Research on Prediction Model of Particulate Matter in Dalian Street Canyon

Abstract

In urban areas where populations commonly reside, particle mass concentrations in street canyons can pose significant risks to human health. This study aimed to investigate the diffusion mechanism of particle mass concentrations in urban street canyons by developing and applying a prediction model based on the mathematical modeling of physical processes. The prediction model considered factors such as the influence of traffic wind, natural wind, traffic flow, and other relevant variables influencing particle mass concentrations in street canyons. Field measurements were conducted in Dalian, China, to verify the feasibility of the model. Particle mass concentrations, traffic flow, temperature, relative humidity, and wind speed were measured on Shichang Street (a two-lane one-way road), Tangshan Street (a four-lane two-way road), and Shengli Road (a six-lane two-way road). The results indicated that the majority of traffic peaks occurred around 19:00 on all road types. The PM_1.0 mass concentration was well diluted on the four-lane two-way road, with the least dilution observed on the two-lane one-way road. A strong correlation between the particle mass concentrations and traffic flow was discovered. Furthermore, a prediction model was established, accurately predicting the particle mass concentrations when the prediction step was from 5 to 15 s. The coefficient of determination (R²) between the predicted and measured values on the two-lane one-way road, four-lane two-way road, and six-lane two-way road was 0.9319, 0.6582, and 0.9238, respectively. Additionally, the prediction model allowed for a detailed analysis of traffic flow limitations, corresponding to the recommended World Health Organization (WHO) PM_2.5 values. Overall, the findings of this study offer valuable insights for forecasting particle exposure levels in street canyons.

1. Introduction

In recent years, the proliferation of vehicles has led to a significant rise in pollution stemming from vehicle exhausts, with a particular focus on urban street canyons where traffic emissions constitute the primary source of particle mass concentrations. As the COVID-19 pandemic recedes, increased travel activity is being observed, resulting in a notable uptick in pollutants generated by road traffic [1,2,3,4,5]. Numerous studies have underscored the direct health implications of particulate matter (PM) on human health [6,7,8,9,10,11,12,13,14,15,16,17,18,19,20]. Among the many particles that appear in urban street canyons, PM_1.0, PM_2.5, and PM₁₀ are deemed the most harmful. PM_1.0, containing a plethora of toxic substances, constitutes a significant proportion of PM_2.5 [21,22]. PM₁₀ and PM_2.5, capable of direct lung penetration, are not easily excreted by the human body [23]. Particularly concerning is PM_1.0′s ability to not only penetrate deeply into alveoli and remain there, but also to enter the human blood circulation. Owing to its expansive specific surface area, it has a strong ability to attach harmful substances per unit area. Such particles are closely associated with human morbidity and mortality [24,25].

Several studies [26,27,28] have conducted real-time monitoring of particulate matter in street canyons and compared it with the particle mass concentrations in the city center, ultimately obtaining particle mass concentrations at various city locations. However, it is not practical for most individuals to monitor particle mass concentrations using instruments. Therefore, this study aimed to predict particle mass concentrations through model establishment. Accurate air quality forecasting technology is crucial for early pollution warning and control. To date, methods for predicting particle mass concentrations primarily fall into deterministic and statistical categories [29]. Deterministic models utilize the mathematical modeling of physical and chemical processes to forecast particle mass concentrations. Conversely, statistical approaches typically employ multiple linear regression (MAR), neural network (NN), fuzzy logic (FL), and autoregressive moving average (ARIMA) models [30,31,32,33,34,35].

While statistical approaches primarily rely on air quality and meteorological data, deterministic methods encompass not only meteorological data, but also other factors that impact particle mass concentrations. Remarkable progress has been made in the use of deterministic models to predict particle mass concentrations in recent years [36,37,38]. These models are grounded in physical and chemical principles integrating various atmospheric processes such as emissions from natural and human activities, the transport of pollutants through the atmosphere, turbulent dispersion, and chemical reactions. Deterministic models for physical and chemical processes have been established and verified [39,40], demonstrating a high correlation between calculated and measured values. Both deterministic and statistical model-based forecasts have their respective advantages and limitations. Compared with deterministic models, statistical models require a minimal amount of input data and computing power, offering a relatively straightforward and flexible user experience. The primary drawback is that statistical models fail to provide direct insights into the underlying physics of the data. However, deterministic models have a higher prediction accuracy [41,42]. Particulate matter is a multisource pollutant involved in physical and chemical processes [43], and deterministic methods can more accurately predict particle mass concentrations by building physical and chemical models. From a computational standpoint, deterministic models require substantial amounts of input data and boundary conditions, rendering them highly demanding.

This study proposed a higher-precision spatiotemporal prediction model for particle mass concentrations in street canyons based on measured data. Different types of traffic flow data were quantified according to PM exposure standards, providing a quantifiable basis for urban planning and road network air pollution assessments. Field measurements were conducted on three urban street canyons (Shichang Street, Tangshan Street, and Shengli Road) in Dalian. The measured data were used to develop and validate the prediction model for particle matter. The findings of this study offer valuable guidance for minimizing individual exposure to particle mass concentrations and optimizing travel plans.

2. Methodology

2.1. Prediction Model of Particle Mass Concentrations

To further explore the diffusion law of particle mass concentrations in street canyons, this study adopted the theory of a multilayer canopy [44,45] for analysis. According to the multiple canopy theory, canopy layers along the vertical direction include the urban, street canyon, and vehicle canopy layer [46]. Therefore, in this study, the vertical height of the urban canyon layer was divided into two distinct layers: vehicle canopy and street canyon canopy. A schematic representation of the street canyon setup is depicted in Figure 1. There was vertical diffusion of particle mass concentrations between them, and the control equation for particle mass concentrations between the vehicle canopy and street canyon canopy was derived using mass conservation. Owing to the volume exchange between the vehicle canopy and the upper layer of the street canyon during vehicle movement, the particle mass concentrations disperse with the air of traffic and natural winds. Moreover, this study emphasized the effect of vehicle-to-vehicle exhaust emissions on pollution dispersion and did not consider the effects of other external pollutants. Therefore, the particle mass concentrations generated by vehicles in street canyons were regarded as the only source of pollution in this investigation.

Equation (1) is defined as follows:

$$\stackrel{.}{q}{\lambda }_e=u_{d1}(C_1-C_2)(1-{\lambda }_{p1})$$

(1)

where $\dot{q}$ indicates the emission flux of traffic-related pollutants (in kg/s/m²), ${\lambda }_e$ is the ratio of pollutant area to total area, $C_1$ and $C_2$ represent pollutant concentrations in the vehicle and canyon canopies (in μg/m³), ${\lambda }_{p1}$ represents the site coverage of the vehicle canopy, the site coverage is the ratio of the vehicle cross-section to the total area in the street canyon, and $u_{d1}$ shows the diffusion velocity between the vehicle and street canyon canopies (in m/s).

The diffusion velocity is obtained by the following:

$$u_d=\frac{u^{\ast }}{\sqrt{2}\pi }(2)u_d=\frac{u^{\ast }}{\sqrt{2}\pi }$$

(2)

where $u^{\ast }$ indicates the friction velocity of the shear layer, and when ${\lambda }_f\ge 0.4$, the ratio of $u^{\ast }/U_{ref}$ is 0.12; $U_{ref}$ represents the average wind speed at the top of the rough sublayer, and ${\lambda }_f$ shows the front density (density of the windward area).

The front density is obtained as follows:

$${\lambda }_f=\sum _{\theta =1}^{16}{\lambda }_{f(\theta )}P(\theta )$$

(3)

where ${\lambda }_{f\left(\theta \right)}$ and $P\left(\theta \right)$ represent the front density and the annual wind probability at $\theta$ angles, which can be obtained from wind data climate statistics.

2.2. Model Deduction

This study made the following assumptions regarding the process of a vehicle driving in a street canyon:

• Vehicle exhaust was the sole source of pollution in the street canyon.
• Wind speed was uniform in the urban canopy.
• Different canopy heights were distinguished by site coverage ${\lambda }_p$ and front density ${\lambda }_f$.
• The particle mass concentrations above buildings’ roofs were ignored.
• The particle mass concentrations delivered horizontally to the target volume were equal to the pollutant concentrations exported by the target volume.

Based on the above five assumptions, a prediction model for the particle mass concentrations in street canyons was established.

In this model, the change in the particle mass concentrations in the street canyon is calculated as follows:

$$V\frac{d{C}_d}{dt}=n{C}_p{V}_p-V_j(C_d-C_{j,0})-V_f(C_d-C_{f,0})+V_u(C_u-C_d)$$

(4)

where $V$ indicates the volume of the target pedestrian area (in m³), $C_d$ denotes the particle mass concentrations in the target pedestrian area (in μg/m³), $n$ shows the number of vehicles in the street canyon (in #), $C_p$ represents the particle mass concentrations in the vehicle exhaust (in μg/m³), $V_P$ indicates the volume of vehicle exhaust gas (in m³/s), $V_j$ shows the traffic wind flow (in m³/s), $C_{j,0}$ represents the concentration of external pollutants from traffic wind (in μg/m³), $V_f$ denotes natural ventilation (in m³/s), $C_{f,0}$ shows external particle mass concentrations from natural wind (in μg/m³), $V_u$ represents the amount of air exchanged between the lower and upper layers of the street canyon (in m³/s), and $C_u$ shows the particle mass concentrations in the middle and upper layers of the street canyon (in μg/m³).

The number of vehicles in the street canyon are calculated as follows:

$$n=\frac{QL}{60v_c}$$

(5)

where $Q$ is the number of vehicles recorded per min (in #/min), $L$ is the length of the street canyon (in m), and $v_c$ is the vehicle speed (in m/s).

The traffic wind flow is established as follows:

$$V_j=v_c\sqrt{\frac{n{A}_m}{A_r}}{A}_r$$

(6)

where $A_r$ is the longitudinal cross-sectional area of the street canyon (in m²) and $A_m$ is the longitudinal cross-sectional area of the vehicle (in m²).

The natural ventilation is determined as follows:

$$V_f=v_0\mathit{cos}\theta A_r$$

(7)

where $v_0$ indicates the wind speed (in m/s) and $\theta$ shows the wind angle.

The amount of air exchanged between the lower and upper layers of the street canyon is obtained as follows:

$$V_u=u_d{A}_s$$

(8)

where $A_s$ is the cross-sectional area of the street canyon (in m²).

By inserting Equation (2) through Equation (6) into Equation (1), Equation (9) is obtained as follows:

$$L{A}_r\frac{d{C}_d}{dt}=\frac{QL}{60v_c}{C}_P{v}_{p}s-v_c\sqrt{n{A}_r{A}_m}(C_d-C_{j,0})-v_0\mathit{cos}\theta A_r(C_d-C_{f,0})+A_s\frac{C_p{v}_p{\lambda }_e}{{\lambda }_p-1}$$

(9)

Then, Equation (10) is obtained through simplification as follows:

$$\frac{d{C}_d}{dt}+C_d(\frac{v_c\sqrt{n{A}_r{A}_m}}{L{A}_r}+\frac{v_0\mathit{cos}\theta }{L})=\frac{Q{C}_p{v}_{p}s}{60v_c{A}_r}+\frac{v_c{C}_{j,0}\sqrt{n{A}_r{A}_m}}{L{A}_r}+\frac{v_0\mathit{cos}\theta C_{f,0}}{L}+\frac{C_p{v}_p{\lambda }_e{A}_s}{L{A}_r({\lambda }_p-1)}$$

(10)

where $v_p$ indicates the exhaust emission speed of motor vehicle (in m/s), $s$ shows the exhaust port area (in m²), ${\lambda }_e$ denotes the ratio of the pollutant emission area to the total land area, and ${\lambda }_p$ indicates the site coverage.

Equation (11) is obtained through simplification as follows:

$$C_d=(D+M\Delta t{e}^{N\Delta t})e^{-N\Delta t}$$

(11)

where $D$ is a constant.

Equations (12) and (13) are defined as follows:

$$M=\frac{Q{C}_p{v}_{p}s}{60v_c{A}_r}+\frac{v_c{C}_{j,0}\sqrt{n{A}_r{A}_m}}{L{A}_r}+\frac{v_0\mathit{cos}\theta C_{f,0}}{L}+\frac{C_p{v}_p{\lambda }_e{A}_s}{L{A}_r({\lambda }_p-1)}$$

(12)

$$N=\frac{v_c\sqrt{n{A}_r{A}_m}}{L{A}_r}+\frac{v_0\mathit{cos}\theta }{L}$$

(13)

Equation (14) is obtained after modification as follows:

$$C_d=\alpha (D+M\Delta t{e}^{N\Delta t})e^{-N\Delta t}+\beta$$

(14)

where $\Delta t$ is the time step. Equation (14) describes the influence of traffic wind, natural wind, vehicle driving, and other factors on the particle mass concentrations in street canyons. Using the above equation, it can be demonstrated that the particle mass concentrations in the street canyon are positively correlated with the traffic flow and wind, and negatively correlated with the natural wind.

2.3. Field Measurement

To verify the above model, field measurements were conducted on three streets in Dalian, primarily during autumn and winter in 2020, and spring and summer in 2021. Seven randomly selected days were chosen from each season for measurement purposes.

The selected field measurement sites in Dalian include Shichang Street, Tangshan Street, and Shengli Road, as illustrated in Figure 2 and Figure 3. Shichang Road comprises a two-lane one-way road with a measured length of 0.33 km. Tangshan Road features a four-lane two-way road and spans a measured length of 0.32 km. Shengli Road is characterized by six lanes of two-way traffic with a measured length of 0.45 km. In this study, particle mass concentrations (PM_1.0, PM_2.5, and PM₁₀), temperature, relative humidity (RH), wind, and traffic flow were measured.

To ensure the precise capturing of changing patterns in the factors, particle mass concentrations were collected using a Korno GT-1000 composite pollutant mass concentration detector with a sampling interval of 1 s. The temperature and RH were recorded using a 179 dt DTH temperature and humidity recorder with a sampling interval of 1 min. Wind speed was recorded using a Testo 410i instrument with a sampling interval of 1 min. The measurement ranges, resolutions, and accuracies of each instrument are listed in

Table 1. Measuring instrument specifications and accuracy.

Instrument	Range	Resolution	Accuracy
179 dt DTH temperature and humidity recorder	0–80 °C	0.01 °C	±0.4 °C
	0–100%	0.01%	±3%
Korno GT-1000 composite pollutant mass	0–500 µg/m3	1 µg/m3	≤±3%
concentration detector
Testo 410i	0.4~30 m/s	0.1 m/s	0.1 m/s

. A system diagram of the measurement instruments is shown in Figure 4.

Figure 5 depicts the measured points, with 12 traffic flow measuring points distributed across the three types of roads. To facilitate the analysis, 2-R represents a two-lane one-way road, 4-R signifies a four-lane two-way road, and 6-R denotes a six-lane two-way road. The characteristics of the traffic flow in the morning and afternoon were consistent [40,47]. Considering the convenience of processing the measured data, the measurement period was selected from 14:00 to 20:00. and was divided into off-peak and peak periods. The off-peak period was from 14:00 to 17:00 and the peak period was from 17:00 to 20:00.

2.4. Data Processing

In this study, the 3$\mathsf{\sigma }$-rule was employed to process data, defining the correct interval as ($\mathsf{\mu }-3\mathsf{\sigma }$,$\mathsf{\mu }$ + 3$\mathsf{\sigma }$). To facilitate data analysis, the particle mass concentration, temperature, and RH data were averaged over 10 min. The counting method for traffic flow was the sum over the 10 min.

A t-test was utilized to test the hypotheses, primarily employing the theory of the T distribution to ascertain whether the difference between two samples was statistically significant. The test results indicated a significant difference between particulate matter and the other factors [48,49,50,51].

3. Results and Analysis

The measured particles included PM_1.0, PM_2.5, and PM₁₀, with corresponding measured heights of 0.3 m (vehicle exhaust’s height), 1.2 m (vehicle canopy’s height), and 1.5 m (human respiration’s height). The sorted data showed that the particle mass concentrations were highest at 1.2 m for the three road types. Through a comparative analysis of the particle mass concentrations at the three road types, PM_1.0, PM_2.5, and PM₁₀ exhibited almost the same change trends. Therefore, this study focused on PM_1.0 measured at 1.2 m as the main research object.

3.1. Traffic Flow and Meteorological Parameters

3.1.1. Traffic Flow

The traffic flow during the measurement period for the three road types, from 14:00 to 20:00, is displayed in Figure 6. There were large differences in the traffic flow on the three road types. The average traffic flow on 2-R was approximately 125 vehicles/10 min, while for 4-R, this value was approximately 211 vehicles/10 min. Remarkably, 6-R exhibited a notably higher value of approximately 525 vehicles/10 min. The data for the three road types showed that the traffic flow in winter was relatively high. This finding may indicate a preference among Dalian citizens for self-driving during colder weather, as opposed to utilizing alternative transportation methods.

3.1.2. Temperature

As shown in Figure 7, the temperatures at 2-R remained relatively stable, with an average temperature change of 4.6 °C. In contrast, the value at 4-R was 12.9 °C, and 13.1 °C at 6-R. The temperature exhibited minimal fluctuation at 2-R and decreased gradually over time. The temperature at 4-R fluctuated in spring and summer, whereas that at 6-R fluctuated in spring and autumn. The fluctuations at 6-R were larger than those at 4-R. This may be attributed to the greater influence of traffic winds at 6-R compared to that at 2-R and 4-R.

3.1.3. Relative Humidity

The RH data are shown in Figure 8. The RH at 2-R ranged from 34.59% to 76.31%. Conversely, the RH change range was relatively high at 4-R, ranging from 11.35% to 85.16%. At 6-R, the RH maximum was the highest, ranging between 25.24% and 92.22%. The relative humidity increased with time in all four seasons, with the variation trend mostly contrasting with that of temperature.

3.1.4. Wind

The hourly wind speed is shown in Figure 9. The wind speed at 2-R ranged from 1.5 m/s to 3.8 m/s, from 1 m/s to 4.3 m/s at 4-R, and from 1.1 m/s to 3.9 m/s at 6-R. The maximum wind speed at 2-R appeared at approximately 14:00, and the minimum wind speed appeared at approximately 20:00 in spring, summer, and winter. The wind speeds at 4-R and 6-R exhibited various degrees of fluctuation, likely influenced by the size of the street canyons. Characteristically, 2-R is a deep street canyon (H/W > 1), whereas 4-R and 6-R are broad street canyons (H/W < 1). The deep street canyon acted as a barrier to the external wind, while the broad street canyon had little effect on the obstruction of external wind.

3.1.5. Particulate Matter

The trends in the particle mass concentrations over time are shown in Figure 10. The particle mass concentrations exhibited a single-peak distribution, as shown in Figure 10a,c. As time increased, the particle mass concentrations generally showed an upward trend and reached their peak between 18:00 and 19:00, likely attributable to the onset of evening rush hour. The inversion layer began to form and rose rapidly. After the evening rush hour, it began to decrease slowly again. However, the traffic flow peaks for the three road types occurred at 17:00, which is related to the cumulative effect of particle mass concentrations in the street canyon. The particle mass concentrations did not reach their maximum until 1 to 2 h after the traffic flow peaked and then started to decline.

Figure 10c shows the double-peak distribution of the particle mass concentrations in spring and summer. This can be attributed to sudden cloudy weather, as the particle mass concentrations were carried by the natural wind, resulting in the particle mass concentrations reaching a peak. Another peak was observed as the evening peak approached. Among the three road types, the particle mass concentrations were the highest at 2-R and lowest at 4-R. This may be because the red-light time on 2-R was 50 s, whereas the red-light time was 30 s on 4-R and 35 s on 6-R. The street canyon H/W of 2-R was higher than that of 6-R and 4-R. Therefore, the particle mass concentrations at 2-R were difficult to dilute. Among the four seasons, the particle mass concentrations in winter were higher than those in the other seasons, which corresponded to higher traffic flow. In summer, the RH was relatively high, resulting in a measurement instrument that identified small droplets of water as particles. Therefore, the particle mass concentrations in summer were higher than those in spring and autumn.

3.2. Analyzing Correlations between PM_1.0 and Other Parameters

To analyze the relationship between PM_1.0 and other parameters, the correlations between them are depicted in Figure 11, Figure 12 and Figure 13.

3.2.1. Correlations between Particle Mass Concentrations and Traffic Flow

As shown in Figure 11, the particle mass concentrations exhibited a positive correlation with traffic flow. The correlation between the particle mass concentrations and traffic flow was significantly greater on 2-R than on 4-R or 6-R. This disparity may be attributed to the positioning of the measurement points; all four measurement points were located at an intersection on 4-R, while three measurement points were located at an intersection on 6-R. Vehicles spent a longer duration at intersections compared to those travelling along roads. Therefore, the change in the particle mass concentrations was not only related to the traffic flow but also to the residence time of the vehicles.

3.2.2. Correlations of Particulate Matter with Temperature and Relative Humidity

As shown in Figure 12, the particle mass concentrations were correlated with the temperature and relative humidity. In most instances, the variation in the particle mass concentrations was positively correlated with temperature and negatively correlated with relative humidity. Increasing temperatures enhanced the diffusion capacity of particulate matter in street canyons, facilitating its dispersion in the air and consequently leading to a reduction in the particle mass concentrations. Conversely, high relative humidity levels resulted in the identification of small aerosol droplets as particulate matter by the particle mass concentration sensor, thereby increasing the recorded particle mass concentrations [40,52]. Among the four seasons, the particle mass concentrations exhibited the highest correlation with temperature and relative humidity in winter, which was clearly a seasonal change in the particle mass concentrations.

3.2.3. Correlation between Particle Mass Concentrations and Wind

As shown in Figure 13, there was a negative correlation between the particle mass concentrations and wind speed. The R² values between the particle mass concentrations and wind speed were 0.1425, 0.2621, and 0.1481 for 2-R, 4-R, and 6-R, respectively, indicating that the wind speed had a certain dilution effect on the particle mass concentrations. The particle mass concentrations in street canyons decrease with an increase in wind speed [53,54].

3.3. Model Calculation and Measured Result

The prediction time step was determined to verify the accuracy of the prediction model and the predicted value of the model was linearly analyzed with the measured value.

The detailed values of the prediction model are shown in

Table 2. Detailed values of particulate matter prediction model.

Model Parameter	Value
Pollutant discharge flux $\dot{q}$	4.5 × 10−6 kg/s/m2
Traffic flow $Q$	10–60 #/min
Wind $v_0$	1–6 m/s
Wind direction $\theta$	0–360°
Speed $v_c$	40–50 km/h
Exhaust emission velocity $v_p$	0.8 m/s
Exhaust port area $\mathrm{s}$	0.02 m2
Long and wide (double rows)	330 and 5 m
Long and wide (four rows)	320 m and 13 m
Long and wide (six rows)	450 m and 21 m
Street canyon canopy height	3 m
Long, wide, and high (car)	4.6 m, 1.8 m, and 1.5 m
External concentration of traffic wind	Measured acquisition
Natural wind external concentration	Obtained by China Weather Network
$u_d$	0.05 m/s
${\lambda }_p$	0.43
${\lambda }_e$	0.0015
${\lambda }_f$	Climate statistics acquisition

.

Due to the numerous factors affecting the diffusion of particle mass concentrations in street canyons, any minor disturbance may make it difficult to predict the diffusion of particle mass concentrations. During the verification process, the non-linear fitting of data with different time intervals was analyzed, and the time interval was determined to be between 5 and 15 s. The D and ∆t of the prediction models for different roads are shown in

Table 3. Different road-type prediction models.

Road Type	Model
2-R	$C_d=\alpha (-5.39+14M{e}^{14N})e^{-14N}+\beta$
4-R	$C_d=\alpha (-3.28+6M{e}^{6N})e^{-6N}+\beta$
6-R	$C_d=\alpha (-4.86+5M{e}^{5N})e^{-5N}+\beta$

. The model presented herein can be effectively applied to predict particle mass concentrations and provide theoretical support for the planning of street canyon structures.

Figure 14 shows the correlation between the values of the prediction model and the measured values of the particle mass concentrations across the four seasons. The R² values for the predicted and measured values were 0.9319, 0.6536, and 0.9218, respectively. During the measurement period, as the measured points were situated at intersections on 4-R, the particle mass concentrations were well diluted by natural and traffic winds, which were not accounted for in the predictive model. Therefore, the correlation between the predicted particle mass concentrations at 4-R and the measured values was relatively low.

3.4. Healthy Environment Analysis for Street Canyon

Current research provides a comprehensive definition of the limit for PM_2.5 mass concentration. As shown in Figure 15, there was a strong correlation between PM_1.0, PM_2.5, and PM₁₀ [55,56,57]. The particle mass concentrations of different particle sizes were highly correlated, suggesting that the particles possibly originated from the same source. Therefore, an analysis of a healthy environment containing PM_1.0 can be achieved by transforming it into an evaluation of PM_2.5.

Through model calculations, the traffic flow limits of the three road types corresponding to different PM_2.5 concentrations are shown in

Table 4. Traffic flow limits and non-accidental mortality corresponding to different PM_2.5 concentrations.

Grade	PM2.5	PM1.0	2-R	4-R	6-R	Non-Accidental Mortality	Level
Recommended	15.0 μg/m3	10.2 μg/m3	63#	127#	281#	0	Limit 1
Interim target 4	25.0 μg/m3	15.8 μg/m3	112#	310#	605#	0.7%	Limit 2
Interim target 3	37.5 μg/m3	22.7 μg/m3	293#	558#	879#	1.3%	Limit 3
Interim target 2	50.0 μg/m3	29.7 μg/m3	417#	1150#	2200#	2.3%	Limit 4
Interim target 1	75.0 μg/m3	43.6 μg/m3	870#	2650#	3030#	3.9%	Limit 5

. The World Health Organization (WHO) recommends that the 24-h daily PM_2.5 mass concentration should not exceed 15 μg/m³ [58]. Additionally, the WHO proposed four interim targets for environmental issues: Interim Target 4 (25 μg/m³), Interim Target 3 (37.5 μg/m³), Interim Target 2 (50 μg/m³), and Interim Target 1 (75 μg/m³) in 2021. Exposure to PM_2.5 concentrations of 25 μg/m³, 37.5 μg/m³, 50 μg/m³, and 75 μg/m³ corresponded to increases in non-accidental mortality of 0.7%, 1.3%, 2.3 %, and 3.9%, respectively. To determine the traffic flow limits, this study considered the four interim targets proposed by the WHO. The changes in traffic flow over time and the corresponding traffic flow limits of the particle mass concentrations are shown in Figure 16. The traffic flows corresponding to Interim Targets 1 and 2 were notably higher than the measured values; hence, Interim Target 1 and Interim Target 2 were not analyzed. The recommended value, the Interim Target 3 value, and the Interim Target 4 value for the traffic flow limits proposed by the WHO correspond to limits 1, 2, and 3, respectively. According to the correlation between PM_2.5 and PM_1.0, the PM_1.0 limits corresponding to different PM_2.5 limits are calculated, as shown in Table 4.

In Figure 16a, the average traffic flow on 2-R was higher than limit 1 but lower than limit 2 in spring and autumn. In summer and winter, the average traffic flow was between limits 2 and 3. The traffic flow in spring during off-peak hours ranged between limits 1 and 2. During peak hours, traffic flow ranged between limits 2 and 3. In summer and winter, more than 90% of the measured periods for traffic flow ranged between limits 2 and 3.

As shown in Figure 16b, the average traffic flow on 4-R was between limits 1 and 2 over the four seasons. In winter, the traffic flow during off-peak hours ranged between limits 1 and 2, whereas it ranged between limits 2 and 3 during peak hours.

The average traffic flow on 4-R in Figure 16c was higher than limit 1 but lower than limit 2 in spring, summer, and autumn. In winter, the average traffic flow was higher than limit 2 but lower than limit 3. In summer, autumn, and winter, the traffic flow was between limits 1 and 2 during off-peak hours. It ranged between limits 2 and 3 during peak hours.

The average traffic flow ranged between limits 1 and 3 over the four seasons. Multiple traffic flows higher than limit 1 were 2.07 on 2-R. On 4-R and 6-R, the values were 1.71 and 1.78, respectively. These data show that pedestrians who had walked along 2-R were the most vulnerable to harm.

3.5. Literature Report Comparison

Particle mass concentrations were compared with published data, and the results are summarized in

Table 5. Particulate concentrations reported in the literature.

Reference	PM2.5	PM1.0
Current study	29.2 ± 5.4	18.1 ± 10.2
Lotrecchiano et al., 2019 [27]	8–15	5.8–9.9
Rovelli et al., 2016 [59]	7.6–28.2	5.6–17.6
Shindler, 2019 [60]	4–25	3.5–15.7

. The comparison indicates that the particle mass concentrations observed in this study fall within a reasonable range when compared to similar studies. Particle mass concentrations reported in the literature typically ranged between limits 1 and 3 [27,59,60].

4. Study Limitations

This study presents several limitations that should be considered in future research endeavors. Firstly, the reliance on field measurements is subject to weather conditions, restricting the operation of measuring instruments to sunny days. Inclement weather, such as rain or snow, hinders the proper functioning of instruments, rendering the model incapable of predicting particle mass concentrations in severe weather conditions. Secondly, the limitations of the camera system prevent comprehensive monitoring of traffic flow within the street canyon during transitions between measurement points. This limitation results in a lack of real-time correlation between traffic flow data and particulate matter data. Thirdly, due to safety concerns, instrument placement is restricted to the sides of the road rather than the center, leading to challenges in accurately capturing particle mass concentrations within the street canyon and introducing errors into the model calculations. Finally, the semi-parametric physical model employed in this study is based on the assumption of a uniform distribution of particle mass concentrations in a street canyon, which may not accurately reflect real-world conditions. To enhance the model’s accuracy, future research should incorporate numerical simulations using initial boundary conditions and key parameters derived from field observations. This approach will broaden the applicability of the semi-empirical model to diverse traffic environments.

5. Conclusions

Based on the analysis conducted in this study, the main conclusions are as follows:

• The traffic flow on the three road types exhibited a single-peak distribution. The peak period occurred between approximately 17:30 to 18:30, with approximately 100–150 vehicles/10 min on 2-R, 300–400 vehicles/10 min on 4-R, and 600–700 vehicles/10 min on 6-R.
• PM_1.0 mass concentrations measured in the field for 2-R and 4-R were 18.1 ± 10.2 μg/m³ and 16.2 ± 13.1 μg/m³, respectively, which were higher than limit 2 (PM_1.0 = 15.8 μg/m³) and lower than limit 3 (PM_1.0 = 22.7 μg/m³). Meanwhile, at 4-R, the PM_1.0 mass concentration was 11.7 ± 8.43 μg/m³ and it ranged between limit 1 (PM_1.0 = 10.2 μg/m³) and limit 2. The PM_1.0 mass concentrations at the three road types demonstrated a significant seasonal pattern in winter. The PM_1.0 mass concentration showed a positive correlation with traffic flow and relative humidity and a negative correlation with temperature and wind speed.
• The prediction model for particle mass concentrations exhibited a high level of accuracy. The prediction step for the PM_1.0 mass concentration at 2-R was 14 s, while for 4-R, it was 6 s, and for 6-R, it was 5 s. The correlations between the predicted and measured values for 2-R, 4-R, and 6-R were 0.9319, 0.6582, and 0.9238, respectively.
• The traffic flow of the three road types exceeded limit 1 (recommended by the WHO), with frequencies of 63 for 2-R, 127 for 4-R, and 281 for 6-R. Among the traffic flow exceeding limit 2, 2-R = 112, 4-R = 310, and 6-R = 605. The highest frequency was found on 2-R, followed by 6-R and 4-R. The PM_1.0 mass concentration exceeding limit 2 in winter had the longest duration. For other seasons, the traffic flow only exceeded limit 2 during peak hours.

From an urban traffic construction perspective, the particle mass concentrations of the vehicle canopy height can be effectively diluted by reducing the arrangement of 2-R in urban street canyons. They can also be rationally arranged according to monsoon direction to increase the air circulation capacity in street canyons, further diluting the particle mass concentrations. From an individual perspective, pedestrians should minimize the time spent on two-lane one-way roads to reduce their exposure to particulate matter. Future research should optimize this model by using numerical simulation methods to enhance its predictive accuracy.