Uncovering Hidden Risks: Non-Targeted Screening and Health Risk Assessment of Aromatic Compounds in Summer Metro Carriages

Abstract

Metro carriages, as enclosed transport microenvironments, have been understudied regarding pollution characteristics and health risks from ACs, especially during high-temperature summers that amplify exposure. This study applied NTS techniques for the first time across three major Chengdu metro lines, systematically identifying sixteen ACs, including hazardous species such as acetophenone, benzonitrile, and benzoic acid that are often overlooked in conventional BTEX-focused monitoring. The TAC concentration reached 41.40 ± 5.20 µg/m³, with half of the compounds exhibiting significant increases during peak commuting periods. Source apportionment using diagnostic ratios and PMF identified five major contributors: carriage material emissions (36.62%), human sources (22.50%), traffic exhaust infiltration (16.67%), organic solvents (16.55%), and industrial emissions (7.66%). Although both non-cancer (HI) and cancer (TCR) risks for all population groups were below international thresholds, summer tourists experienced higher exposure than daily commuters. Notably, child tourists showed the greatest vulnerability, with a TCR of 5.83 × 10⁻⁷, far exceeding that of commuting children (1.88 × 10⁻⁷). Benzene was the dominant contributor, accounting for over 50% of HI and 70% of TCR. This study presents the first integrated NTS and quantitative risk assessment to characterise ACs in summer metro environments, revealing a broader range of hazardous compounds beyond BTEX. It quantifies population-specific risks, highlights children’s heightened vulnerability. The findings fill critical gaps in ACs exposure and provide a scientific basis for improved air quality management and pollution mitigation strategies in urban rail transit systems.

1. Introduction

With the accelerating pace of global urbanisation and modernisation of transportation systems, urban rail transit has become a cornerstone of daily mobility for the general public [1,2]. Owing to its efficiency, punctuality, and low carbon footprint, metro systems have been extensively implemented in densely populated cities, serving as indispensable components of urban transport networks [3,4]. According to recent statistics, by the end of 2020, 538 cities across 77 countries and regions had established urban rail transit systems, with metro networks accounting for 55% of the total, covering a combined operating length of 17,584.77 km [5]. However, prolonged passenger occupancy within enclosed or semi-enclosed metro carriages has raised growing concerns regarding the implications of in-carriage air quality for public health [6,7,8]. Compared to traditional indoor environments, such as residences and offices, metro carriages are characterised by higher occupant densities, more complex pollutant sources, and frequent ventilation cycles, resulting in distinct exposure characteristics representative of transport microenvironments [9].

Metro carriages constitute unique microenvironments defined by their enclosed or semi-enclosed structural form [10], limited capacity for fresh air exchange [11], elevated passenger density [12], and complex interior construction materials [13]. In addition to infrastructural sources, passengers themselves contribute significantly to chemical emissions through perfumes, hair sprays, body odours, clothing off-gassing, and substances released from personal belongings [14]. During the summer season, elevated ambient temperatures substantially increase the volatilisation of VOCs. On one hand, interior materials such as synthetic plastics, foams, adhesives, and rubbers exhibit heightened emission potentials under thermal stress [15]. On the other hand, passenger-derived metabolic by-products, including benzaldehyde and phenolic compounds, become more prone to volatilisation due to heat and perspiration [16]. Concurrently, HVAC systems may introduce greater volumes of outdoor air to maintain thermal comfort. If metro stations are located near source of traffic-related pollution, this practice may inadvertently facilitate the ingress of exogenous VOCs [17]. These interacting factors collectively render metro carriages high-risk environment for VOC exposure during summer. Among the various VOC species identified in metro environments, including ACs, aldehydes and ketones, halogenated compounds, and residual solvents, ACs have attracted growing scientific attention due to their physicochemical stability, high lipophilicity, and marked toxicological potential [18,19].

Most previous metro air quality studies have focused primarily on benzene, toluene, ethylbenzene, and xylene, collectively referred to as BTEX, as the key monitoring targets [20,21,22]. Numerous investigations have examined the spatial distribution, concentration levels, emission sources, and associated health risks of BTEX, leading to the development of targeted mitigation strategies. However, BTEX represents only a small fraction of the broader spectrum of ACs present in metro carriages. In reality, many additional AC species, such as styrene, benzonitrile, phenolic compounds, substituted toluene, aromatic aldehydes and ketones, and certain halogenated aromatics, are also prevalent. Despite sharing comparable volatility, toxicity, and exposure potential, these compounds remained largely unmonitored in current metro air quality surveillance systems, posing an under-recognised public health. A growing body of evidence indicates that ACs can adversely affect multiple physiological systems, including the respiratory, cardiovascular, hepatic, haematological, neurological, and immune systems [23,24,25,26,27,28]. Benzene, in particular, is classified by the IARC as a Group 1 human carcinogen [29], with chronic low-dose exposure linked to leukaemia and other haematopoietic malignancies [30,31]. Other ACs also display significant toxicological profiles; for instance, styrene is categorised as a Group 2A probable human carcinogen [29], with prolonged exposure associated with hearing loss, neurobehavioural impairments, as well as pulmonary and hepatic carcinogenesis [32,33]. Similarly, long-term low-level inhalation of phenol has been linked to liver and kidney damage, haematological disorders such as haemolytic anaemia, neurotoxicity, and immune system disruption [34]. Furthermore, 1,2,3-trimethylbenzene has been implicated with adverse effects on the liver, kidneys, and blood system [35].

Despite growing awareness of the health hazards posed by ACs, methodological limitations persist in their characterisation within metro environments. Most existing studies have relied on targeted analytical approaches, which restrict to predefined set of compounds, primarily BTEX, thereby overlooking less routinely monitored species. The chemical complexity of metro carriage air, particularly under elevated temperature and multisource emissions, often exceeds the analytical reach of such approaches. In contrast, NTS has demonstrated substantial utility in other environmental domains, including atmospheric, food, and indoor air [36,37], for identifying unregulated or previously undetected compounds. However, its application in metro systems, an important urban exposure setting, remains nascent. As such, the comprehensive identification and systematic assessment of ACs in this environment remain critical knowledge gaps awaiting resolution.

Moreover, metro air quality studies focus on annual averages or wintertime with limited attention to high-temperature summer scenarios. Yet, summer heatwaves not only enhance the emission potential of ACs form interior materials and passengers, but also coincide with seasonal surges in passenger volume, particularly tourists, exacerbating exposure risks. Furthermore, existing risk assessments typically treat the metro population as homogenous, neglecting variations in exposure susceptibility across age and mobility patterns. Therefore, assessing the chemical composition and associated health burden of ACs under typical summer conditions is of considerable practical significance. Against this backdrop, the present study aims to uncover the often-overlooked risks of ACs in metro environments, while systematically addressing critical knowledge gaps in the current literature. Specifically, the key contributions of this work are as follows:

(1) This study represents the first application of NTS techniques within metro carriage environments to systematically characterise the complex spectrum of ACs in summer, thereby overcoming the inherent limitations of conventional targeted monitoring approaches;

(2) Based on quantitative NTS results, several previously unmonitored or unreported ACs, such as acetophenone, benzonitrile, and phenol, were identified, enriching the structural understanding of airborne pollutants in the metro setting;

(3) Building upon the pollutant profile established through NTS, an integrated health risk assessment framework was developed to evaluate both non-cancer and cancer risks under high-concentration summer exposure conditions. This enabled the quantitative identification of key risk-driving compounds and priority targets for control interventions.

In summary, by integrating non-targeted chemical screening with health risk assessment, this study seeks to redefine the conceptual understanding of ACs in metro environments. The findings mark a scientific transition from conventional pollutant surveillance towards proactive risk prevention and provide essential empirical evidence and methodological pathways to support the development of health-oriented urban rail transit systems.

2. Materials and Methods

2.1. Sampling Sites

This study was conducted in Chengdu, a major metropolitan centre in southwestern China. Based on spatial layout, operational features, and functional diversity, three representative metro lines, Line 1, Line 4, and Line 7, were selected to investigate the composition and pollution characteristics of ACs in metro carriages using NTS methods. Line 1, the earliest route in Chengdu’s metro system, runs along a north–south axis and connects key residential and commercial areas. Line 4 follows an east–west trajectory and intersects perpendicularly with Line 1, traversing densely populated residential zones and serving multiple transfer stations. Line 7 forms a circle line which interconnects both Lines 1 and 4, facilitating high-frequency transfers and system-wide distribution of passengers. The fundamental characteristics of the three metro lines are summarised in

Table 1. Basic information about metro lines.

Category	Line 1	Line 4	Line 7
Usage time	27 September 2010	26 December 2015	6 December 2017
Length (km)	40.99	43.30	38.61
Number of stations (#)	35	30	31
Interchanges/non-interchanges (#)	12/23	9/21	18/13
Carriage type	6-car B–type train	6-car B–type train	6-car A–type train
Service time	06:10–23:00	06:10–22:50	06:15–23:05
Direction	North–South	East–West	Circle Line
Starting station	Weijianian	Wansheng	North Railway Station
Ending station	Science City	Xihe	North Railway Station

# indicates the number of metro stations.

.

The three lines represent distinct temporal, structural, and operational configurations: Line 1 employs older B-type carriages and reflects mature operational conditions with relatively stable emissions; Line 4, also using B-type trains, serves mixed-function areas with variable occupancy levels; and Line 7, using newer A-type carriages with greater capacity, exhibits high passenger turnover and material freshness, making it more prone to intensive pollutant release. Together, these lines provide a comprehensive framework for assessing spatial–temporal variations in AC concentrations and exposure risks within the metro system.

2.2. Sampling Collection

2.2.1. Sampling Periods

The sampling campaign for this study was conducted from 12 to 18 August 2024, coinciding with typical summer heatwave conditions in Chengdu. Sampling during these thermally stressed conditions aimed to capture the worst-case exposure scenarios and facilitate the identification of a broader range of ACs, particularly in the context of NTS.

Specific protocols were established regarding both the timing and positioning of sampling. Air samples were collected separately during peak periods (07:00–09:00 and 17:00–19:00) and off-peak periods (10:00–12:00 and 14:00–16:00). Each sampling run required a full traversal of the selected metro line, with departure and terminal stations corresponding to those listed in Table 1. To minimise random error and ensure data robustness, each sampling route was repeated three times under comparable environmental conditions. Across the entire campaign, 252 samples were collected.

2.2.2. Sampling Equipment

Active sampling was employed to enrich and capture VOCs in metro carriage air. A QC-2B constant-flow gas sampler (Model QC-2B, Beijing Institute of Labour Protection, Beijing, China) was used in conjunction with TENAX-TA sorbent tubes (TENAX-TA sampling tubes, Markes Inc., Bridgend, UK) as the sampling medium. The TENAX-TA sorbent was selected for its high thermal stability, which helps minimise adsorption bias. The QC-2B sampler was configured with the following calibration parameters: an operating temperature range of 0–40 °C, a flow-rate range of 0–2.0 L/min, a flow-rate error within ±5%, and a timing range of 1–99 min. Samples were continuously collected at a stable flow rate of 0.3 L/min over a 2 h duration, resulting in a total sampling volume of 36 L per run.

2.2.3. Sampling Standards

In accordance with the Standards for indoor air quality (GB/T 18883-2022) [38], sampling points were located in the middle section of the train (third carriage) at an approximate height of 1.5 m above floor level, simulating the average human breathing zone.

2.3. Sample Processing and Analysis

2.3.1. Sample Processing, Extraction, and Analysis Procedures

The collected air samples were analysed using TD–GC–MS. TD was employed to release ACs from the sampling tubes, after which the desorbed analytes were directly introduced into a GC–MS system (Agilent 7890B/5975B, Santa Clara, CA, USA) for qualitative and quantitative analysis. Prior to analysis, each sorbent tube was placed in the thermal desorption unit, where a two-stage desorption procedure was applied. In the primary stage, dry helium was used to purge the sampling tube while heating at 250 °C for 10 min. The desorbed VOCs were then transferred for secondary desorption at 300 °C for 3 min, with a carrier gas flow rate of 55 mL/min.

Separation was achieved on an HP-VOC capillary column (60.0 m × 200 μm × 1.1 μm film thickness; Agilent, Santa Clara, CA, USA) using high-purity helium as the carrier gas at a constant flow rate of 1 mL/min. A split ratio of 1:30 was applied for analyte injection into the GC column. The oven temperature programme was as follows: an initial temperature of 40 °C held for 3 min, ramped at 15 °C/min to 160 °C with a 2 min hold, followed by a ramp of 10 °C/min to 240 °C with a final 4 min hold. Mass spectrometric detection was performed in full scan mode, scanning at 2.5 Hz (30 ≤ m/z ≤ 300). Compound identification was achieved by matching mass spectral data with the reference database of the NIST. Retention times were further confirmed by comparison with those obtained from authentic standard compounds [39,40].

2.3.2. Quality Assurance (QA)/Quality Control (QC)

A rigorous QA/QC framework was applied to ensure the accuracy and reproducibility of the collected data.

Prior to sampling: The TENAX-TA tubes underwent standardised pre-treatment to ensure analytical consistency. First, each tube was rinsed with methanol to remove residual contaminants. Tubes were then thermally conditioned in a TD-100 thermal desorption unit (Markes Inc., UK) at 250–300 °C under a flow of high-purity nitrogen or helium to eliminate remaining VOCs. All sorbent tubes were thermally desorbed and conditioned prior to sampling. Daily analytical blanks were included to assess potential background interference. After conditioning, the tubes were cooled under clean conditions and immediately sealed at both ends with sterile silicone or PTFE caps to prevent ambient contamination.

Sampling: The QC-2B constant-flow gas sampler was calibrated using a bubble flow meter (Mini-BUCK M-5, Post Falls, ID, USA) to maintain a flow rate of 0.3 L/min, within a ±5% error margin. Before each sampling run, the pump was flushed for 30 s to stabilise airflow. To control for background interference and potential contamination from sampling tubes or analytical procedures, one unused tube per day and per metro line was designated as a blank control, yielding a total of 21 blank samples.

Post-sampling: Immediately after sampling, tubes were sealed with aluminium foil, wrapped, and stored at 4 °C to minimise compound loss or degradation. All samples were transported to the laboratory at Southwest Jiaotong University within 24 h and analysed without delay to ensure data reliability. After standardised cleaning and thermal desorption, the tubes were reused in subsequent sampling cycles.

Analytical: Quantification was performed using external standard calibration. Mixed standard solutions at three concentrations (10, 100, and 1000 mg/L) were purchased from the Research Centre for Standard Substances, Ministry of Ecology and Environment of China. A calibration curve was established using standard masses of 0.02, 0.05, 0.08, 0.1, 0.2, 0.5, and 0.8 μg, with coefficients of determination (R²) exceeding 0.99 for all target compounds. The method detection limit for individual analytes was approximately 1 ng, defined at a signal-to-noise ratio (S/N) of 3:1. Concentrations of ACs were quantified as toluene equivalents based on the corresponding calibration curves [41].

Reproducibility validation: To ensure the reproducibility of the sampling protocol, parallel sampling was conducted on two randomly selected consecutive days. Three replicate samples were collected from the same location during the same time window, one set during peak periods and one during off-peak periods, yielding a total of 12 parallel samples. The results showed an average RSD of less than 15%, indicating satisfactory repeatability.

Additional measures: Routine laboratory blank analyses were performed to exclude the possibility of contamination during sample handling or analysis. The GC–MS instrument was regularly maintained and calibrated, and standard calibration curves were periodically updated to ensure sustained analytical accuracy.

2.4. PMF Method

PMF is a widely applied receptor model for source apportionment, originally developed by Paatero in 1993 [42]. It is recognised and recommended by the US EPA as a standard source apportionment tool. Unlike traditional models that require prior knowledge of source profiles, PMF estimates both source contributions and profiles solely based on ambient concentration data and associated uncertainties.

In this study, PMF 5.0 was employed to identify and quantify the potential sources of ACs. The model is based on factor analysis and solves the matrix using a least-squares approach to minimise the residuals between observed and modelled concentrations. A key advantage of PMF over conventional models is its imposition of non-negativity constraints on both factor profiles and contributions, thereby ensuring physically interpretable and statistically robust results. This constraint eliminates the possibility of negative source contributions, which may occur in other receptor model outputs. PMF belongs to the class of multivariate statistical models, designed to deconvolute complex environmental mixtures by distinguishing different source contributions. The model assumes that each measured sample concentration can be expressed as the linear combination of contributions from multiple factors, plus a residual term, as shown in Equation (2). When the observed concentration was lower than the MDL, the uncertainty is calculated using Equation (3). If the observed concentration exceeds the MDL, Equation (4) is used [43,44].

$$X_{ij}={\displaystyle \sum _{k=1}^n{g}_{ik}{f}_{kj}}+e_{ij}$$

(1)

$$U_{nc}={\displaystyle \frac{5}{6}}\times MDL$$

(2)

$$U_{nc}=\sqrt{{\left(EF\times X_{ij}\right)}^2+{\left(0.5\times MDL\right)}^2}$$

(3)

where X_ij is the concentration of species j component in sample i, g_ik is the relative contribution of factor k to sample i; f_kj is the concentration of species j in factor k; e_ij is the residual of species j in sample i; U_nc is the uncertainty; MDL is the method detection limit and EF is the error fraction.

The concentrations of each AC, along with their corresponding uncertainty values, were input into the US EPA PMF 5.0 model for source apportionment analysis. The number of factors was varied from 3 to 8 to explore optimal solutions. The model was run in Robust mode with 20 iterations, aiming to minimise the Q value while ensuring that the elements of the residual matrix (E) predominantly fell within the range of −3 to +3, consistent with a normal distribution. This approach ensured that the model outputs were well-correlated with the observed data and statistically reliable.

2.5. Health Risk Assessment Model

2.5.1. Inhalation Exposure

In this study, respiratory exposure to ACs in indoor air was assessed using the CDI method. The calculation followed the guidelines outlined in the RAGS issued by the US EPA, with the CDI formula presented in Equation (4) [45]. This approach has been widely adopted in studies concerning exposure assessment in indoor environments.

$$CD{I}_{inh}={\displaystyle \frac{C\times IR\times ET\times EF\times ED}{BW\times AT}}$$

(4)

During the summer season, Chengdu, as a popular tourist destination, experiences a substantial influx of external visitors, leading to elevated occupancy levels in metro carriages. Consequently, it is essential to include such populations within the health risk assessment framework. As physiological characteristics such as inhalation rates and body weight, vary notably by age and sex, exposure parameters must be tailored accordingly. To ensure the scientific validity of the estimates, this study adopted demographic-specific parameters from the Exposure Factors Handbook of Chinese Population [46,47]. The detailed exposure parameters are presented in

Table 2. Significance and values of exposure parameters for different population.

Symbol	Meaning	Unit	Commuter		Tourists		Ref.
			Adult	Child	Adult	Child
C	Concentration	mg/m3	Actual measured concentration				This study
IR 1	Inhalation rate	m3/h	0.384	0.233	0.492	0.398	[46,47]
ET 2	Exposure time	hours/day	2	2	3	3	2
EF 3	Exposure frequency	days/year	350	350	350	350	[48]
ED	Exposure duration	years	59.93	18	59.93	18	3
BW	Body weight	kg	59.3	20.8	60.6	21.2	[46,47]
AT	Average exposure time (non-cancer)	days	15,330	3285	15,330	3285	[49,50]
AT	Average exposure time (cancer)	days	25,550	25,550	25,550	25,550	[49,50]

¹ The IR of commuters refers to the activity level of “sitting”, while that of travelers refers to the activity level of “light activity” [46,47]. ² According to Li et al. [48], a survey conducted in Chengdu found that the average daily metro travel time for commuters is approximately 2 h. For tourists, a longer ET of 3 h per day was assumed to reflect the extended travel typical during peak summer periods, including cross-line journeys, sightseeing activities, and interchange delays. Although direct observational or individual-level data were not available to validate these durations precisely, the adopted estimates are grounded in contextual reasoning and are considered reasonable. ³ The ED is estimated according to the actual conditions. The ED for children is taken from birth to adulthood and it is taken from adulthood to death for adults. According to the China Statistical Yearbook, the average life expectancy for Chinese population is 77.93 [51].

.

2.5.2. Non-Cancer and Cancer Risk

The health risks associated with human exposure to ACs were assessed in accordance with methodologies established by the US EPA in its IRIS [52] and RAGS [53], as well as the WHO Global Air Quality Guidelines [54] Two risk metrics were applied: the HI for evaluating non-cancer risk, and the TCR for estimating cancer risk resulting from the inhalation of ACs in metro carriages. The calculation methods for HI and TCR are provided in Equations (5)–(8) [55,56]. The reference dose (RfD) and cancer slope factor (SF) values used in these calculations are summarised in

Table 3. RfD and SF values.

ACs	CAS 1	RfD [mg/(kg⋅d)] 2	SF [(kg⋅d)/mg] 3
benzene	71-43-2	0.00388	0.0145
toluene	108-88-3	0.064	NA
ethylbenzene	100-41-4	0.05	0.00385
m-xylene	108-38-3	0.0286	NA
p-xylene	106-42-3	0.0286	NA
chlorobenzene	108-90-7	0.0062	NA
phenol	108-95-2	0.27	NA
1,4-dichlorobenzene	106-46-7	0.07	0.0054
ethoxybenzene	103-73-1	NA	NA
styrene	100-42-5	0.16	NA
benzaldehyde	100-52-7	0.1	NA
benzonitrile	100-47-0	NA	NA
acetophenone	98-86-2	0.1	NA
benzoic acid	65-85-0	4	NA
3-ethyltoluene	620-14-4	NA	NA
1,2,3-trimethylbenzene	526-73-8	0.01	NA

¹ CAS: A surrogate for the unique identifier of a substance in the field of biochemistry. ² Values of RfD and SF are referenced from the IRIS of the US EPA. ³ NA: Not available.

.

$$H{Q}_{inh}={\displaystyle \frac{CD{I}_{inh}}{RfD}}$$

(5)

$$H{I}_{inh}={\displaystyle \sum H{Q}_{inh}}$$

(6)

$$C{R}_{inh}=CD{I}_{inh}\times SF$$

(7)

$$TC{R}_{inh}={\displaystyle \sum C{R}_{inh}}$$

(8)

where CDI is chronic daily intake, mg/(kg⋅d); HQ is hazard quotient; RfD is reference dose, mg/(kg⋅d); HI is hazard index; CR is cancer risk; SF is slope factor, (kg⋅d)/mg; TCR is total cancer risk.

According to established health risk assessment criteria [57,58,59,60], an HQ or HI value below 1 indicates an acceptable exposure level, suggesting negligible non-carcinogenic risk. A value equal to 1 suggests exposure approaching the reference dose, potentially posing minor health effects. Values exceeding 1 imply a potential non-cancer health risk, whereby chronic exposure may lead to adverse effects. For cancer risk, a CR or TCR below 10⁻⁶ is considered negligible and does not warrant intervention. When CR or TCR falls between 10⁻⁶ and 10⁻⁴, a low level of cancer risk is assumed, for which precautionary control measures are recommended. A CR or TCR above 10⁻⁴ indicates an unacceptable level of risk, necessitating urgent pollution control actions to protect human health.

2.6. Data Analysis

To determine whether the concentrations of Acs in metro carriages varied significantly under different conditions, appropriate statistical tests were selected based on the distributional characteristics of the sample data. The Shapiro–Wilk test was first employed to assess the normality of each dataset. When both groups satisfied the assumption of normality, an Independent Samples t-test was applied. This test is suitable for comparing the means of two unrelated groups, where no interdependence exists between observations. For datasets that did not meet the normality assumption, the Wilcoxon signed-rank test was used as a non-parametric alternative to assess differences between the groups. All statistical analyses were conducted using SPSS Statistics 26.0 (IBM Corp., Armonk, NY, USA). Significance levels were set at 0.05 and 0.01, corresponding to confidence levels of 95% and 99%, respectively. A significance level of 0.01 denotes a more stringent criterion for detecting statistical differences than 0.05. The p-value was used to determine statistical significance; results with p-values ≤ 0.05 or ≤0.01 were interpreted as indicating statistically significant differences.

3. Results

3.1. Composition and Pollution Levels of ACs Based on NTS

Using the NTS method, a total of sixteen ACs were qualitatively and quantitatively identified in the air of Chengdu metro carriages during the summer. The concentration profiles of these compounds are summarised in

Table 4. NTS of ACs concentrations in metro carriages (μg/m³).

ACs	Min.	Max.	Mean	SD	95% CI
benzene	1.31	2.30	1.78	0.39	(1.37, 2.19)
toluene	2.64	4.98	4.17	0.99	(3.14, 5.21)
ethylbenzene	0.79	1.29	0.90	0.19	(0.69, 1.10)
m-xylene	0.97	1.98	1.21	0.39	(0.80, 1.63)
p-xylene	2.06	4.68	2.67	1.01	(1.62, 3.73)
chlorobenzene	0.41	0.52	0.46	0.04	(0.42, 0.50)
phenol	1.93	3.55	2.61	0.61	(1.97, 3.25)
1,4-dichlorobenzene	0.66	1.10	0.87	0.17	(0.69, 1.05)
ethoxybenzene	0.36	0.40	0.38	0.02	(0.36, 0.40)
styrene	1.48	6.60	2.88	1.92	(0.87, 4.89)
benzaldehyde	1.45	2.62	2.04	0.44	(1.58, 2.49)
benzonitrile	0.98	1.90	1.28	0.35	(0.92, 1.65)
acetophenone	4.78	8.57	6.07	1.50	(4.50, 7.64)
benzoic acid	11.42	14.84	12.48	1.27	(11.15, 13.81)
3-ethyltoluene	0.42	0.89	0.56	0.17	(0.38, 0.74)
1,2,3-trimethylbenzene	0.71	1.90	1.04	0.45	(0.56, 1.51)

. Benzoic acid exhibited the highest average concentration, reaching 12.48 ± 1.27 µg/m³ (95% CI: 11.15–13.81 µg/m³), followed by acetophenone at 6.07 ± 1.50 µg/m³ (95% CI: 4.50–7.64 µg/m³) and toluene at 4.17 ± 0.99 µg/m³ (95% CI: 3.14–5.21 µg/m³). In contrast, the lowest concentrations were recorded for ethoxybenzene (0.38 ± 0.02 µg/m³), chlorobenzene (0.46 ± 0.04 µg/m³), and 3-ethyltoluene (0.56 ± 0.17 µg/m³). Analysis of data dispersion revealed considerable variability for styrene (2.88 ± 1.92 µg/m³; range: 1.48–6.60 µg/m³) and p-xylene (2.67 ± 1.01 µg/m³; range: 2.06–4.68 µg/m³). Among the typical BTEX compounds, benzene, ethylbenzene, and m-xylene were detected at 1.78 ± 0.39 µg/m³ (95% CI: 1.37–2.19 µg/m³), 0.90 ± 0.19 µg/m³ (95% CI: 0.69–1.10 µg/m³), and 1.21 ± 0.39 µg/m³ (95% CI: 0.80–1.63 µg/m³), respectively. Other detected compounds included phenol (2.61 ± 0.61 µg/m³), 1,4-dichlorobenzene (0.87 ± 0.17 µg/m³), benzaldehyde (2.04 ± 0.44 µg/m³), benzonitrile (1.28 ± 0.35 µg/m³), and 1,2,3-trimethylbenzene (1.04 ± 0.45 µg/m³). The identification of compounds such as benzonitrile, acetophenone, and benzoic acid, which are typically absent in standard BTEX profiles, underscores the unique capacity of NTS to reveal overlooked yet toxicologically relevant ACs in metro environments.

3.2. ACs Concentration Levels in Metro Carriages

3.2.1. Differences in AC Pollution on Different Metro Lines

Differences in the total concentrations of sixteen ACs across the three metro lines are illustrated in Figure 1a. The average TAC concentration in metro carriages was 41.40 ± 5.20 µg/m³. Among the three lines, Line 7 exhibited the highest average AC concentration at 46.51 ± 1.95 µg/m³, representing increases of 15.2% and 24.6% compared to Line 4 (40.37 ± 6.60 µg/m³) and Line 1 (37.33 ± 0.49 µg/m³), respectively. Line 4 also showed a moderate increase of 8.1% relative to Line 1. Pairwise comparisons using the independent samples t-test revealed statistically significant differences between Line 1 and Line 4, as well as between Line 4 and Line 7 (p < 0.05). The difference between Line 1 and Line 7 was more pronounced, reaching a higher level of significance (p < 0.01).

The distributions of individual AC concentrations across the three metro lines are presented in Figure 1b–d. Benzoic acid consistently emerged as the dominant compound, contributing 29.5–31.3% of the TAC concentration (Line 1: 11.70 ± 0.39 µg/m³; Line 4: 12.03 ± 0.83 µg/m³; Line 7: 13.70 ± 1.61 µg/m³). Chlorobenzene and ethoxybenzene displayed minimal spatial variability, with concentrations narrowly ranging from 0.45–0.49 µg/m³ and 0.37–0.39 µg/m³, respectively. Styrene exhibited a concentration peak on Line 7 (4.83 ± 2.51 µg/m³), which was approximately 2.5 times that observed on Line 1 (1.92 ± 0.62 µg/m³). p-xylene reached its maximum concentration on Line 4 (3.45 ± 1.75 µg/m³), showing a substantial 62.7% increase over Line 1 (2.12 ± 0.09 µg/m³). Toluene concentrations were highest on Line 7 (4.57 ± 0.49 µg/m³), representing a 19.9% increase relative to Line 1 (3.81 ± 1.66 µg/m³). In contrast, m-xylene peaked on Line 4 (1.48 ± 0.71 µg/m³), which was 51.2% higher than its concentration on Line 1 (0.98 ± 0.02 µg/m³). Benzene exhibited relatively moderate spatial variation, with concentrations ranging from 1.59 to 1.90 µg/m³, and a maximum inter-line difference of only 19.5%.

3.2.2. Different Periods

The concentration distributions of the sixteen identified ACs during peak and off-peak periods are illustrated in Figure 2. Temporal variations revealed a pronounced peak-period accumulation of ACs on Lines 4 and 7. On Line 4, the TAC concentrations during the peak period reached 45.03 ± 0.89 µg/m³, representing an absolute increase of 9.33 µg/m³ (a 26.14 rise) compared to the off-peak level of 35.70 ± 0.62 µg/m³ (p < 0.05). Similarly, Line 7 exhibited a peak-period concentration of 47.89 ± 1.38 µg/m³, 6.10% higher than its off-peak value of 45.13 ± 0.75 µg/m³. In contrast, Line 1 displayed an atypical trend: the off-peak concentration (37.68 ± 0.47 µg/m³) was 1.89% higher than the peak-period level (36.98 ± 0.82 µg/m³), suggesting a unique temporal pattern potentially attributable to operational characteristics specific to that metro line.

3.2.3. Pollution Levels of ACs During Different Peak and Off-Peak Periods

The concentration distributions of the sixteen identified ACs during peak and off-peak periods are illustrated in Figure 2. Temporal variations revealed a pronounced peak-period accumulation of ACs on Lines 4 and 7. On Line 4, the TAC concentrations during the peak period reached 45.03 ± 0.89 µg/m³ (Figure 2d), representing an absolute increase of 9.33 µg/m³ (a 26.14 rise) compared to the off-peak level of 35.70 ± 0.62 µg/m³ (p < 0.05) (Figure 2c). Similarly, Line 7 exhibited a peak-period concentration of 47.89 ± 1.38 µg/m³ (Figure 2f), 6.10% higher than its off-peak value of 45.13 ± 0.75 µg/m³ (Figure 2e). In contrast, Line 1 displayed an atypical trend: the off-peak concentration (37.68 ± 0.47 µg/m³) (Figure 2a) was 1.89% higher than the peak-period level (36.98 ± 0.82 µg/m³) (Figure 2b), suggesting a unique temporal pattern potentially attributable to operational characteristics specific to that metro line.

Among the sixteen detected ACs, eight species (phenol, ethoxybenzene, benzaldehyde, benzonitrile, acetophenone, benzoic acid, 3-ethyltoluene, and 1,2,3-trimethylbenzene) consistently exhibited higher concentrations during peak periods across all three lines (

Table 5. ACs with increased concentrations during the peak period and increase rates.

Compound	Line 1			Line 4			Line 7
	Concentration (µg/m3)		Increase Rate (%)	Concentration (µg/m3)		Increase Rate (%)	Concentration (µg/m3)		Increase Rate (%)
	Off-Peak	Peak		Off-Peak	Peak		Off-Peak	Peak
phenol	2.29	2.33	1.42	1.93	2.43	26.47	3.14	3.55	12.93
ethoxybenzene	0.36	0.39	7.05	0.36	0.40	11.98	0.38	0.40	3.87
benzaldehyde	1.87	2.46	31.41	1.45	2.04	40.49	1.78	2.62	47.48
benzonitrile	1.01	1.08	7.53	0.98	1.90	93.14	1.27	1.46	14.38
acetophenone	4.78	5.46	14.20	4.79	7.13	48.88	5.71	8.57	49.91
benzoic acid	11.42	11.98	4.88	11.45	12.62	10.24	12.56	14.84	18.12
3-ethyltoluene	0.48	0.89	84.44	0.46	0.55	17.97	0.42	0.55	31.25
1,2,3-trimethylbenzene	0.87	1.90	119.70	0.71	1.12	58.49	0.73	0.89	21.40

). For instance, the peak vs. off-peak concentrations of benzoic acid were as follows: Line 1, 11.98 ± 0.64 µg/m³ vs. 11.43 ± 0.72 µg/m³ (4.88% increase); Line 4, 12.62 ± 0.51 µg/m³ vs. 11.45 ± 0.82 µg/m³ (10.24% increase); and Line 7, 14.84 ± 0.68 µg/m³ vs. 12.56 ± 0.62 µg/m³ (18.12% increase). Acetophenone exhibited peak-period increases of 14.2%, 7.13%, and 49.91% on Lines 1, 4, and 7, respectively. Phenol concentrations rose by 1.42%, 26.47%, and 12.93% across the three lines. These patterns suggest a potential correlation between increased passenger density and elevated emission levels of these compounds.

Notably, Line 4 demonstrated a general increase in compound concentrations during peak periods, implying that higher occupancy may facilitate pollutant accumulation. By contrast, Line 1 and Line 7 displayed a reversed pattern for the majority of compounds, with higher concentrations observed during off-peak periods. This divergence from the conventional peak-period accumulation model may be attributed to variations in HVAC operational strategies or continuous emissions from interior materials during less ventilated off-peak conditions.

Figure 2. Sixteen ACs concentration distribution across three lines in different periods (a,b) Line 1 in off-peak and peak period; (c,d) Line 4 in off-peak and peak period; (e,f) Line 7 in off-peak and peak period.

Figure 2. Sixteen ACs concentration distribution across three lines in different periods (a,b) Line 1 in off-peak and peak period; (c,d) Line 4 in off-peak and peak period; (e,f) Line 7 in off-peak and peak period.

3.3. Source Apportionment of ACs in Metro Carriages

3.3.1. Diagnostic Ratios

To preliminarily identify potential sources of ACs in metro carriages, three widely used diagnostic ratio methods were employed in this study: the toluene/benzene ratio (T/B), xylene/ethylbenzene ratio (X/E), and styrene/benzene ratio (S/B). The results, summarised in

Table 6. Metro line diagnostic ratio results.

Line	Period	T/B	X/E	S/B
Line 1	Off-peak	2.66	4.01	1.26
Line 1	Peak	2.01	3.57	1.13
Line 1	Whole	2.40	3.78	1.21
Line 4	Off-peak	2.92	4.06	0.89
Line 4	Peak	1.57	5.18	1.07
Line 4	Whole	2.17	4.74	0.99
Line 7	Off-peak	1.84	4.61	2.87
Line 7	Peak	3.52	4.07	2.19
Line 7	Whole	2.47	4.32	2.61

, cover the variations in these ratios across Line 1, Line 4, and Line 7 during peak periods, off-peak periods, and the whole periods.

A T/B ratio lower than or close to 1 is typically indicative of biomass or fossil fuel combustion sources [61]; a value around 2 suggests dominant traffic emissions [62]; while T/B > 2.5 is generally associated with industrial emissions and solvent usage [63]. In this study (Table 6), T/B values on Line 1 and Line 4 exceeded 2.5 during off-peak periods, suggesting a predominance of solvent-related sources, such as emissions from carriage materials or fragrance products carried by passengers in low-ventilation conditions. Conversely, the T/B value on Line 4 decreased to 1.57 during peak periods, falling closer to the traffic emission range and indicating greater infiltration of external pollutants. On Line 7, the T/B value peaked at 3.52 during peak periods, significantly exceeding 2.5 and implying strong influence from industrial or solvent-related sources, potentially exacerbated by high passenger density and off-gassing from newer materials.

The X/E ratio is commonly used to assess the freshness of emissions. Values greater than 3 typically reflect fresh emissions, whereas values below 3 suggest photochemical ageing of the air mass [64,65]. In this study (Table 6), all X/E values across lines and periods were above 3, indicating that the detected xylene and ethylbenzene were predominantly from fresh emissions. These may stem from recently infiltrated traffic-related pollutants or continuous volatilisation from carriage interiors.

The S/B ratio serves as an indicator of emissions from polymer degradation or rubber product volatilisation [66]. Styrene is generally released from heated plastics, foams, or rubber components; thus, S/B > 1 indicates a significant contribution from internal material emissions. As shown in Table 6, Line 7 consistently exhibited S/B values well above 1 across all periods, suggesting that styrene emissions primarily originated from within the carriage, particularly in newer and more airtight vehicles. In contrast, Line 1 and Line 4 showed average S/B values of 1.21 and 0.99, respectively, implying weaker contributions from styrene, likely due to material emission stabilisation over longer vehicle service life.

In summary, the combined analysis of T/B, X/E, and S/B ratios reveals that AC pollution in Chengdu metro carriages originates from multiple sources. These include both external inputs, such as traffic emissions entering from tunnels, and internal sources, including off-gassing from interior materials, human-related emissions, and accumulation in confined spaces. The spatial and temporal variability of diagnostic ratios highlights the heterogeneity of pollution sources and emission dynamics across different metro lines and operational periods.

3.3.2. PMF

Building upon the preliminary source identification presented in Section 3.2.1, this subsection employs PMF to further resolve and quantify the major sources of ACs pollution and their respective contributions (Figure 3).

The PMF analysis identified five distinct sources contributing to AC concentrations. Factor 1 was characterised by high loadings of benzaldehyde (38.60%), acetophenone (33.38%), and phenol (31.86%). These compounds are commonly associated with human metabolic emissions, as well as volatilisation from personal care and cosmetic products, and were therefore attributed to human emissions. Factor 2 exhibited significant loadings of p-xylene (32.75%), benzene (30.82%), m-xylene (28.44%), and toluene (23.30%). As key members of the BTEX group, well-established markers of vehicular exhaust emissions [61,67], this factor was assigned to traffic exhaust infiltration. Factor 3 showed dominant contributions from 1,2,3-trimethylbenzene (36.84%) and 3-ethyltoluene (28.57%), both frequently used as industrial solvents [68,69]. This factor was thus interpreted as reflecting industrial emissions. Factor 4 was dominated by styrene, with an exceptionally high loading of 74.89%. Additional contributions were observed from toluene (45.17%), benzene (44.56%), phenol (40.78%), 1,4-dichlorobenzene (37.87%), and m-xylene (34.24%). These compounds are typically associated with emissions from carriage interior materials such as seat coverings, flooring, and polymer-based fixtures. As such, this factor was assigned to composite emissions from interior materials. Factor 5 was primarily associated with p-xylene (33.39%), benzonitrile (31.34%), and m-xylene (30.68%). These compounds found in solvent mixtures, suggesting that this factor corresponds to an organic solvent source.

In summary, the sources of ACs in metro carriages during the summer were identified as carriage materials release (36.62%), human emissions (22.50%), traffic exhaust penetration (16.67%), organic solvent sources (16.55%), and industrial emissions (7.66%).

3.4. Health Risk Assessment for Different Population Groups

In this study, the health risks associated with exposure to ACs in Chengdu metro carriages were quantitatively assessed, encompassing both commuters and tourists across different age groups (adults and children), as shown in

Table 7. Health risks of ACs in metro carriages for different population groups.

Population	Group	HI	TCR
Commuter	Adult	1.45 × 10−2	3.09 × 10−7
Commuter	Child	7.53 × 10−3	1.88 × 10−7
Tourists	Adult	2.73 × 10−2	5.83 × 10−7
Tourists	Child	1.89 × 10−2	4.72 × 10−7

. Non-cancer risks were evaluated using HI, with all assessed populations exhibiting HI values below 1. The highest HI was observed in adult tourists (2.73 × 10⁻²), while the lowest was recorded for commuting children (7.53 × 10⁻³). According to established risk assessment criteria, an HI value below 1 indicates that exposure is within an acceptable threshold [70], suggesting that current ACs exposure levels do not pose significant non-cancer risks to the evaluated populations. Cancer risks were assessed using the TCR, which estimates the probability of developing cancer under long-term exposure. All groups had TCR values below 10⁻⁶, with the highest risk again observed in adult tourists (5.83 × 10⁻⁷) and the lowest in commuting children (1.88 × 10⁻⁷), both well below the commonly accepted threshold of 10⁻⁴ [49]. These findings indicate that the cancer risks associated with current AC exposure in metro environments are negligible and do not warrant specific intervention measures. Comparative analysis revealed that health risk values for tourists (adults and children) were approximately 1.89 and 2.51 times higher, respectively, than those for their commuting counterparts, underscoring the enhanced exposure vulnerability of transient urban populations during peak travel periods.

Health risk assessment of 13 ACs included in this study revealed that the HQ values for all individual pollutants across four demographic groups, adult commuters, adult tourists, child commuters, and child tourists, were well below the safety threshold (HQ < 1) [71], indicating no observable non-cancer risk from any single compound (Figure 4a−d). Among all compounds, benzene exhibited the highest HQ values, measured at 8.13 × 10⁻³ (adult commuters), 1.53 × 10⁻² (adult tourists), 4.22 × 10⁻³ (child commuters), and 1.06 × 10⁻² (child tourists). Risk contribution analysis identified benzene as the dominant contributor to non-cancer risk in metro carriages, accounting for approximately 52% of the HI across all population groups (52.07–52.18%). This was approximately 4.4 times higher than the second-highest contributor, 1,2,3-trimethylbenzene (11.75–11.78%). The top five contributors (benzene, 1,2,3-trimethylbenzene, p-xylene, acetophenone, and toluene) jointly accounted for 90.80%, 90.91%, 90.87%, and 90.71% of the total HI in adult commuters, adult tourists, child commuters, and child tourists, respectively. All remaining compounds contributed less than 1.5% individually and collectively accounted for less than 10% of the total non-cancer risk (Figure 4a–d).

For CR (Figure 4e–h), three known carcinogens: benzene, ethylbenzene, and 1,4-dichlorobenzene, were evaluated. All CR values were below the negligible risk threshold (1 ×10⁻⁶), with benzene again showing the highest risk, at 2.35 × 10⁻⁷ (adult commuters), 4.43 × 10⁻⁷ (adult tourists), 1.43 × 10⁻⁷ (child commuters), and 3.58 × 10⁻⁷ (child tourists). This was followed by 1,4-dichlorobenzene (ranging from 2.60 × 10⁻⁸ to 8.09 × 10⁻⁸) and ethylbenzene (1.91 × 10⁻⁸ to 5.93 × 10⁻⁸). Benzene was identified as the principal contributor to cancer risk, accounting for approximately 73% of the TCR across all groups (73.59–73.73%), significantly exceeding the contributions of 1,4-dichlorobenzene (13.43–13.49%) and ethylbenzene (9.84–9.94%). Notably, the intergroup variation in contribution proportions across the three carcinogens was less than 0.3%, suggesting a consistent risk pattern among different demographic populations.

4. Discussion

4.1. Compositional Characteristics of ACs Based on NTS

NTS conducted in Chengdu metro carriages during the summer identified sixteen ACs (Table 4), revealing a compositional profile dominated by BTEX compounds alongside notable oxygenated derivatives. While classical traffic-related pollutants such as BTEX remained the predominant components, the consistently high concentrations of benzoic acid and acetophenone were particularly noteworthy. These two compounds, largely overlooked in previous metro studies, now contribute substantially to TAC concentrations, alongside toluene and benzene. Indeed, the top five contributors, benzene, 1,2,3-trimethylbenzene, p-xylene, acetophenone, and toluene, accounted for approximately 90% of the identified aromatic hydrocarbon mass, suggesting a compositional skew toward a limited number of dominant species. Benzoic acid (an aromatic carboxylic acid) and acetophenone (an aromatic ketone) are commonly emitted from coatings, plasticisers, and synthetic resin-based materials, indicative of ongoing off-gassing from interior carriage components [72,73]. Unlike traditional traffic-derived VOCs, these compounds point to the increasing significance of emissions from internal carriage materials. Notably, benzoic acid accounted for over 30% of the TAC burden across all three metro lines and exhibited limited spatial variability, suggesting its potential as a reliable tracer for in-carriage pollution.

This compositional profile differs from those reported in other urban metro systems, where BTEX compounds dominate more exclusively. For instance, in the Guangzhou metro, toluene alone accounted for approximately 60% of aromatic VOCs within carriages, highlighting the dominant influence of vehicle exhaust [74]. Similarly, studies in the Shanghai metro have identified BTEX and formaldehyde as the main pollutants, reflecting common external sources [20]. While the results from Chengdu confirm the importance of BTEX, they also underscore the growing presence of less conventional aromatic species. The elevated concentrations of acetophenone and benzoic acid suggest that metro-specific factors, such as emissions from carriage components or the use of scented disinfectants, may play a more substantial role in shaping VOC composition than previously recognised. This broader aromatic profile provides a more complete pollution spectrum within metro environments, capturing both traditional traffic-related compounds and secondary emissions specific to enclosed transit settings. The findings highlight the need for integrated management strategies that address both external and internal sources to improve in-carriage air quality in underground transit systems.

4.2. ACs Concentration Levels in Metro Carriages

4.2.1. Different Lines

The results of this study revealed significant differences in AC concentrations among the three metro lines, with mean levels of 37.33 ± 0.49 µg/m³ on Line 1, 40.37 ± 6.60 µg/m³ on Line 4, and 46.51 ± 1.95 µg/m³ on Line 7 (Figure 1). These differences can be attributed to the combined influence of several factors, including vehicle service age, operational characteristics, and the ventilation system performance.

Firstly, the service life of metro vehicles plays a critical role in determining the intensity of AC emissions (Figure 1a and Table 1). Line 1, the oldest line in the Chengdu metro system, operates trains with the longest operational service duration. Interior materials and structural components on this line have undergone extensive ageing, and previous studies have shown that the VOC emissions from such materials typically peak during early use and decline over time to a relatively stable, lower-emission phase [75]. The comparatively low AC levels observed on Line 1 may thus reflect material ageing and reduced emission potential. In contrast, trains on Line 7 are relatively new, with many interior components, including seat upholstery, rubber seals, and plastic panels, remain in their high-emission phase, substantially increasing pollutant levels. Secondly, the operating environment also plays a significant role in shaping AC levels. Line 7 functions as a circle line with high daily service frequency and a dense transfer network, leading to more frequent door openings and intensified passenger turnover. This dynamic operational pattern increases the frequency of air exchange between carriages and station platforms, facilitating the ingress of external pollutants and elevating in-carriage AC levels [48]. Additionally, the high operating intensity may increase thermal loads within carriages, potentially accelerating the volatilisation of interior materials. More importantly, elevated passenger density on Line 7 contributes additional emissions from human-related sources, including metabolic by-products, residual chemicals on clothing, and exhaled VOCs. In contrast, Line 1, as a trunk route, exhibits more stable passenger flow and less intensive disturbance, resulting in lower levels of human-sourced emissions. Finally, differences in ventilation system configurations significantly influence in-carriage pollutant concentrations. Although Line 7 is equipped with a high-efficiency HVAC system, it is typically fitted with HEPA filters designed primarily for particulate removal [76] and lacks dedicated modules for gaseous pollutant adsorption. This may result in a “sealed accumulation effect”, in which enhanced airtightness combined with insufficient air exchange, reduces dilution capacity and allows pollutants to build up. In contrast, the older carriages on Line 1, with relatively poor airtightness, may benefit from a degree of “passive ventilation”, facilitating pollutant dispersion and natural air exchange. This passive dilution effect may partially offset the emissions from ageing materials. In summary, the inter-line differences in AC concentrations are not governed by a single determinant but instead result from the combined effects of vehicle ageing, operational burden, and ventilation–dilution capacity.

4.2.2. Different Periods

As illustrated in Figure 1, not all metro lines exhibited a consistent trend of higher AC concentrations during peak periods (Figure 2). Line 4 and Line 7 demonstrated significantly higher concentrations during peak periods, reaching 45.03 ± 0.89 µg/m³ and 47.89 ± 1.38 µg/m³, respectively (Figure 2d,f), compared to 35.70 ± 0.62 µg/m³ and 45.13 ± 0.75 µg/m³ during off-peak periods (p < 0.05) (Figure 2c,e). This pattern aligns with conventional expectations, whereby increased passenger density and intensified human-related emissions facilitate the accumulation of ACs within confined carriage environments. However, Line 1 displayed an anomalous inverse trend. TAC concentration during off-peak periods (37.68 ± 0.47 µg/m³) (Figure 2a) was slightly higher than during peak periods (36.98 ± 0.82 µg/m³) (Figure 2b), with a difference of 1.89 µg/m³. Although the disparity was modest, it clearly deviated from the patterns observed on the other two lines. Line 1 connects major transport hubs (North Railway Station and South Railway Station), where passenger flow does not strictly follow conventional peak and off-peak patterns. Consequently, AC concentrations on this line remained relatively stable across time periods.

These temporal differences are unlikely to be governed by a single mechanism (Figure 2). Rather, they are more plausibly driven by the interaction between passenger density and the emission characteristics of specific compounds. Notably, the impact of human occupancy appeared to be species-selective, with significant responses observed only among compounds closely associated with human activity. In this study, phenol, ethoxybenzene, benzaldehyde, acetophenone, and benzoic acid all exhibited marked increases during peak periods. Phenol may originate from the oxidative metabolism of human sweat and sebum [77], or from the thermal degradation of chlorine-containing laundry disinfectants [78]. Ethoxybenzene, commonly used as a fixative in perfumes [79], becomes more volatile with elevated body temperature. Benzaldehyde can form via the metabolic oxidation of benzyl alcohol or direct exposure to air and is also a constituent of perfumes and skincare products [80], representing a hybrid emission route involving both exhalation and surface volatilisation. Acetophenone used as a fragrance and fixative in sunscreens and hairsprays [81,82], and benzoic acid, potentially released from the decomposition of hippuric acid in human sweat [83], also demonstrated peak-hour enhancements. The emission fluxes of these compounds are significantly amplified under high passenger density conditions, leading to elevated airborne concentrations during peak periods. These findings highlight the importance of considering compound-specific emission sources and their coupling with passenger activity in interpreting temporal concentration variations within metro environments.

4.3. Source Analysis of ACs

To identify the sources of ACs, this study first applied diagnostic ratio analysis using three commonly used indicators: the T/B, X/E, and S/B (Table 6). Across the three metro lines, T/B ratios generally ranged from 1.57 to 3.52, with mean values exceeding 2, and in some cases surpassing 3.5. Such elevated ratios cannot be explained solely by infiltration of traffic-related emissions, suggesting additional internal sources of toluene, likely including emissions from interior paints, adhesives, and synthetic materials. These results preliminarily indicate that ACs in metro carriages predominantly originate from a combination of external vehicular exhaust infiltration and solvent-related emissions from internal materials. The X/E ratio serves as an indicator of the photochemical age of air masses. In this study, X/E values typically ranged from 3.57 to 5.18 within the carriages, suggesting that the detected xylene and ethylbenzene were relatively fresh and had undergone minimal atmospheric ageing. Given the short time between emission and exposure, along with continuous HVAC operation that recirculates emissions throughout the carriage, these findings support the conclusion that xylene and ethylbenzene are primarily derived from sustained internal emissions. The S/B ratio was used to assess the influence of polymer-based materials within the carriage environment. In this study, S/B values generally ranged from 0.89 to 2.87. Styrene is a well-established marker of plastic and resin emissions, typically released from rubber seals, foam seating, or insulation materials [84,85]. The presence of styrene, and the lack of significant concentration differences between aboveground and underground operation, suggests that its source is predominantly internal. A similar result was reported in the Shanghai metro, where styrene concentrations remained stable regardless of external conditions [20]. Therefore, elevated S/B values in the present study reinforce the contribution of in-car material emissions. In summary, while vehicular emissions remain the dominant source of ACs, emissions from interior carriage materials are also substantial and cannot be overlooked. The relative contributions vary by line: Line 1 appears more strongly influenced by external traffic sources, while Line 7 exhibits a clearer signature of internal emissions. Variations between peak and off-peak periods did not fundamentally alter the source profiles, suggesting a relatively stable composite source structure. These preliminary source apportionment results highlight the need for dual control strategies targeting both traffic pollutant infiltration and emissions from metro carriage materials.

A quantitative source apportionment of ACs in metro carriages during summer was conducted using the PMF model, which resolved five major sources (Figure 3). Among them, emissions from carriage materials were the dominant contributor, accounting for 36.62% of the total ACs. This highlights the importance of “secondary off-gassing” processes in enclosed transit environments. The key indicator was the extremely high loading of styrene (74.89%), a monomer of polystyrene and acrylonitrile–butadiene–styrene (ABS) plastics, which are widely used in metro seats, wall panels, and lighting components. Styrene emissions are known to increase under high temperature and humidity conditions, underscoring a structural deficiency in VOC control embedded in current metro design practices. The second-largest contributor was identified as human emission sources (22.50%), characterised by elevated levels of benzaldehyde and acetophenone. These compounds are known derivatives of human sebum oxidation and are commonly found in personal care products, as well as evaporative components of perspiration. This source highlights the considerable impact of high passenger density on in-car air quality, particularly during peak commuting periods. Although traffic exhaust infiltration (16.67%) and industrial emission sources (7.66%) were comparatively less dominant, their representative species, such as BTEX and polymethylated benzenes, warrant toxicological attention. These pollutants suggest a potential risk of ambient contamination entering the cabin via the ventilation system. BTEX compounds, typically produced by the incomplete combustion of petrol and vehicle exhaust, are recognised markers of outdoor infiltration. Given the design of metro HVAC systems, which may draw air from ground-level inlets, these findings suggest a mechanism of “non-contact penetration”, allowing ambient traffic pollution to enter the underground environment without direct exposure. Notably, the identification of an organic solvent source (16.55%) points to atypical emission events. Benzonitrile, a common plastic additive and chemical intermediate used in the production of synthetic fibres, flame retardants, and coating materials, alongside p-xylene and m-xylene, both widely employed as solvents and diluents, suggest inputs beyond routine operation. These compounds may originate from maintenance coatings or refurbishment activities carried out during metro construction and service. This finding underscores the importance of implementing pollution control strategies that extend beyond daily operations to include the entire lifecycle of metro systems, including construction, renovation, and maintenance phases.

In this study, the PMF model played a central role in elucidating the underlying sources of ACs within metro carriages. Its ability to apportion pollution contributions without requiring prior source profiles makes it particularly well-suited to non-targeted screening datasets. The five-factor solution revealed by PMF not only corroborated the diagnostic ratio analysis but also enabled a more quantitative understanding of the relative influence of human activity, interior materials, and external infiltration.

These findings have practical implications for pollution control and system design. The source structure revealed by this study reflects a composite pollution pattern characterised by a triad of material off-gassing, human emissions, and external infiltration. This underscores that air quality control in metro carriages must extend beyond the optimisation of material selection and ventilation systems, and incorporate behavioural interventions, green maintenance protocols, and comprehensive lifecycle management strategies. Future mitigation efforts should prioritise the substitution of low-emitting interior materials, the identification of individual-level pollutant sources, and the implementation of zone-based ventilation control, to enable a systemic and sustainable reduction of pollutants in enclosed transit environments. Overall, PMF strengthened the interpretability of complex chemical profiles and enhanced the translational value of the risk assessment framework.

4.4. Health Risk Assessment

This study employed a quantitative risk assessment model to systematically evaluate the health effects of ACs exposure among different population groups in Chengdu’s metro carriages. The results indicate that, under current exposure conditions, both the HI and TCR values for passengers remain within acceptable or negligible thresholds. However, a more detailed analysis reveals notable inter-group differences in risk levels, and a high concentration of overall health risk attributable to a small number of high-toxicity compounds. These findings emphasise the need for targeted intervention strategies in future air quality management.

Tourists were consistently found to face higher health risks than commuters (Table 7). Specifically, the HI values for adult and child tourists were 2.73 × 10⁻² and 1.89 × 10⁻², respectively, while their corresponding TCR values were 5.83 × 10⁻⁷ and 4.72 × 10⁻⁷, approximately 1.89 and 2.51 times those of the commuting population. This disparity is likely driven by two factors: (1) increased physical activity and higher respiratory rates during travel; and (2) longer and more complex travel routes, resulting in prolonged exposure in enclosed environments. Although overall risk levels remain below critical thresholds, these findings suggest that highly sensitive subpopulations, particularly child tourists, should be prioritised in future risk mitigation efforts.

In terms of toxicological contributions, a small number of AC species accounted for the majority of health risks (Figure 4). Benzene was the predominant contributor to both non-cancer and cancer risks. Across all groups, its HQ ranged from 8.13 × 10⁻³ to 1.53 × 10⁻², accounting for roughly 52% of the total HI. Its TCR ranged from 1.43 × 10⁻⁷ to 4.43 × 10⁻⁷, contributing over 73% of total cancer risk. As a Group 1 human carcinogen, benzene is strongly associated with haematopoietic disorders and increased risk of leukaemia from chronic low-dose exposure. The findings are consistent with previous metro-based risk assessments worldwide. While the risks observed in this study did not exceed acute toxicity thresholds, the potential cumulative effects of long-term exposure in confined environments warrant ongoing attention.

Other notable contributors to HI included 1,2,3-trimethylbenzene, p-xylene, acetophenone, and toluene, which together accounted for over 90% of the total HI (Figure 4a–d). While 1,2,3-trimethylbenzene is not currently classified as a regulated pollutant, its consistent HQ contribution (11.75–11.78%) across all groups indicates its environmental persistence and potential toxicological relevance. This suggests the need to re-evaluate the chronic impacts of such under-regulated pollutants in metro environments. Of note, acetophenone, widely used as a fixative in personal care products, may originate from both passenger emissions and material off-gassing. Its potential neurotoxic and metabolic effects under high background concentrations call for further toxicological investigation.

For cancer risk, two additional compounds (Figure 4e–h), 1,4-dichlorobenzene and ethylbenzene, were assessed. Although their individual CR values remained below 10⁻⁷, well within the acceptable threshold of 1 × 10⁻⁶ [86], they contributed approximately 13% and 10% to the overall TCR, respectively. The presence of 1,4-dichlorobenzene may reflect non-traffic-related sources, such as cleaning agents or fragranced products. Ethylbenzene, classified as a possible human carcinogen, also warrants attention. Overall, the cancer risk profile exhibited a “core compound plus low-frequency high-toxicity species” dual-layer pattern. This pattern highlights the need to focus on primary carcinogens while expanding monitoring to include low-frequency, high-toxicity pollutants.

In conclusion, no single compound exceeded health-based critical thresholds. From a public health perspective, the concentrations of ACs currently found in Chengdu’s metro carriages are unlikely to cause adverse health effects. This may reflect the benefits of improved system design and enhanced source control strategies implemented in place. Nevertheless, the findings reveal a clear risk structure shaped by a few dominant toxicants, especially benzene, and identify vulnerable populations such as children as key targets for protective measures. It is recommended that long-term monitoring and dynamic regulatory mechanisms be established to ensure the continued protection of passenger health and the sustainable management of air quality in underground transit environments.

4.5. Limitations

Despite its contributions, this study has several limitations that should be addressed in future research. First, the investigation was conducted exclusively during the summer, without coverage of other climatically distinct periods. This limits the ability to fully characterise year-round variations in AC pollution. Temperature and humidity substantially affect the volatilisation behaviour and emission rates of interior materials, which may alter not only the pollutant concentration but also compositional profiles and source dominance. Second, the study was restricted to Chengdu, a representative city in southwest China with unique metro system characteristics. However, differences in climate, passenger load, carriage design, and operational strategies across cities may constrain the generalisability of the findings. To enhance the universality and comparative value of results, future studies should incorporate multi-site investigations across diverse climatic regions and urban contexts. Lastly, although this study identified pollution levels and source contributions, it did not incorporate dynamic HVAC operation parameters or in-cabin temperature and humidity data. These environmental control variables are known to exert substantial influence on pollutant dispersion, dilution, and accumulation processes. Incorporating real-time monitoring of ventilation performance and cabin microclimate data in future research would help clarify the interaction between AC concentration dynamics and environmental control. Such insights will facilitate the scientific optimisation of air quality management in enclosed transit environments.

5. Control Strategies for ACs in Metro Carriages

To effectively mitigate AC-related exposure risks in metro carriages, particularly those posed by compounds such as benzene and acetophenone, a multi-level control strategy must address both emission sources and in-cabin dispersion mechanisms. This section proposes a set of actionable interventions tailored for metro system operators and infrastructure planners, with emphasis on technical feasibility and health protection.

5.1. Source Control During Train Manufacture and Maintenance

Minimising emissions at the source is essential. Materials and coatings used in train manufacturing or refurbishment should be selected based on their thermal stability and low VOC emission potential. Waterborne or high-solids-content paints provide safer alternatives to conventional solvent-based coatings, particularly under high-temperature conditions that promote off-gassing. Additionally, chemical products used in maintenance, such as cleaning agents, air fresheners, and surface treatments, should be screened for VOC emissions. High-risk substances, such as 1,4-dichlorobenzene, commonly found in deodorisers, should be strictly limited. Metro authorities are encouraged to establish an internal catalogue of certified low-emission materials, drawing on international standards such as ISO 16000-9 [87] or Leadership in Energy and Environmental Design (LEED) indoor air quality criteria.

5.2. Ventilation System Enhancement

Upgrading the carriage ventilation system is critical for disrupting pollutant exposure pathways. High-efficiency adsorptive modules, such as activated carbon filters, can be incorporated into HVAC systems to capture aromatic VOCs through physical adsorption. This is particularly valuable for older rolling stock, where degraded sealing may permit greater infiltration of external pollutants. Improvements in cabin airtightness should also be prioritised to minimise bypass airflow. Furthermore, intelligent ventilation systems guided by real-time VOC monitoring, or benzene-specific sensors, should be deployed to dynamically adjust fresh air intake based on occupancy, internal pollutant loads, or external air quality conditions.

5.3. Operational and Maintenance Protocols

Routine cleaning activities represent a notable source of AC emissions, especially when fragranced or solvent-rich agents are used. Cleaning protocols should favour low-emission products and be scheduled during non-operational hours, such as overnight service suspensions, to allow sufficient ventilation time for pollutant clearance. Procurement decisions can be guided by certification schemes such as the EU Ecolabel or GREENGUARD. Passenger education campaigns, such as discouraging the use of personal fragrance sprays on board, may also help reduce episodic spikes in VOC concentrations.

5.4. Monitoring, Auditing, and Health Risk Communication

Sustained risk control requires robust air quality monitoring and transparent communication. Metro operators should conduct periodic VOC monitoring campaigns, particularly during seasonal extremes (e.g., summer heatwaves) or high-traffic periods. Centralised digital platforms can support real-time pollution mapping, issue early warning alerts, and track the effectiveness of interventions. These initiatives should be paired with public reporting mechanisms to promote awareness and accountability. Health-oriented performance metrics, such as cumulative exposure thresholds or risk-weighted ventilation indices, can also be incorporated into rolling stock and station design standards by urban planners.

5.5. Integration into Policy and Design

For long-term effectiveness, the above strategies must be embedded within the policy and design frameworks governing metro development. This includes updating technical specifications for rolling stock procurement, incorporating indoor air quality (IAQ) benchmarks into carriage design tenders, and aligning urban planning regulations with air pollution resilience, particularly in regions prone to high temperatures that exacerbate VOC emissions. Health-centred planning, through coordination between environmental, transport, and public health departments, is essential for building metro systems that not only ensure efficient mobility but also safeguard public wellbeing.

6. Conclusions

This study developed a comprehensive framework for managing AC risks in metro carriages by NTS, source apportionment, and population-specific health risk assessment. Sixteen ACs were identified across three major metro lines in Chengdu, encompassing both conventional species (e.g., BTEX) and a series of oxygenated ACs that are often overlooked, such as acetophenone, benzonitrile, and benzoic acid. These findings highlight the chemical complexity of in-carriage air pollution.

Concentration analyses revealed significant inter-line and temporal variations in ACs exposure. Line 7 exhibited the highest AC levels due to new material emissions and high passenger loads. Line 4 experienced pollutant accumulation during peak periods. By contrast, Line 1 displayed the lowest concentrations, likely due to material ageing and passive ventilation. PMF analysis revealed five primary contributors: emissions from carriage materials release (36.62%), human emissions (22.50%), traffic exhaust penetration (16.67%), organic solvent sources (16.55%), and industrial emissions (7.66%). These findings support a tripartite mechanism involving material off-gassing, passenger activity, and external intrusion. Health risk assessment indicated that all HI and TCR values remained below health-based thresholds, and risk disparities emerged among population groups. Tourists, especially children, exhibited higher exposure. Benzene emerged as the dominant risk contributor, accounting for over 70% of TCR values. However, non-priority compounds such as 1,2,3-trimethylbenzene and acetophenone showed significant non-cancer risk contributions. These substances likely originate from adhesives, synthetic polymers, and cleaning residues, with emissions intensified under high temperatures and enclosed conditions. Their frequent presence and toxicological potential suggest a need to revise conventional VOC priority lists, which are typically limited to BTEX species.

Based on these insights, several targeted control strategies are proposed, each linked to specific emission mechanisms. First, reducing emissions from carriage materials requires the adoption of low-emission, thermally stable components, especially in newly deployed trains. Second, elevated VOCs during peak hours call for demand-responsive ventilation systems that adjust airflow based on real-time occupancy. Third, the use of green-certified cleaning products and optimised maintenance schedules can reduce solvent-related emissions. These strategies function through two main mechanisms: suppressing emissions at the source and disrupting pollutant transport pathways. Compared with conventional BTEX-focused approaches, the proposed multidimensional strategy addresses a wider array of overlooked yet health-relevant VOCs identified through non-targeted analysis. In summary, this work provides a mechanistic and compound-specific perspective to guide future metro air quality management and supports a more health-oriented and sustainable transit environment.