 2 = 0.57 and 0.55, respectively), as well as AQS (R 2 = 0.39), DIS21 (R 2 = 0.20), GLB (R 2 = 0.24), and TRBM (R 2 = 0.28).It is noteworthy that similar patterns in the relationship between PNC and NOx were observed in studies conducted at surface stations in Gothenburg 26 , London 60 , and Stockholm 61 , reinforcing the consistency of our findings with existing research.This pattern can be attributed to the fact that a substantial proportion of urban NOx emissions is associated with diesel vehicles 62,63 .Despite comprising only 2.4% of Tehran's vehicle fleet, diesel vehicles contribute significantly, accounting for Table 2. Total average percentages of N1 and N2 on an hourly basis during the day.www.nature.com/scientificreports/more than 41%, 64%, and 85% of the NOx, SOx, and PM emissions, respectively 41 .Moderate-low to moderatehigh correlations between PNC and NO and NO 2 were similarly reported for both urban (R 2 = 0.27 and 0.35, respectively) and traffic-oriented stations (R 2 = 0.70 and 0.63 for NO and NO 2 , respectively) across European countries 64 .The mean PNC values and R 2 values for PNC-NOx correlations from other studies are detailed in Table 3, demonstrating consistently elevated levels in metropolitan areas and proximity to highways.Notably, strong correlations between particle number concentrations and NOx were observed in most studies.Marylebone Road recorded the highest PNC among the locations listed in Table 3, situated alongside a road with a traffic flow exceeding 80,000 vehicles per day within a street valley 65 .While Tehran's PNC exceeded values in all the cities mentioned in Table 3, it only fell below Hornsgatan and Marylebone Rd.Discrepancies in PNC among cities may arise not only from differing source profiles but also from variations in the instrumentation used for measurement, especially considering the potential impact of lower cut sizes on total measured PNC.For PNC and PM 2.5 , correlations were either insignificant (e.g., AQS, PUK, and TRBM) or significantly moderate-low at DIS21 and RAY (R 2 = 0.20 and 0.28, respectively).Global study 66 , including ten cities across North America, Europe, Asia, and Australia, also reported low correlations of PM 2.5 and PNC (R 2 = 0.01 to 0.48).Thus, PNC and PM 2.5 measurements do not represent each other adequately, highlighting the need for more precise pollutant indicators such as PNC or BC mass 41 instead of total PM 2.5 mass for more effective policy implementation.Finally, negative and very low correlations for O 3 with PNC were observed at FSQ, GLB, SHU, and TRBM (R 2 < 0.1).In contrast, notably higher PNC and O 3 correlations were identified at DIS21, DIS19, and RAY (R 2 = 0.24, 0.27, and 0.46, respectively).The negative slopes for the former group indicate that O 3 -rich sources of PNC emissions are not significant in Tehran, at least during the measurement period in this study.

CBPF analysis results
In Fig. 5, we showed the dominant directions of wind and the CBPF analysis for PNC located in the north (AQS, residential), center (SHU, and DIS19, traffic), and south (RAY, traffic-industrial), where only the 75 th percentile was used to distinguish the most important sources of pollution at each station.The traffic sources and domestic heating emissions at AQS, GLB, and DIS19 stations were predominantly local in nature since low wind speeds prevail.These stations showed 45%, 25%, and 30% probabilities coinciding with 2.4 × 10 4 , 2.3 × 10 4 , and 7.7 × 10 4 particle/cm 3 (75th percentiles) concentrations when the wind is from ESE, S, and WNW directions and the wind speed is in the range of 5-10 m/s, respectively.The above-mentioned directions contain a number of major roads that could have an impact on traffic-related sources.For instance, Sadr and Sayyad Shirazi highways are located 1.5 km away from AQS, Baqeri Expressways are located 0.5 km away from GLB, Sa'idi and Kazemi Expressways are located just 0.5 km away from DIS19.Several sources of pollution have been identified in the NW of DIS21, which could be attributed to industrial complexes (food and automotive industries), as well as the Lashkari Expressway.Despite the fact that the Azadegan Expressway is located in the SE of this station, there is more than a 30% probability that wind speed ranges between 20 and 30 m/s and winds from the NW could significantly increase particle concentrations to 5.3 × 10 4 particle/cm 3 , highlighting the importance of meteorological impacts on pollutant long-range transport and dispersion.Similarly, in spite of SHU station's proximity to the Mehrabad airport, however; particle concentrations are more than 80% probable to be originated by NW wind directions and 20-30 m/s wind speed to be 5.8 × 10 4 particle/cm 3 , which are the locations of the Tarasht power plant and the Nuri Expressway.FSQ station, located just close to the airport and residential environment, is primarily affected by domestic sources such as natural gas, liquid petroleum, and propane gas which are used for building spaces heating or emitted from kitchens during cooking, as well as a high rate of emission from the airport.Airports are responsible for emitted pollutants such as PN, PM 2.5 , and black carbon; PN concentrations at airport sites were approximately four times greater compared to the freeway 70 .The W wind component and 30-40 m/s wind speed highlight the impacts of the Fath highway aside from the sources mentioned above, which can significantly increase particle concentrations probabilities by more than 25% and 15% to be 3.2 × 10 4 particle/cm 3 , respectively.Similar to FSQ, PUK and TRBM stations also suffer from domestic sources from SW and SE with wind speeds between 15 and 25 m/s, respectively, which contribute to only 9.2 × 10 3 particle/cm 3 by 60-80% and 3.7 × 10 4 particle/cm 3 by more than 80% probabilities, respectively.Because the strongest wind direction comes from the W-NW directions at both stations, and the 75 th percentile at PUK station is considerably lower than other stations, the role of the major road, Ashrafi Esfahani Expressway, located in SE direction of PUK station, cannot be discussed precisely in terms of long-range transport contribution to pollution.Along with the vicinity of the Avini Expressway, RAY station is also affected by the Be'sat power plant.A particle concentration of 4.1 × 10 4 particle/cm 3 is more than 50% likely to result from N wind directions with a speed of 15-20 m/s at this urban traffic station.As a result of distillate oil and natural gas usage in Iran during the cold and warm seasons, they emit high levels of NOx, SO 2 , PM, and greenhouse gases from stationary internal and external combustion 71 , which emphasize the significance of power plants studied above.

Limitations of this study
It is important to highlight that in stations characterized by elevated BC concentrations, the calculated values of N 2 appeared to be negative.This suggests that the application of the method proposed by 47 may not be universally applicable under conditions with high BC concentrations.However, to draw more definitive conclusions on this matter, further investigation through a comprehensive study with a larger dataset is warranted.Moreover, a combination of particle size and number may shed light on the primary or secondary production of UFPs; thus, considering that particles with primary mode diameter peaks at 30-35 nm and 60-80 nm are linked to sparkignition and diesel vehicle emissions, respectively 72 , conducting particle number size distribution analyse would be the future work of the present study.This additional step can enhance the ability to discern local or regional traffic sources, providing a more nuanced understanding of the contributors to particle number concentration.

Conclusions
In Tehran, the primary source of PNC is predominantly linked to vehicle exhaust emissions, particularly heightened during rush hours.Secondary particle formation in the ambient air is observed mainly during noon or early afternoon.The diurnal PNC trend follows a pattern with peak values occurring during morning and evening rush hours.This study establishes a positive correlation between changes in urban PNC and BC as well as NOx.
To differentiate between primary and secondary sources of PNC, the segregated method is applicable when BC and PNC are measured simultaneously at the same stations.In Tehran, specifically at RAY and SHU stations, the average contribution of primary and secondary sources to PNC was determined to be 67% and 33%, respectively.The CBPF analysis identified local traffic as the primary source of PNC emissions in Tehran.Additionally, the study underscored the influence of meteorological factors that may contribute to the transport of pollution over long distances from distant sources to the receptor.This highlights the significance of the MLH as a determining factor during the daytime in the cold season in Tehran.Notably, the study revealed that traffic regulations for HDDVs played a significant role in influencing PNC levels at traffic stations during the night-time.It was found that PNC, as a local pollutant, is directly impacted by the emissions from the diesel fleet, particularly heavy-duty trucks, indicating that HDDVs traffic stands as the main source of PNC emissions in Tehran.Consequently, the study suggests that phasing out old HDDVs and replacing them with newer technology vehicles could yield beneficial outcomes.The average PNC values observed in most stations in Tehran exceeded those in many cities reviewed in the study.In conclusion, the study recommends that reducing primary emissions in Tehran would be a practical approach to decrease the population's exposure to UFPs.Additionally, controlling the formation of new particles could also significantly contribute to reducing such exposure.

Figure 1 .
Figure 1.Locations of nine monitoring stations, main point sources which may contribute to PNC, and the major traffic roads around the stations, including Mehrabad airport (MIA), Tarasht power plant (TPP), B`esat thermal power plant (BTPP), South Bus Terminal (SBT), Ray and Tehran cement factories (RCF, and TCF).

Figure 2 .
Figure 2. Total values of PNC (× 10 4 particle/cm 3 ) at all stations during the measurement periods with the corresponding IDW interpolation values.

Figure 3 .
Figure 3.Diurnal average concentrations of PNC (blue, left side), and NOx (red, right side) at all stations.The brown dashed line represents the WHO high value threshold for UFP concentration for 1-h average.

Figure 4 .
Figure 4. Correlations between hourly average PNC and criteria air pollutants with 0.95 confidence interval at nine stations.

Figure 5 .
Figure 5. Dominant wind directions with CBPF polar plots analysis showing how contributions of different distant and local sources are affected by wind direction and wind speed.

Table 1 .
Main details of the selected monitoring stations.

Table 3 .
Particle Number Concentrations and R 2 values for PNC-NO x reported in other studies.