Concentration distribution and group disparity of traffic-derived NO₂ exposure in Baoshan District

Application of the CALPUFF model

In current environmental impact assessments and air pollutant dispersion simulation studies, small- to medium-scale air quality models are widely applied in scenarios such as industrial emissions and urban pollution control. Commonly used models include AERMOD, ADMS (Atmospheric Dispersion Modelling System), and CALPUFF. Among these, AERMOD, a steady-state Gaussian model recommended by the U.S. Environmental Protection Agency (EPA), is extensively employed to simulate pollutant dispersion over short ranges and relatively flat terrains^[5,6]. Its advantages include a well-established framework, computational efficiency, and ease of integration with meteorological data. However, AERMOD performs poorly under conditions of low wind speeds, complex topography, or non-steady-state atmospheric dynamics. ADMS, developed by the UK’s Cambridge Environmental Research Consultants (CERC), is also a steady-state model but features pollutant chemistry compared to AERMOD, demonstrating superior performance in urban micro-meteorological environments^[7]. Nevertheless, ADMS is less suitable for long-range transport modeling and often incurs higher costs due to commercial licensing requirements.

In contrast, CALPUFF employs a non-steady-state Lagrangian puff modeling approach, enabling dynamic simulation of pollutant dispersion under changing wind directions and speeds. This makes it particularly suited for simulating long-distance transport, complex terrains, and coastal or land-sea interface pollution scenarios^[8]. The model supports time-varying emission sources, dynamic meteorological inputs, and integration with diverse meteorological models^[9], thereby enabling high-resolution spatiotemporal simulations. With its non-steady-state modeling capability and flexible input mechanisms, CALPUFF exhibits strong adaptability and accuracy, positioning it as one of the most promising tools for simulating complex pollution dispersion scenarios.

In the transportation sector, CALPUFF has been widely applied in four key areas: (1) spatiotemporal simulation of air pollution; (2) assessment of exposure risks to traffic-related pollutants; (3) investigation into emission inventories; and (4) analysis of traffic impact scenarios.

Regarding spatiotemporal simulation, research primarily targets pollutants closely associated with vehicular emissions, including nitrogen oxides (NO_x^[10,11], carbon monoxide (CO)^[12], carbon dioxide (CO₂)^[13], sulfur dioxide (SO₂)^[14], and particulate matter (PM)^[15,16]). Most studies examine the spatial and temporal distribution characteristics of two or three types of pollutants simultaneously.

For assessing exposure risks, a study in the Montreal region of Canada employed a four-stage traffic model, an emissions model, and CALPUFF to quantify the total emissions associated with residents' daily travel. The findings revealed that exposure risk increases significantly during outdoor activities, with average outdoor exposure levels 23%-44% higher than those indoors, regardless of indoor air quality conditions^[17]. These results underscore the importance of incorporating travel trajectory data when evaluating individual exposure risks.

In the context of emission inventory development and traffic scenario analysis, it is crucial to allocate responsibility for emissions and formulate targeted mitigation strategies. The Montreal case study revealed that improving vehicle performance yields more substantial reductions in traffic-related pollution than investments in public transportation or other policy interventions^[17].

Air pollution exposure assessment

Accurate assessment of air pollution exposure requires not only high-resolution measurements of pollutant concentrations but also a comprehensive understanding of human activity patterns^[18]. Previous studies have largely focused on static population data, such as census statistics^[19,20] and nighttime light imagery^[21,22]. At the macro-static level, exposure risk is typically evaluated using indicators such as air quality concentration^[23], population exposure intensity^[24], and population-weighted exposure level (PWEL)^[25]. However, air quality concentration alone does not account for the spatial heterogeneity of population distribution and has been criticized for its theoretical limitations^[26]. Population exposure intensity is highly sensitive to population density, and substantial regional differences can lead to polarized exposure risk assessments. By comparison, PWEL is capable of capturing fine-scale spatial variations in exposure risk and is widely used in urban-level studies in China^[27].

To overcome the limitations of macro-static assessments that overlook individual mobility, recent studies have incorporated dynamic population distribution data, such as travel surveys^[28,29] and mobile phone signal data^[30,31], to analyze exposure during various daily activities. For example, Guo et al. used mobile phone signal data to examine spatial and temporal variations in PM_2.5 exposure risk among different socioeconomic groups in Shenzhen^[32]. Additionally, studies investigating exposure during commuting have revealed that traffic congestion and longer commuting distances significantly increase pollution exposure levels^[33].

Exposure equity and environmental justice

The principle of environmental justice asserts that all individuals, regardless of socioeconomic status, should have equitable access to public resources and share an equal burden of the health impacts resulting from environmental degradation^[34]. Originating in the United States in the 1980s, the concept initially focused on the spatial inequities associated with the placement of polluting facilities in marginalized communities, exemplified by the 1,982 protests against the siting of a hazardous waste landfill in Warren County, North Carolina^[35]. Over time, this framework has expanded to encompass disparities in air quality, access to transportation, flood risk management, and other dimensions, evolving into a critical interdisciplinary field that bridges social geography and public health.

A major subfield of environmental justice research is the assessment of equity in air pollution exposure, which seeks to identify disparities in exposure levels among different sociodemographic groups (e.g., by gender, age, income, ethnicity). Extensive evidence confirms that low-income and less-educated populations, along with children, the elderly, and ethnic minorities, often bear a disproportionate share of pollution exposure risks^[36]. For example, studies in the United Kingdom have revealed a “Matthew effect” in air quality improvement policies, whereby middle- to high-income groups - who typically have lower baseline exposure - benefit disproportionately, while high-exposure, impoverished communities suffer unintended consequences during policy transitions. This trend risks deepening regional resource allocation conflicts and perpetuating environmental injustices.

However, current research on exposure equity is largely concentrated in developed nations, particularly the United States and the United Kingdom, with relatively limited attention to developing countries such as China. Amid China’s rapid urbanization and industrialization, air pollution from transportation, ports, and construction has become increasingly intertwined with daily human activity, intensifying exposure disparities across social groups. Consequently, examining group-specific disparities in traffic-related pollution exposure in Chinese cities is now a critical pathway for promoting health equity and enhancing environmental governance.

Case study

Baoshan District, located in the northeastern part of Shanghai, falls under the city’s jurisdiction. Positioned at the confluence of the Yangtze River, Huangpu River, and Yunzaobang, the district features a 46.5-kilometer-long shoreline. Its port area connects to more than 400 ports across 164 countries and regions. As a key “waterway gateway” for Shanghai, Baoshan features an integrated transport infrastructure that links waterborne shipping with highways, railways, and urban roads [Figure 1]. Due to the area's distinctive industrial profile, a high concentration of warehousing and logistics enterprises has developed, resulting in frequent freight vehicle activity. Consequently, traffic-related air pollution in this district is notably severe.

Figure 1. Road network in the study area.

Simulation of traffic-derived NO_x pollution using the CALPUFF model

The CALPUFF modeling system consists of four main components: a preprocessing module, a meteorological module (CALMET), a dispersion module (CALPUFF), and a postprocessing module (CALPOST) [Figure 2]. First, the preprocessing module converts land use and terrain elevation data into georeferenced formats compatible with the CALMET module. It also processes surface and upper-air meteorological data. Next, CALMET generates hourly, three-dimensional diagnostic fields of wind and temperature, which characterize the local geographic and meteorological conditions. Then, the CALPUFF module simulates the atmospheric dispersion of pollutants by generating air pollutant puffs and modeling their physical transport and chemical transformation in the atmosphere. Finally, the resulting concentration data are input into the CALPOST module, which produces geographic information output files suitable for statistical analysis and ArcGIS mapping.

Figure 2. Framework of the CALPUFF model.

The CALPUFF model utilizes a three-dimensional grid system, where the X, Y, and Z axes represent the east-west, north-south, and vertical directions, respectively. In this study, the Universal Transverse Mercator (UTM) coordinate system is adopted as the projection system, with the World Geodetic System 1984 (WGS-84) serving as the geodetic datum.

The study area encompasses a 25 × 27 km region centered on Baoshan District. The UTM coordinates of the origin are (337.160, 3459.74) km, and the horizontal grid resolution is set to 0.5 km. The domain consists of 50 grid points along the X-axis (east-west) and 54 grid points along the Y-axis (north-south). Vertically, the Z-axis includes 11 levels with top heights defined as 0, 20, 40, 80, 160, 320, 640, 1,200, 2,000, 3,000, and 4,000 m.

Terrain elevation data are sourced from the USGS SRTM1 dataset. In the terrain preprocessing module, these data are converted into the “terrel.dat” file to represent the elevation of each grid cell. Land use data are derived from the USGS Global Land Cover Characteristics (GLCC) database, specifically the Asian region (USGS Global (Lambert Azimuthal) for Eurasia-Asia), with a spatial resolution of 1 km. These data are processed into the “lu.dat” file to describe land use across the grid.

Ground-level meteorological data are obtained from 11 observation stations across Shanghai, including Minhang, Baoshan, Jiading, Chongming, Xujiahui, Nanhui, Pudong, Jinshan, Qingpu, Songjiang, and Fengxian [Table 1]. The dataset spans from 00:00 on November 11 to 23:00 on November 30, 2019, and includes hourly records of wind speed (m/s), wind direction (degrees), temperature (Kelvin), cloud cover (tenths), cloud base height (hundreds of feet), surface pressure (hPa), and relative humidity (%). To maintain data continuity for model input, missing values in variables such as wind speed, temperature, or cloud cover were filled using linear interpolation based on the average of adjacent time points.

Upper-air meteorological data must include at least two observations per day, covering geopotential height, temperature, wind direction, and wind speed at the 500, 700, 850, 925, and 1000 hPa pressure levels. In this study, the data were sourced from publicly accessible datasets provided by the University of Wyoming (UWyo) and the National Oceanic and Atmospheric Administration (NOAA).

The road traffic emission data for Baoshan District [Table 2] were obtained from the Shanghai Environmental Monitoring Center. These data encompass all 4,511 major roads in the district and span the period from 00:00 on November 11 to 23:00 on November 30, 2019. The dataset includes information such as road identification number, date, time, vehicle type, road length, average vehicle speed, traffic flow, vehicle mileage, and NO_x emissions.

Conversion rate of NO₂/NO_x

Nitrogen oxide (NO_x) emissions are commonly used as a statistical metric for pollution source data. When air quality models are employed to simulate the spatial and temporal distribution of NO₂ concentrations, and photochemical models are unable to predict NO and NO₂ separately, NO₂ concentrations are typically estimated based on NO_x data. According to the Technical Guidelines for Environmental Impact Assessment - Atmospheric Environment, issued by the Ministry of Ecology and Environment of China, a conversion rate of 90% for NO₂/NO_x should be used when calculating hourly or daily mass concentrations of NO₂. For annual average mass concentrations, an estimated conversion rate of 75% is recommended^[37]. However, this standardized approach may constrain the flexibility of air pollution simulations, potentially leading to inaccurate assessments of ambient air quality. To improve this, Luo et al. proposed a proportional simulation method for converting NO_x to hourly NO₂ concentrations^[38]. This method derives the NO₂/NO_x conversion rate from hourly concentration data collected at meteorological observation stations. Results show that it more accurately and reasonably simulates the dispersion and transport of NO₂.

In this study, NO₂ concentrations are derived using an hourly NO₂/NO_x conversion rate. The data used to calculate the conversion rates were obtained from traffic monitoring stations in Shanghai, as shown in Table 3. These include five roadside air monitoring stations, one airport station, and two port stations, covering the period from 00:00 on November 11 to 23:00 on November 30, 2019. The five roadside monitoring stations are located within the Outer Ring Road, while the airport and port stations are also situated nearby.

The conversion rate calculation formula is as follows, where i represents the hour, NO_2i is the NO₂ concentration at hour i, and NO_Xi is the NO_X concentration at hour i:

Mobile phone signal data

The mobile phone signal dataset for Shanghai, covering the period from November 11 to November 30, 2019, was provided by JISMART (http://daas.smartsteps.com/, accessed on January 19, 2021). This dataset includes information on time, location, and users. Initially, it comprised data on approximately 5.28 million mobile phone users, which can be considered representative of Shanghai’s population of around 23.55 million, excluding the Chongming District^[39].

Using data from the Seventh National Population Census as a reference, SPSS 26.0 was employed to perform a correlation analysis and paired sample t-test to compare the spatial distribution, gender composition, and age structure of the resident population in Shanghai, as inferred from the mobile signaling data. These analyses aimed to verify whether significant differences exist between the two datasets and, in turn, to assess the extent to which mobile signaling data can represent the demographic distribution of Shanghai's resident population.

Regarding population distribution, the proportions of residents in each administrative district, as derived from both the census and the mobile signaling data, are presented in Table 4. The Pearson correlation coefficient was calculated at 0.998, indicating a strong linear relationship. Furthermore, the significance level of the paired sample t-test was 0.999, which is much greater than the conventional threshold of 0.05, suggesting that there is no statistically significant difference between the two datasets at the 0.05 level. Overall, these results confirm that mobile signaling data can reliably represent the population distribution across Shanghai’s various districts.

In analyzing the gender structure, the gender ratio, defined as the number of males per 100 females, was used as the evaluation metric. Table 5 presents the gender ratios for each administrative district in Shanghai based on data from the Seventh National Population Census and mobile signaling data. The Pearson correlation coefficient between the two datasets was 0.794, indicating a strong positive correlation. The paired sample t-test result was 0.374, greater than the 0.05 significance level. This suggests there is no statistically significant difference in the mean gender ratios between the two datasets, supporting the conclusion that mobile signaling data can effectively represent gender structure across different districts in Shanghai.

Regarding age structure, a comparative analysis was conducted for three age groups (0-14, 15-64, and 65 years and older) across Shanghai’s administrative districts, as shown in Table 6. The Pearson correlation coefficients for the three age groups are 0.518, 0.912, and 0.942, respectively - all above 0.5 - indicating strong correlations. The paired sample t-test results are 0.997, 0.627, and 0.499, all exceeding the 0.05 significance threshold. These results suggest that there are no significant differences in the age structure of the three groups, confirming that mobile signaling data can reliably reflect the age structure of different regions in Shanghai.

Traffic-derived NO₂ pollution exposure assessment

This study adopted the population-weighted exposure level (PWEL) as the indicator for assessing exposure to traffic-derived NO₂ pollution, as proposed by Fu and Kan^[25]. PWEL was calculated by combining predicted NO₂ concentrations with dynamic population distribution data. The CALPUFF model was used to estimate NO₂ concentrations for each standardized 1,000 × 1,000 m grid across different time periods. Hourly gridded population data were obtained from mobile phone records. PWEL was then used to evaluate the risk of exposure to traffic-derived NO₂ pollution in each grid over various time intervals. The formula is as follows:

where i denotes the grid index, n is the total number of grids, E_i represents the potential population exposure in grid i, P_i is the population of grid i during a certain period, and C_i is the NO₂ concentration in grid i. The overall population-weighted exposure level of traffic-derived NO₂ pollution in Shanghai was assessed using Formula (3), where E represents the total potential exposure across the city:

Temporal variation in the NO₂/NO_x conversion rate in Shanghai

The NO₂/NO_x conversion rate in Shanghai predominantly falls within the range of 0.3-0.7 [Figure 3], which is comparable to the range reported in Seoul (0.4-0.8)^[38]. However, it significantly deviates from the value suggested by China's Ministry of Ecology and Environment, as previously mentioned. Therefore, estimating NO₂ concentrations using hourly data from air quality monitoring stations and a ratio-based hourly simulation method is crucial for accurately modeling the temporal and spatial distribution of NO₂ exposure risk.

Figure 3. Temporal variation in the NO₂/NO_x conversion rate in Shanghai. (The broken line represents the mean hourly NO₂/NO_x conversion rate, and the grey shading area indicates the full range of hourly values).

Spatiotemporal distribution of traffic-related NO₂

There is significant spatiotemporal variability in NO₂ concentrations related to road traffic. Temporally, NO₂ concentrations exhibit a “bimodal” distribution that aligns with the typical daily traffic pattern: the highest concentrations occur during morning and evening rush hours, reaching 15.51 and 18.36 μg/m³, respectively. Concentrations are slightly lower during the daytime, averaging around 12.62 μg/m³, and drop to their lowest levels at night, averaging just 7.12 μg/m³.

Spatially, NO₂ concentrations tend to decrease from the center of the road outward to both sides. Overall, higher concentrations are observed along major roadways, including the east-west Shanghai Ring Expressway (G1503) and Outer Ring Expressway (S20), as well as the north-south Hutai Road and Yunchuan Road. Elevated concentrations are also found near warehousing and logistics hubs, such as Baoyang Road at the confluence of two rivers. Furthermore, areas prone to frequent traffic congestion, such as expressway toll stations, entrance and exit ramps, and overpasses, exhibit higher pollutant levels compared to other road segments [Figure 4].

Figure 4. Spatiotemporal simulation results of road traffic-related NO₂.

Disparities in exposure and inequality

Wang et al. explored disparities in air pollution exposure and inequality from three aspects: temporal variation, spatial variation, and group disparity^[39]. Building on their framework, this paper integrates all three dimensions, with a particular focus on examining the underlying factors that influence the impact of traffic-derived NO₂ pollution on vulnerable populations.

Temporal variation

Exposure to traffic-related NO₂ pollution peaks during rush hours, with the highest levels occurring in the evening, followed by the morning rush, daytime, and the lowest levels at night. Compared to other time periods, the evening peak displays a greater variability in exposure levels [Table 7]. This suggests that the degree to which permanent residents of Baoshan District are affected by traffic pollution during this period varies significantly, resulting in unequal levels of environmental exposure within residential areas.

Table 7

Summary statistics of traffic-derived NO₂ pollution exposure (µg/m³)

Period	Mean	Minimum	Maximum	Standard deviation
Daytime	12.62	12.31	12.75	0.15
Nighttime	6.62	6.58	6.67	0.02
Morning peak	16.00	15.83	16.14	0.10
Evening peak	18.41	18.13	18.63	0.15

This study postulates that such disparities are associated with the travel behavior typical of the evening peak. This period is characterized by more dispersed travel times, a higher proportion of non-commuting trips, and greater diversity in travel patterns. In contrast to relatively uniform commuting routines of the morning rush, the evening rush reveals substantial heterogeneity in individuals’ exposure to traffic-derived NO₂ pollution.

In terms of daily variation, exposure is higher on weekdays than on weekends [Figure 5]. This is primarily due to rigid weekday travel patterns driven by work and school schedules, resulting in dense traffic, frequent vehicle use, and elevated emissions. Consequently, residents are often forced to travel through heavily polluted areas. On weekends, travel tends to be more flexible and discretionary, with a reduction in travel frequency and lower concentrations of traffic-related NO₂ pollution, thereby decreasing exposure risk. However, this weekday-weekend distinction is less evident at night due to the continued operation of freight and logistics services. Nighttime travel tends to be limited in scope, and there is no significant variation in pollution exposure between weekdays and weekends, leading to a relatively stable risk level.

Figure 5. Time variation in traffic-derived NO₂ pollution exposure risk.

Spatial variation

In this study, the mean-standard deviation classification method is employed to depict the spatial distribution characteristics of traffic-derived NO₂ pollution exposure risk, with an emphasis on identifying areas with extreme exposure values. The mean exposure level serves as a baseline, and each grid is classified into one of four risk levels: low risk (0.5 and ≤ 1.5 SD), high risk (> 1.5 and ≤ 2.5 SD), and critical risk (> 2.5 SD).

Areas with high traffic pollution exposure are concentrated in the southern part of Baoshan District, specifically including Wusong Street and Baoshan Town near Wusong Port, Gucun Town near Baoan Road-Hutai Road, and Yanghang Town near the Yunchuan Road overpass on the Outer Ring Road [Figure 6]. These locations are not only densely populated but are also surrounded by major arterial roads that function as the main external passages for Baoshan District, making them particularly vulnerable to disruption by transit traffic.

Figure 6. Spatial variation in traffic-derived NO₂ pollution exposure risk.

Group disparity

To investigate group disparity, user attribute information, including gender and age, was extracted from mobile phone signal data. Age was categorized into four groups: juvenile (≤ 18 years), young adults (19-44 years), middle-aged adults (45-64 years), and elderly (≥ 65 years).

Gender

Overall, males experienced a higher risk of exposure to traffic-derived NO₂ pollution compared to females [Figure 7]. The gender disparity in exposure was most evident during peak hours, when the average NO₂ exposure risk for males reached 0.76 μg/m³, approximately 4.59% higher than that of females. During the daytime hours, the exposure risk for males was around 0.61 μg/m³, about 6.20% higher than that of females. The lowest disparity occurred at night, with a male exposure level of 0.40 μg/m³, approximately 9.53% higher than that of females.

Figure 7. Gender-specific assessment of exposure risk to traffic-derived NO₂ pollution.

In Baoshan District, heavy industries such as steel smelting and port cargo handling are predominantly male-dominated. These occupations typically require early-morning start times and late-evening finishes, coinciding with urban traffic peak hours. During the morning commute, male workers traveling to industrial sites are exposed to elevated NO₂concentrations due to the simultaneous start of industrial productions and increased freight vehicle activity in industrial zones. Consequently, significant pollution exposure occurs both en route and within the workplace. Similarly, at the end of the workday, another surge in pollution levels occurs in industrial districts and surrounding roadways, leading to additional NO₂ exposure for male workers. In contrast, female employment in such high-exposure industries is relatively low, contributing to their lower pollution exposure.

Male commuters are more inclined to drive private vehicles or use company-operated shuttles along expressways and arterial roads^[40]. During rush-hour congestion, these drivers are immersed in environments with elevated vehicular exhaust emissions, where NO₂concentrations are significantly higher than during off-peak hours. Additionally, male commuters typically have longer commuting distances and durations, leading to greater cumulative inhalation of air pollutants. By contrast, female commuters more often use public transportation, such as subways, which are electrically powered and primarily operate underground, thereby minimizing direct exposure to surface-level vehicle emissions. Shorter travel distances further limit exposure duration for female commuters. Thus, even during the same peak-hour periods, differences in commuting patterns and distance contribute to higher NO₂exposure among male commuters.

Beyond commuting, occupational and lifestyle differences further intensify gender disparities in pollution exposure during high-risk periods. Men are more likely to engage in occupations necessitating substantial outdoor or road-based activities - such as logistics, taxi driving, and express delivery - which typically peak during morning and evening rush hours. These jobs inherently involve prolonged exposure to ambient air pollution at times when urban air quality is most deteriorated. In contrast, a greater proportion of women are employed in indoor settings or benefit from flexible work schedules, enabling some to avoid peak commuting times altogether, particularly due to family caregiving responsibilities. As a result, the differences in daily activity patterns between men and women substantially contribute to the observed disparities in cumulative NO₂exposure.

Age

Different age groups exhibit varying levels of susceptibility to traffic-related NO₂ pollution, with the elderly experiencing the lowest exposure risk. Specifically, recorded exposure concentrations during different time periods were 10.58, 5.51, 14.67, and 15.78 μg/m³, respectively [Figure 8].

Figure 8. Age-specific risk assessment of exposure to road traffic-derived NO₂ pollution.

The observed disparities in NO₂ exposure across age groups are primarily attributable to differences in travel behavior. Working-age adults are frequently present in traffic environments during peak periods, when vehicular emissions are most intense. In contrast, older adults typically avoid peak-hour travel or remain within residential areas. Motor vehicle emissions are a major contributor to urban NO₂ pollution, accounting for over 40% of total nitrogen oxide emissions in Shanghai^[41]. Consequently, the substantial volume of NO_x generated by commuter traffic significantly elevates ambient NO₂ concentrations during rush hours. Individuals who commute during these high-exposure windows - mainly young and middle-aged adults - therefore experience markedly higher levels of NO₂ inhalation.

On the other hand, older adults and retired middle-aged individuals, whose activities are largely limited to indoor or community spaces^[42], benefit from the shielding effects of buildings and the attenuation of pollutants over distance, resulting in significantly lower exposure levels compared to individuals in close proximity to traffic corridors. Although office workers spend much of the day indoors, their cumulative NO₂ exposure is substantially increased by elevated concentrations during commuting periods. Adolescents, while spending most school hours indoors, still experience brief but significant exposure during travel to and from school. Thus, variations in daily activity patterns and mobility rhythms among age groups play a critical role in shaping the total NO₂ exposure accumulated throughout the day.

Comparative analysis

This study finds a high level of consistency with existing literature regarding the weekend effect on NO₂ pollution and the exposure characteristics of the elderly population. Regarding the weekend effect, Pommier (2023)^[43] demonstrated that NO_x emissions in UK cities were 1.54 to 3.05 times higher on weekdays than on weekends. Similarly, Demetillo et al. (2021)^[44], based on TROPOMI data, confirmed a significant decrease in NO₂ concentrations in US cities on weekends, largely due to reduced heavy-duty diesel vehicle activity. These findings align with the weekend effect observed in this study. Together, the studies suggest that traffic emissions, particularly from freight transport and commuting, are the main contributors to elevated weekday NO₂ levels.

Concerning gender differences, the findings align with Mo et al. (2023), who found that men face higher NO₂ exposure risks due to occupational factors (e.g., transportation and outdoor work) and longer durations of outdoor activity, based on a Chinese cohort study^[45]. Hu et al. (2020) reported similar results in a multi-province survey across China^[46]. However, this contrasts with the findings of Kim et al. (2019) in South Korea, which indicated that women experienced significantly higher average NO₂ exposure than men (30.11 vs. 25.83 µg/m³), potentially due to indoor environmental exposures such as cooking fuel pollution^[47]. Furthermore, Chau et al. (2002) in Hong Kong noted that long-distance commuters (typically male) are overexposed, but also suggested that women might encounter different exposure pathways related to residential environments or domestic responsibilities^[48]. In addition, Setton et al. (2010), in a simulation study, found no gender differences at all, highlighting how exposure assessment methods (e.g., variations in spatial-temporal resolution between satellite data and individual monitoring) and differences in pollution source structures (traffic-dominant versus mixed-source cities) may influence observed outcomes^[49].

The finding that elderly individuals experience lower NO₂ exposure is consistent with most literature. Mo et al. (2023), in a national cohort study, found that older individuals are less exposed to high NO₂ concentrations along transportation corridors due to limited mobility^[45]. Similarly, Chau et al. (2002) observed that reduced commuting needs among the elderly in Hong Kong contribute to lower traffic-related exposure^[48]. However, such conclusions should be interpreted with caution, as they may overlook the impact of residential pollution distribution. If elderly populations are concentrated in older urban areas or near industrial zones, their static exposure risk could be underestimated.

In summary, the weekend effect and the exposure characteristics of the elderly observed in this study align with findings from various global regions. However, the complexity of gender-based exposure differences underscores the need to account for regional cultural contexts (e.g., occupational roles and family responsibilities) and exposure pathways (i.e., the interplay between occupational and indoor microenvironment exposures). Future research could benefit from integrating high-resolution pollution data with individual behavioral trajectories (e.g., GPS-based activity patterns) to further reveal the mechanisms driving exposure disparities among population subgroups.

Spatiotemporal heterogeneity of conversion processes

The derived NO₂/NO_x ratio range (0.3-0.7), based on Luo Minghan's proportional simulation method, reflects the dynamic interplay between localized emission characteristics and atmospheric chemical processes in megacities such as Shanghai. This complexity is not adequately captured by national standard rates. Specifically:

Enhanced Photochemical Conversion: During summer in Shanghai, elevated temperatures, high relative humidity, and high solar radiation increase O₃ and VOC concentrations, which in turn promote the conversion of NO to NO₂^[50,51]. This process pushes the NO₂/NO_x ratio toward the upper limit (0.7).

Nonlinear Effects of Particulate Matter: During heavy pollution episodes, heterogeneous reactions on PM_2.5 surfaces, such as soot-catalyzed NO oxidation, further enhance NO₂ formation, contributing to an elevated ratio^[51].

Temporal Resolution Advantage: The national standard relies on annual averages, which overlook daily variations. Studies have shown that midday O₃ peaks can raise the NO₂/NO_x ratio to 0.6-0.7, while nighttime values drop to 0.3-0.4. These temporal patterns are supported by monitoring data^[52].

This broader ratio range more accurately captures the chemical complexity in high-emission, high-ozone urban environments, justifying its adoption for policy-relevant urban air quality modeling.

Group exposure disparities

This study reveals significant disparities in traffic-related NO₂ exposure across gender and age groups, highlighting the significance of activity-path-based environmental exposure assessment in environmental equity research.

Uncertainty analysis

This study integrates the CALPUFF dispersion model with dynamic mobile phone signaling data to enhance the spatiotemporal resolution of exposure assessments. However, several sources of uncertainty remain and warrant further refinement:

(1) Data Representativeness and Sampling Bias

Although the spatial distribution, gender, and age structure of the mobile phone signaling data are broadly consistent with the Seventh National Population Census, the sample covers only 23.5% of the city’s population. This may lead to underrepresentation of low-mobility groups (e.g., older adults or children). Furthermore, the inference-based classification of gender and age may introduce additional errors.

(2) Uncertainty in NO₂ Concentration Estimation

NO₂ concentrations are estimated from NO_x emissions using hourly NO₂/NO_x conversion rates (range: 0.3-0.7), which improves the temporal accuracy compared to fixed conversion factors. While this approach better captures diurnal pollution variations, it remains influenced by meteorological conditions and photochemical processes, potentially overlooking short-term spikes or atypical weather patterns.

(3) Regional Limitations

This study focuses on Baoshan District, an industrial area on the outskirts of Shanghai characterized by freight-intensive transportation, which may limit the generalizability of the findings to central urban areas or non-port cities. Nevertheless, the proposed methodology is highly transferable and offers a replicable framework for application in other regions.

(4) Temporal Resolution Constraints

While hourly simulations effectively capture broad trends in dynamic exposure, they lack the resolution needed to detect minute-level pollution peaks, such as those occurring in traffic-congested areas. Future work could incorporate higher-frequency trajectory data and real-time monitoring to refine exposure modeling.

Key findings

This study, using Shanghai’s Baoshan District as a case study, developed a dynamic exposure risk assessment framework by integrating the CALPUFF dispersion model with mobile phone signaling data, systematically revealing the spatiotemporal distribution patterns and demographic disparities associated with traffic-related NO₂ pollution. The key findings include:

Temporal Dimension: NO₂ exposure exhibited a clear bimodal diurnal pattern, with noticeable weekday-weekend variations. Concentrations peaked during the morning and evening rush hours, while weekend levels remained consistently lower. This “rhythmic” pattern highlights the strong correlation between traffic emissions and human activity, providing a quantitative basis for “time-targeted pollution control”.

Spatial Dimension: Areas of high exposure risk clustered around ports, logistics corridors, and the outer-ring expressways, forming a spatial distribution pattern closely tied to the city’s functional zoning. This spatial correlation suggests that uneven transportation infrastructure may contribute to environmental inequities.

Demographic Dimension: Exposure risks varied significantly by gender and age group. Males and working-age adults (19-44 years) experienced the highest exposure risks, whereas older adults faced the lowest risks due to their infrequent and short-distance travel behavior. These disparities reflect differences in mobility patterns and highlight the “social-behavioral and structural” nature of traffic pollution exposure.

Policy implications

Limitations

Limited Spatial Applicability: This study focuses on Baoshan, an industrial logistics-oriented district, which may limit the applicability of the findings to urban cores or non-port cities.

Data Constraints: The analysis relies on mobile signaling data to infer group attributes and movement trajectories, which may lack precision. More granular data sources (e.g., travel diaries or wearable devices) are needed to enhance accuracy.

Unquantified Health Impacts: While the study assesses exposure levels, it does not directly model the causal links between exposure and health outcomes.