Spatiotemporal Characterization of Atmospheric Emissions from Heavy-Duty Diesel Trucks on Port-Connected Expressways in Shanghai

Abstract

Heavy-duty diesel trucks (HDDTs) are recognized as significant sources of air pollutants and greenhouse gases (GHGs) along freight corridors in port cities. Despite their impact, few studies have provided detailed spatiotemporal insights into their emissions within port-adjacent highway systems. This study presents a high-resolution, hourly emission inventory at the road-segment level for six major expressways in Shanghai, one of China’s leading port cities. The emission estimates are derived using a locally adapted COPERT V model, calibrated with HDDT GPS trajectory data and detailed road network information from OpenStreetMap. The inventory quantifies emissions of CO₂, NO_x, CO, PM, and VOCs, highlighting distinct temporal and spatial variation patterns. Weekday emissions consistently exceed those of weekends, with three prominent traffic-related peaks occurring between 5:00–7:00, 10:00–12:00, and 14:00–16:00. Spatial analysis identifies the G1503 and S20 expressways as major emission corridors, with S20 exhibiting particularly high emission intensity relative to its length. Combined spatiotemporal patterns reveal that weekday emission hotspots are more concentrated, reflecting typical freight activity cycles such as morning dispatch and afternoon return. The findings provide a scientific basis for formulating more precise emission control measures targeting HDDT operations in urban port environments.

1. Introduction

Amid escalating global climate change and worsening environmental pollution, the reduction in air pollutants and greenhouse gas (GHG) emissions has become a pressing global imperative. In recent years, China has strengthened its commitment to ecological and environmental protection and actively pursued a green, low-carbon transition. In 2022, the Ministry of Ecology and Environment, in collaboration with multiple government agencies, released the Implementation Plan for the Synergistic Reduction of Pollution and Carbon Emissions. This plan promotes the integrated management of air pollutants and GHGs as a unified strategy to advance green transformation across economic and social sectors. The policy framework provides critical institutional support for improving environmental quality and achieving national carbon peaking targets [1]. Freight transport constitutes a vital pillar of the transportation system, playing an essential role in enabling manufacturing, distribution, and consumption. Heavy-duty diesel trucks (HDDTs), as the backbone of many freight operations, are not only indispensable to logistics but also represent a major source of CO₂ and key atmospheric pollutants within the road transport sector [2,3].

In China, road freight accounts for approximately 75% of the nation’s total freight volume. As the primary carriers within this system, HDDTs play an important role in emission reduction efforts and serve as a strategic leverage point for achieving carbon peaking and neutrality targets in the transportation sector [4]. According to previous studies, HDDTs contribute significantly to certain pollutants relative to other vehicle types because of the high calorific value of diesel fuel used by heavy-duty engines [5]. The Chinese government has steadily tightened emission standards for diesel vehicles, advancing from China I to the current China VI standard [6], while simultaneously improving diesel fuel quality to reduce pollutant formation at the source [7]. Shanghai, one of the world’s leading container port cities, maintains a comprehensive and well-developed road network to support freight collection and distribution. The city experiences high volumes of large truck traffic and intense vehicle operation, both of which significantly impact local air quality. In response, the Shanghai Clean Air Action Plan (2023–2025) mandates the implementation of the China VI b emission standard beginning in July 2023, calls for the complete phase-out of China III-compliant diesel trucks by the end of 2025, and explores accelerated retirement of China IV vehicles [8]. Furthermore, Shanghai’s 14th Five-Year Plan for Coordinated Control of Atmospheric Pollutants outlines a suite of synergistic emission reduction strategies, including the elimination of outdated diesel trucks, the replacement of port container trucks with Liquefied Natural Gas (LNG)-powered alternatives, and the electrification of non-road mobile machinery. These comprehensive measures provide a robust technical foundation for future policy development [9]. However, the effective design and evaluation of such strategies necessitate a thorough understanding of the spatiotemporal distribution of pollutant emissions. Currently, emission inventories used for policy development are constrained by limited spatial and temporal resolution, as well as by moderate levels of accuracy [10]. In light of these limitations, the development of a high-resolution spatiotemporal emission inventory for HDDTs is essential not only for advancing emission mitigation in the freight transport sector but also for supplying critical input data for regional air quality modeling and evidence-based policy refinement.

Vehicle emission inventories are generally developed using two primary methodological approaches: top-down and bottom-up [11]. The top-down approach estimates emissions based on macro-level statistical data, typically including vehicle population and vehicle kilometers traveled. This method is well-suited for large-scale assessments at national or regional levels [12,13]. However, in complex urban traffic environments characterized by diverse and dynamic emission sources, the top-down approach may not fully satisfy the precision requirements needed for effective policy formulation and evaluation [14]. In contrast, the bottom-up approach leverages detailed micro-level data—such as traffic volume, fleet composition, road network characteristics, and vehicle operating conditions—to estimate emissions at finer spatial and temporal scales, often on a road segment and hourly basis. This method provides improved spatiotemporal resolution and estimation accuracy [15,16], and has increasingly become the preferred approach for constructing urban-scale emission inventories.

With the advancement of Intelligent Transportation Systems, the capacity to collect large-scale traffic activity data has significantly improved, offering a valuable data basis for bottom-up emission estimation [17]. The widespread adoption of GPS technology has enabled the collection of vehicle trajectory data with high temporal continuity and broad spatial coverage. GPS data are characterized by large sample sizes and multidimensional attributes, capturing key operational parameters such as speed, acceleration, travel distance, and idling status. These parameters are critical for accurately identifying vehicle operating modes and corresponding emission processes [18]. The integration of GPS data with Geographic Information Systems (GIS) and advanced visualization techniques further enhances the analysis of emission patterns and traffic behavior by revealing their spatial distribution [19]. Recent studies have highlighted the potential of GPS data in the development of high-resolution emission inventories. For instance, Deng et al. [20] proposed the TrackATruck framework, leveraging 19 billion GPS trajectories from HDDTs to construct a nationwide emission inventory with high spatiotemporal resolution. Similarly, Sui et al. [21] utilized GPS trajectories and order data from ride-hailing vehicles and taxis in Chengdu to evaluate fuel consumption and emission characteristics across different vehicle types, and further investigated the underlying factors contributing to observed disparities. Si and Lin [22] integrated ride-hailing GPS trajectories and geospatial data to quantify mobility patterns and built-environment indicators in Chengdu, China.

Commonly used bottom-up vehicle emission estimation models include CMEM [23], MOVES [24], EMFAC [25], IVE [26] and COPERT [27]. These models vary in terms of modeling accuracy, data requirements, and applicability to different contexts. However, the implementation of these models in China is often constrained by factors such as limited data availability, model complexity, and insufficient adaptability to local conditions. For instance, although CMEM supports real-time emission modeling, it demands highly detailed input data, including microscopic driving behavior and vehicle-specific technical specifications. The difficulty in localizing these parameters, along with the model’s proprietary nature, limits its broader application in China [28]. Similarly, MOVES require extensive data on vehicle configurations and operational behavior—such as start frequency, idling duration, and canister purge rates—which are difficult to obtain given the current stage of development of China’s current statistical and monitoring systems [28]. EMFAC further increases the data burden by requiring dynamic inputs such as detailed fleet composition, vehicle age distributions, maintenance records, and environmental variables like temperature and humidity [29]. The IVE model requires transient vehicle operating data, such as high-frequency speed, acceleration, and engine load, which involve considerable data collection efforts and technical complexity [30]. In contrast, COPERT (Computer Programme to Calculate Emissions from Road Transport) is recognized for several notable advantages, including relatively low data requirements, simplified computational procedures, and high adaptability. The model estimates emissions by categorizing vehicles based on type, fuel, road characteristics, and other relevant parameters, utilizing regression analysis of vehicle speed and travel distance to quantify emissions of key pollutants (CO, NO_x, HC, PM, etc.) as well as fuel consumption (FC). Originally developed under the framework of European emission standards, COPERT aligns well with China’s current vehicle emission standards, thereby exhibiting strong compatibility. As a result, it has been widely adopted in emission estimation studies across China [31,32]. Cheng et al. [32] integrated HDDT trajectory data, road network characteristics, and traffic conditions to develop a high-resolution spatiotemporal emission inventory for Beijing using the COPERT model. Zhang et al. [33] applied the model to estimate NO_x, PM, and SO₂ emissions across 200 administrative divisions in the Beijing–Tianjin–Hebei region based on HDDT activity data. Additionally, Peng et al. [34] utilized 1.08 billion GPS trajectory records to estimate CO₂ emissions from HDDTs in Xi’an during September 2022. Han et al. [35] analyzed emissions along highway corridors and found that HDDT emissions on major national expressways remain a significant challenge for regional air pollution control, highlighting the importance of port-related and urban arterial corridors in emission inventory studies. Cheng et al. [36] evaluated the health and economic benefits of HDDT emission reductions under policy scenarios, demonstrating the critical significance of controlling heavy truck emissions on expressways and urban arterials. Collectively, these studies highlight COPERT’s compatibility with available data and computational efficiency, indicating its suitability for estimating HDDT emissions at both regional and national scales in China.

While most studies have primarily focused on CO₂, NO_x, and PM, with some extending to CO and VOCs, comprehensive multi-pollutant inventories remain relatively limited. The COPERT model provides validated parameters and localized adjustment functions for these five pollutants, making them particularly suitable for high-resolution inventory development. Moreover, these pollutants are closely linked to environmental policy: CO₂ is the dominant greenhouse gas, NO_x and PM are major contributors to urban air pollution, while CO and VOCs are precursors to ozone formation with significant health impacts. Therefore, this study considers CO₂, NO_x, CO, PM, and VOCs to construct a multi-pollutant emission inventory and conduct a spatiotemporal analysis of their emission characteristics. At the same time, despite the introduction of several regulatory measures in Shanghai and the Yangtze River Delta—such as China VI standards and green port initiatives—practical gaps remain in their implementation. For instance, vehicle fleet renewal lags behind policy requirements, enforcement and monitoring capacities are limited, and highway corridors and port-access roads have not been adequately addressed in existing emission control programs. As a result, reductions in NO_x and PM emissions from HDDTs often fall short of policy targets [37,38]. These gaps highlight the necessity of refined, data-driven analyses of HDDT emissions, particularly along critical freight corridors.

Building on this background, the present study focuses on Shanghai and aims to develop a high-resolution emission inventory for HDDTs in the freight transport system. A data-driven estimation framework is established by integrating GPS trajectory data, detailed geospatial information, and vehicle operation status identification. Using the COPERT V model with localized fuel parameters, segment-level emissions are estimated for six major freight expressways, including the Shanghai Ring Expressway (G1503) and the Outer Ring Expressway (S20), covering multiple pollutants (CO₂, NO_x, CO, PM, and VOCs). Unlike previous inventories aggregated at city or regional scales, our approach enables emission analysis at hourly and road-segment resolutions, capturing peak-hour patterns, traffic hotspots, and port-related operational influences that are not discernible in coarser datasets. The locally adapted COPERT V model ensures that emission factors reflect actual vehicle and operational conditions in Shanghai, thereby enhancing both the accuracy and representativeness of the results. Collectively, these methodological innovations provide a more detailed and actionable understanding of HDDT emissions in port-city contexts, offering valuable insights for targeted emission control and urban air quality management strategies.

2. Method

This study utilizes GPS trajectory data from HDDTs operating in Shanghai between 1 May and 31 May 2023, to construct a high-resolution emission inventory with both hourly and road-segment-level granularity. The inventory enables a systematic examination of the spatiotemporal patterns of major pollutant emissions from HDDTs. The research framework is organized as follows: First, the raw GPS trajectory data are preprocessed to ensure data quality and consistency. A customized, data-driven emission estimation system is then developed. Next, baseline emission factors are calculated using the COPERT V model, with localization adjustments applied to key parameters to reflect the actual fuel composition and regional vehicle characteristics. The processed trajectory data are subsequently integrated with traffic volume statistics and road segment length information to estimate hourly and segment-level emissions of multiple pollutants, resulting in the construction of a high-resolution HDDT emission inventory. Finally, the study conducts an in-depth spatiotemporal analysis of HDDT emissions using six major freight expressways in Shanghai as case studies, focusing on emission patterns during the study period in May 2023. The overall research framework is depicted in Figure 1.

2.1. Study Area

Shanghai, a major port city in China, has developed a comprehensive and well-integrated freight collection and distribution system, in which road transport plays a critical role within the broader multimodal transportation network. In 2023, Shanghai’s total freight volume reached approximately 1.533 billion tons, with road transport contributing around 504 million tons—accounting for nearly 33% of the total [39]. HDDTs, as the primary mode of road-based freight transport, are responsible for a substantial share of medium- and long-distance logistics operations. Due to their high operational frequency, HDDTs are also a major contributor to local air pollutant emissions [40]. This study selects Shanghai as the research area, given its representative and high-intensity expressway network for port-related collection and distribution. The analysis focuses on the emission characteristics of major freight corridors. Based on field investigations and a comprehensive assessment of regional traffic volume patterns and port connectivity, six key freight expressways were identified for in-depth analysis: the Shanghai Ring Expressway (G1503), Outer Ring Expressway (S20), Shenjiahu Expressway (S32), Shenyang–Haikou Expressway (G15), Hulu Expressway (S2), and Huxiang Expressway (S6) (Figure 2). These expressways connect major ports and logistics hubs and carry high volumes of HDDT traffic, making them critical segments for evaluating freight-related emissions within port city transportation systems. The selection of these six corridors is also supported by Jin and Liu [41], who identified them as the main freight roads in Shanghai based on GPS-based traffic and emission analyses, further validating their relevance for high-resolution emission studies.

2.2. Data

2.2.1. Data Description

This study integrates GPS-based vehicle trajectory data with road attribute information to develop a high-resolution spatiotemporal emission inventory for HDDTs operating along Shanghai’s primary freight expressways, with a further focus on analyzing their spatiotemporal emission characteristics. The GPS trajectory dataset captures the movement patterns of all HDDTs within Shanghai between 1 May and 31 May 2023, with sampling intervals ranging from 10 to 120 s. Key recorded variables include vehicle ID, timestamp, geographic coordinates (latitude and longitude), and speed, as detailed in

Table 1. Sample GPS Trajectory Data of HDDTs in Shanghai.

Date	Time	Truck ID	Longitude	Latitude	State	Direction	Speed
30 April 2023	23:59:37	4115511560	121.28732	31.35992	1	2	0.0
30 April 2023	23:59:44	4026134699	121.85678	30.90282	1	6	53.6
30 April 2023	23:59:44	4026984598	121.52759	31.37525	1	2	23.0
30 April 2023	23:59:43	6470922003	121.76619	31.16807	1	1	0.0

. The raw GPS trajectory dataset comprises approximately 320 million records, covering 16 expressways and around 1000 km of key urban freight corridors in Shanghai, including both expressways and primary arterial roads. To focus specifically on port-related freight transport, the selected six major freight expressways—G1503, S20, S32, G15, S2, and S6—which are well represented in the dataset and serve as the primary logistics routes for Shanghai’s port freight system. Consequently, an initial filtering and spatial matching process was required to isolate relevant data. Using Python (v3.8.20) scripting within the ArcGIS (v10.8) environment and leveraging the arcpy library, large-scale trajectory data processing was performed. Spatial matching was achieved by generating buffer zones and overlaying GPS points onto vector representations of the road network, enabling the identification of trajectory points located within the specified expressway corridors. This process resulted in approximately 43.6 million valid GPS trajectory records for the six expressways, forming the foundational dataset for the subsequent development and analysis of the emission inventory.

The primary vehicle type captured in the dataset is China National V emission standard heavy-duty diesel freight trucks, with payload capacities typically ranging from 12 to 40 tons. These vehicles predominantly operate on the aforementioned freight corridors. Typical truck models include mainstream logistics fleet vehicles such as the FAW Jiefang J6 series, Sinotruk HOWO series, and Shaanxi Auto Delong series, which are widely used in regional and port-related freight transport. The associated diesel engine exhaust after-treatment technologies include SCR (Selective Catalytic Reduction), EGR (Exhaust Gas Recirculation), and the combined DPF+SCR system (Diesel Particulate Filter plus SCR), with SCR being the most prevalent configuration.

In this study, road network data were primarily obtained from OpenStreetMap, encompassing features such as road centerlines, node coordinates, road classifications, names, types, widths, and directional attributes. These data elements provide crucial support for subsequent map-matching procedures and for accurately estimating segment-level emissions. OSM was adopted as the underlying road network not only for its open accessibility and compatibility with large-scale GPS trajectory analysis but also because it has been widely applied and validated in emission studies and transportation research [31,32,41]. Although local inaccuracies or incomplete segments may exist in OSM, the selected expressways and major freight corridors in Shanghai are well represented, thereby limiting the potential impact of such uncertainties on our results.

2.2.2. Data Processing

During the acquisition of GPS trajectory data, various anomalies—such as invalid fields, duplicate entries, abnormal speed values, and spatial drift—may arise due to device malfunctions, signal loss, or environmental interference. Among these, spatial drift is defined as a condition in which the Euclidean distance between two consecutive sampling points exceeds the maximum theoretical travel distance of a vehicle within the given sampling interval. To ensure the reliability and analytical precision of the dataset, the raw GPS trajectory data underwent a structured preprocessing procedure prior to subsequent analysis. ArcGIS 10.8 and Python were employed to execute data filtering and cleaning operations. Only variables critical to emission estimations such as vehicle identifiers, timestamps, geographic coordinates, and instantaneous speeds—were retained, while all anomalous and irrelevant records were systematically removed. The preprocessing workflow included the following key steps:

(1) Invalid Field Removal: Non-essential fields unrelated to emissions analysis were excluded to improve data processing efficiency. Only core attributes—vehicle ID, time, location, and speed—were preserved.

(2) Duplicate Entry Elimination: For duplicate entries recorded for the same vehicle at identical timestamps, only the first valid observation was retained to avoid computational bias.

(3) Filtering of Abnormal Speed Values: In accordance with the Implementation Regulations of the Road Traffic Safety Law of the People’s Republic of China [42], freight vehicles are legally restricted to a maximum speed of 100 km/h on expressways. Data points exceeding this threshold were identified as abnormal and removed. It is worth noting that this study focuses primarily on the hot running phase of heavy-duty diesel trucks, which constitutes the dominant source of on-road emissions during vehicle operation. This emphasis aligns with the research objective of characterizing emissions along major freight corridors serving port-related logistics in Shanghai, where trucks are predominantly in motion rather than idling or parked. Furthermore, the COPERT model defines the applicable speed range for hot emission factors of heavy-duty diesel vehicles as 5–85 km/h. To ensure compatibility with the model framework and to better reflect effective heat emission processes during active driving, vehicle speeds recorded below 5 km/h—typically associated with short-term congestion or stop-and-go traffic—were adjusted to 5 km/h. As a result, emissions from stationary or parked vehicles were not considered in the current estimation.

(4) Due to limitations in positioning accuracy or signal obstructions, some trajectory points exhibited unrealistic spatial displacement. The Euclidean distance between successive trajectory points was calculated as defined in Equation (1), and compared against the theoretical maximum travel distance. Points exceeding this threshold were flagged as drift and removed, while the previous, more reliable point was retained.

$$d={\mathit{cos}}^{-1}(\left(\mathit{sin}{y}_t\mathit{sin}{y}_{t-1}\right)+(\mathit{cos}{y}_t\mathit{cos}{y}_{t-1}\mathit{cos}(x_t-x_{t-1})\left)\right) \ast D$$

(1)

where d denotes the Euclidean distance, D represents the Earth’s radius, $x_t,y_t$ represent the longitude and latitude, respectively.

After data cleaning, the number of valid GPS records decreased from 43.6 million to 38.7 million, representing a loss of about 11.2%. These removed records mainly consisted of duplicated points, speed anomalies, and drift errors. The remaining data still provide high spatial–temporal coverage across the six expressways, ensuring robust representativeness for emission estimation. GPS positioning may suffer from errors due to multipath effects, building obstructions, or device inaccuracies, leading to deviations of several meters in dense urban environments. However, since this study primarily focused on six major expressways in Shanghai, the open-road conditions substantially reduced GPS drift, typically within 3–10 m. Considering the high spatial density of the dataset (approximately 38.7 million valid re-cords after preprocessing), such small-scale deviations are unlikely to significantly bias distance estimation or subsequent emission results. Nevertheless, to further improve spatial accuracy and ensure the validity of emission calculations, all GPS trajectory points were map-matched to the corresponding road segments, consistent with practices widely adopted in previous GPS-based emission studies [14,41].

2.2.3. Map Matching

In this study, the primary objective is to extract instantaneous speed and traffic volume information at the road-segment level for HDDTs, which forms the basis for estimating average segment speeds and emissions—without the need for reconstructing full vehicle trajectories. Given the sparsity of the trajectory data and the need for computational efficiency, a geometry-based map-matching approach was adopted. Specifically, a nearest-neighbor strategy was employed, wherein each GPS point is projected onto the closest road centerline segment using Euclidean distance, and the corresponding road attributes are assigned accordingly. This method offers an effective trade-off between spatial accuracy and processing efficiency, making it particularly well-suited for large-scale trajectory datasets. As shown in Figure 3, the matched results demonstrate high spatial consistency, with most points accurately aligned to the designated road network—thus providing a reliable foundation for emission estimation.

2.3. Method of Emission Calculation

This study adopts a bottom-up approach to estimate emissions from HDDTs operating on major expressways, thereby constructing a high-resolution spatiotemporal emission inventory. As a preliminary step, each expressway was divided into multiple segments of approximately 4 km in length to facilitate localized emission estimation. Based on GPS trajectory analysis, this segmentation was found suitable for calculating the average speed of freight vehicles within each section, ensuring representative emission estimates without sacrificing spatial accuracy. Although finer segmentation could capture more localized variations, it would increase computational demand; thus, the 4 km division provides a practical balance between resolution and efficiency.

The calculation method of emission calculation as follows: First, based on GPS trajectory data, the average travel speed for HDDTs is determined for each road segment. This average speed serves as a proxy for the segment’s operating speed and is used to estimate emissions of key pollutants, including CO₂, NO_x, CO, PM, and VOCs. Second, the segment length and the corresponding hourly traffic volume are combined to estimate total vehicle activity, which is then used to quantify emissions across pollutant categories. Finally, emissions from all individual segments within the six principal freight corridors are aggregated spatially to generate high-resolution emission inventories at both segment and network levels. The detailed emission calculation method is described in Equation (2).

$$E_{p,i,t}=E{F}_{p,i,t}^{Corr}\times T{V}_{t,i}\times L_i\times 10^{-3}$$

(2)

where $E_{p,i,t}$ denotes the emission of pollutant $p$ from all HDDTs on road segment $i$ during time interval $t$, measured in kilograms per hour (kg/h); $E{F}_{p,i,v}^{Corr}$ represents the corrected emission factor for pollutant $p$ on segment $i$ at a speed of $v$, expressed in grams per kilometer (g/km); $T{V}_{i,t}$ represents the traffic volume of HDDTs on road segment $i$ during time interval $t$; and $L_i$ represents the length of road segment $i$, in kilometers (km). The total emissions of pollutant $p$ from HDDTs across all road segments during time interval $t$ denoted as $E_{p,t}$, can be calculated by summing the segment-level emissions:

$$E_{p,t}={\sum }_i{E}_{p,i,t}$$

(3)

In Equation (2), emission factors are obtained by inputting localized parameters into the COPERT V model to reflect actual regional fuel and vehicle characteristics. HDDT traffic volumes and travel speeds are extracted from GPS trajectory data, while road segment attributes, including length and direction, are sourced from shapefiles provided by OpenStreetMap.

2.3.1. Calculation of Emission Factors

Vehicle emission factors are influenced by multiple variables, including emission standards, fuel types, vehicle weight classifications, and driving behaviors [43]. Given the strong alignment between China’s vehicular technology regulations and European standards—as well as the widespread domestic use of the COPERT model for estimating emissions across diverse vehicle categories [31,44]—this study adopts the COPERT V model [45] to estimate the baseline emission factors for HDDTs. Specifically, detailed traffic activity data and vehicle-specific attributes, such as instantaneous speed and traffic volume, were extracted from the processed GPS trajectory datasets. These data were then mapped to the corresponding emission parameters within the model’s database to compute baseline HDDT emission factors, as described in Equation (4):

$$E{F}_{p,i,v}^{Base}={\displaystyle \frac{\left(a\times {v_i}^2+\beta \times v_i+\gamma +\delta /v_i\right)}{\epsilon \times {v_i}^2+\theta \times v_i+\varphi }}\times (1-RF)$$

(4)

where $E{F}_{p,i,v}^{Base}$ represents the baseline emission factor for pollutant $p$ from HDDTs on road segment $i$ at an average speed of $v$, expressed in grams per kilometer (g/km); $v_i$ represents the average speed on segment $i$, in kilometers per hour (km/h). The coefficients $a,\beta ,\gamma ,\delta ,\epsilon ,\theta ,\varphi$, along with the correction factor $RF$, are model parameters derived from the COPERT V emission factor database, with their specific values listed in

Table 2. Key Parameters for Each Pollutant Used in COPERT V Model.

$\mathbf{Pollutant} \mathit{p}$	$\mathit{a}$	$\mathit{\beta }$	$\mathit{\gamma }$	$\mathit{\delta }$	$\mathit{\epsilon }$	$\mathit{\theta }$	$\mathit{\varphi }$	$\mathit{R}\mathit{F}$
CO	−1.06 × 10−3	2.60 × 10−2	5.87 × 100	−2.47 × 100	−1.49 × 10−3	1.30 × 10−1	8.62 × 10−2	0.0%
NOx	−1.06 × 10−2	1.30 × 100	−9.65 × 10−1	2.85 × 100	4.39 × 10−3	1.79 × 10−2	8.13 × 10−3	0.0%
PM	2.15 × 10−2	7.15 × 10−1	−2.23 × 100	5.67 × 100	3.76 × 10−3	−1.09 × 10−2	5.48 × 10−2	99.7%
VOCs	8.95 × 10−4	7.15 × 10−2	5.50 × 10−1	2.73 × 10−1	1.03 × 10−1	4.23 × 10−1	3.30 × 100	0.0%
EC	2.15 × 10−2	7.15 × 10−1	−2.23 × 100	5.67 × 100	3.76 × 10−3	−1.09 × 10−2	5.48 × 10−2	0.0%

. Notably, in the COPERT V framework, the methodology for calculating CO₂ emission factors differs slightly from that used for other pollutants such as NO_x, CO, PM, and VOCs. The specific calculation formula is presented in Equation (5) [45].

$$E{F}_{C{O}_2,i,v}^{Corr}=44.011\times {\displaystyle \frac{E{F}_{EC,v,i}^{Base}}{12.011+1.008r_{H:C}}}\times r_{FC:EC}$$

(5)

where $E{F}_{C{O}_2,i,v}^{Corr}$ denotes the corrected emission factor for CO₂ from HDDTs on road segment $i$ at a speed of $v$, expressed in grams per kilometer (g/km); $E{F}_{EC,i,v}^{Base}$ represents the baseline emission factor for energy consumption under the same conditions; $r_{H:C}$ is the hydrogen-to-carbon (H/C) ratio of the fuel; and $r_{FC:EC}$ is the conversion factor between energy consumption and fuel consumption. The H/C ratio of fuel adopted in this study was based on default values in the COPERT V model. While we do not currently possess laboratory-measured data from Shanghai to confirm the exact local H/C ratio, prior studies suggest that under national diesel standards, variation across regions is not extreme. Therefore, using the COPERT default H/C parameter is considered a practical assumption in this context. Future work will aim to obtain local fuel composition measurements to refine and validate this parameter.

The COPERT V model is calibrated based on fuel standards corresponding to specific calendar years. For instance, the Euro V standard reflects fuel properties prevalent in Europe in 2009. Given the discrepancies between the default fuel parameters in the model and the actual fuel quality used in Shanghai in May 2023, localized adjustments to the emission factors for CO, NO_x, PM, and VOCs were necessary. Since CO₂ emissions are directly proportional to fuel consumption, the influence of fuel quality is inherently accounted for in the emission factor calculation; thus, no additional correction is required. The adjustment formula is presented as follows:

$$E{F}_{p,i,v}^{Corr}={\displaystyle \frac{F_{p,fuel}^{Corr}}{F_{p,base}^{Corr}}}\times E{F}_{p,i,v}$$

(6)

where $E{F}_{p,i,v}^{Corr}$ denotes the corrected emission factor for pollutant $p$ from HDDTs on road segment $i$ at an average speed $v$, expressed in g/km. The terms $F_{p,fuel}^{Corr}$ and $F_{p,base}^{Corr}$ are pollutant-specific correction functions applied to account for differences in diesel fuel specifications. specifically, $F_{p,fuel}^{Corr}$ represents the adjustment factor derived from the actual fuel properties used by HDDTs in China in 2023. $F_{p,base}^{Corr}$ denotes the reference correction factor based on the default diesel fuel specifications assumed in the COPERT V model for European HDDTs. Both correction functions are pollutant-specific and are summarized in

Table 3. Fuel-Based Emission Correction Functions for HDDTs.

Pollutants	Correction Factor Equation
CO	$F_{corr}=2.24407-0.0011\times DEN+0.00007\times PAH-0.00768\times CN-0.00087\times T95$
VOC	$F_{corr}=1.61466-0.00123\times DEN+0.00133\times PAH-0.00181\times CN-0.00068\times T95$
NOx	$F_{corr}=-1.75444+0.00906\times DEN-0.0163\times PAH+0.00493\times CN+0.00266\times T95$
PM	$F_{corr}=\left(0.06959+0.00006\times DEN+0.00065\times PAH-0.00001\times CN\right)\times \left[1-0.0086\times \left(450-S\right)/100\right]$

Note: DEN = Density at 15 °C [kg/m³]; S = Sulphur content in ppm; PAH = Polycyclic aromatics content in %; CN = Cetane number, T95 = Back; end distillation in °C.

. The associated parameters used in these correction functions are provided in

Table A1. Diesel fuel specifications for different fuel standards (in 2023).

Fuel Specification	CN	DEN	T95	PAH	S
Chinese fuel	51	840	355	11	10
European fuel	53	840	320	5	3

, which details the comparison between the baseline diesel fuel properties in China (as of 2023) and those embedded within the COPERT V model.

2.3.2. Estimation of Segment-Level Average Speed

In this study, the arithmetic mean of the instantaneous speeds of all HDDTs traveling on a given road segment was used to represent the average speed for that segment. This average speed was subsequently input into the COPERT V model to estimate pollutant emissions. The corresponding calculation formula is presented below:

$$v_i={\displaystyle \frac{1}{Q}}{\sum }_{k=1}^M{v}_{k,i}$$

(7)

where $v_i$ denotes the average speed of HDDTs on road segment $i$, expressed in km/h; $Q$ represents the total traffic volume of HDDTs on segment $i$ and $v_{k,i}$ indicates the instantaneous speed of vehicle $k$ on segment $i$, also in km/h.

To facilitate accurate speed estimation, each expressway was divided into 4 km segments, taking into account the maximum GPS sampling interval of 120 s and the speed limit of 100 km/h. This segmentation ensures that each vehicle contributes at least two trajectory points per segment, thereby enabling the reliable calculation of average speed using Equation (8). The corresponding formula is presented below:

$$v_{k,i}={\displaystyle \frac{D_{k,i}}{t_{l,k,i}-t_{f,k,i}}}$$

(8)

where $v_{k,i}$ denotes the average speed of vehicle $k$ on road segment $i$, expressed in kilometers per hour (km/h); $D_{k,i}$ represents the distance traveled by vehicle $k$ on segment $i$, in kilometers (km); $t_{f,k,i}$ and $t_{l,k,i}$ represent the timestamp of the first and last trajectory points of vehicle $k$ on segment $i$, respectively, both measured in seconds.

3. Results

3.1. Results of HDDT Emission Status

In the parameterization of the COPERT model, this study selects the heavy-duty diesel truck (HDDT) category corresponding to the China National V (CN V) emission standard, with a gross vehicle weight (GVW) range of 28–34 tons and equipped with SCR (Selective Catalytic Reduction) technology. This choice is grounded in both empirical data and regulatory context. First, according to fleet structure investigations and literature [41], the operational status of HDDTs in Shanghai is characterized by a loading pattern in which 40% of vehicles are running empty (approx. 12 tons) and 60% are fully loaded (approx. 40 tons), resulting in a weighted average operational tonnage of 28.8 tons. This aligns closely with the selected GVW range in the COPERT model. Second, in line with Shanghai’s regulatory framework, which mandated the adoption of CN V standards for HDDTs by 2019, it is reasonable to assume that the majority of trucks operating on the studied expressways comply with CN V requirements. This assumption is consistent with the findings of previous studies reporting that CN V has become the dominant emission standard in major Chinese port cities [41]. Moreover, vehicles within the 28–34-ton GVW range are widely used for regional and urban freight transport, and are among the most frequent contributors to emissions in high-traffic corridors. Finally, COPERT requires vehicle classification to be aligned with the prevailing fleet characteristics to ensure emission factors are accurately represented. The integration of loading ratio, technological configuration (SCR), and regulatory fleet standards ensures that the chosen parameters are both technically consistent and representative of actual conditions. This enhances the reliability of the emission estimation outcomes in the context of Shanghai’s freight transportation sector.

Based on the aforementioned parameter settings, including the selected vehicle type and emission standard, the baseline configuration in COPERT was established with a road gradient of 0% and a load ratio of 50%, approximating typical operating conditions. Under this setup, roadway characteristics and traffic dynamics were extracted from GPS trajectory data to reflect real-world driving behavior. Subsequently, emissions of key pollutants were estimated at both hourly and daily temporal resolutions, providing a detailed understanding of the spatiotemporal distribution of emissions along major freight corridors.

Table 4. Daily Emissions of HDDTs on Major Freight Expressways in Shanghai, May 2023.

Date	Pollutants Emission
	CO2 (ton)	CO (kg)	NOx (kg)	PM (kg)	VOCs (kg)
0501	199.31	412.77	759.22	7.85	5.61
0502	230.90	474.61	835.14	9.09	6.43
0503	247.10	507.57	885.65	9.73	6.87
0504	501.44	1034.04	1864.42	19.74	14.02
0505	592.68	1225.89	2264.69	23.33	16.66
0506	564.46	1166.12	2136.23	22.22	15.84
0507	378.23	781.28	1428.31	14.89	10.60
0508	579.48	1202.34	2281.14	22.81	16.39
0509	612.62	1272.89	2440.82	24.12	17.37
0510	629.26	1306.51	2490.38	24.77	17.82
0511	625.02	1295.94	2441.16	24.61	17.66
0512	595.34	1234.76	2333.69	23.44	16.82
0513	400.16	826.76	1513.74	15.75	11.23
0514	293.52	604.45	1079.30	11.56	8.19
0515	554.65	1149.74	2164.14	21.84	15.65
0516	635.52	1319.03	2506.22	25.02	17.97
0517	632.10	1319.88	2622.31	24.88	18.05
0518	658.49	1367.26	2605.10	25.92	18.65
0519	645.02	1339.53	2558.71	25.39	18.27
0520	423.25	876.50	1616.74	16.66	11.92
0521	318.22	654.60	1157.43	12.53	8.86
0522	549.55	1141.55	2180.96	21.64	15.56
0523	613.77	1275.41	2448.77	24.16	17.41
0524	638.29	1324.93	2522.55	25.13	18.07
0525	633.41	1314.25	2491.91	24.94	17.91
0526	603.70	1257.81	2455.45	23.77	17.19
0527	405.74	838.92	1544.62	15.97	11.39
0528	293.01	602.47	1057.81	11.54	8.15
0529	529.68	1097.23	2051.67	20.85	14.93
0530	604.72	1251.03	2317.34	23.81	17.01
0531	584.17	1210.08	2262.00	23.00	16.48
Sum	15,772.80	32,686.16	61,317.61	620.96	444.99

presents the total daily emissions of five major pollutants from HDDTs operating along six primary freight expressways in Shanghai during May 2023. Carbon dioxide (CO₂) emissions totaled 15,772.80 tons, while emissions of carbon monoxide (CO), nitrogen oxides (NO_x), particulate matter (PM), and volatile organic compounds (VOCs) were 32,686.16 kg, 61,317.61 kg, 620.96 kg, and 444.99 kg, respectively. Among these, CO₂ constituted the dominant share of total emissions, serving as the primary greenhouse gas. Notably, NO_x and CO emissions were significantly higher than those of PM and VOCs, underscoring their roles as the principal air pollutants associated with HDDT activity.

As mentioned above, the COPERT V model was localized for Shanghai by adjusting fuel characteristics, and typical operating speeds based on local statistics and literature. While direct emission measurements for validation were unavailable, the calculated emissions were checked against the trends reported in carbon monitor website [46]. According to the statistics in this website, the average CO₂ emissions from ground transportation in Shanghai during May 2023 were approximately 84,500 tons per day, while the estimate derived from this study for HDDTs was 509 tons per day. Although the scope differs, the magnitude is consistent with the expected share of freight-related emissions.

To further verify the temporal consistency between the two datasets, a comparative analysis was conducted, as illustrated in Figure 4, which presents the daily CO₂ emissions from Shanghai’s ground transport (Carbon Monitor) and from six expressways (this study). The two curves exhibit a generally consistent variation pattern over the analysis period, both showing coincident peaks and troughs, suggesting that the model-based estimates effectively capture the real-world dynamics of transportation-related emissions despite the difference in absolute magnitude. This indirect comparison provides additional evidence supporting the robustness and reliability of the emission estimation results.

In addition, daily PM2.5 and PM10 sub-index values for May 2023 were collected from the Shanghai Municipal Bureau of Ecology and Environment (SMBEE). These observational data were qualitatively compared with the estimated daily emissions from HDDTs in this study, as shown in Figure 5. Although the absolute magnitudes of PM concentrations and modeled emissions differ, the overall temporal trends exhibit consistent patterns, with peaks and troughs occurring on similar days. This qualitative agreement provides supportive evidence that the emission estimates derived in this study reasonably capture the spatiotemporal variations in particulate matter emissions from freight trucks in Shanghai.

Box-and-whisker plots (Figure 6) were employed to detect outliers in the daily emissions of various pollutants based on the data presented in Table 4. The analysis revealed no significant outliers in daily CO₂ or other pollutant emissions, indicating strong data stability and consistency. Additionally, the median daily emissions for all pollutants exceeded their respective means, suggesting a left-skewed distribution. This pattern is likely attributable to the national holidays in early May, during which reduced HDDT activity resulted in lower emissions and, consequently, decreased average values. It should be noted that formal uncertainty intervals and statistical significance tests were not applied in this analysis; future work will incorporate these methods to further validate the robustness and reliability of the observed emission patterns.

To further examine the emission characteristics of each pollutant, descriptive statistical analysis was performed on their daily emission levels, as summarized in

Table 5. Descriptive statistics of daily pollutant emissions from HDDTs in Shanghai (May 2023).

Pollutant	Unit	Mean	Median	Range	Std. Dev.	C.V.	Skewness	Kurtosis	95% C.I. of Mean	Normality (p < 0.001)
CO2	ton	508.8	579.48	459.19	144.06	0.28	−0.92	−0.59	±52.84	Non-normal
CO	kg	1054.39	1202.34	954.49	300.9	0.29	−0.92	−0.61	±110.37	Non-normal
NOx	kg	1977.99	2262.00	1863.09	602.5	0.3	−0.84	−0.75	±221.00	Non-normal
PM	kg	20.03	22.81	18.08	5.67	0.28	−0.93	−0.59	±2.08	Non-normal
VOCs	kg	14.35	16.39	13.04	4.12	0.29	−0.91	−0.63	±1.51	Non-normal

. The average daily emissions of CO₂, CO, NO_x, PM, and VOCs were 508.80 tons, 1054.39 kg, 1977.99 kg, 20.03 kg, and 14.35 kg, respectively. Among these pollutants, NO_x exhibited the greatest variability, with a range of 1863.09 kg (peaking on 17 May and reaching a minimum on 1 May), reflecting substantial day-to-day fluctuations likely related to variations in truck traffic intensity and port activity. In contrast, VOCs showed the smallest range (13.04 kg), suggesting more stable emission patterns.

In terms of distributional characteristics, the median values for all pollutants were higher than their corresponding means, and the skewness coefficients ranged from –0.84 to –0.92, indicating mildly left-skewed distributions. The negative kurtosis values (–0.59 to –0.75) further suggest relatively flat distribution profiles, implying that daily emissions were moderately dispersed rather than sharply peaked. The Shapiro–Wilk and Kolmogorov–Smirnov normality tests both yielded p < 0.001 for all pollutants, confirming that the daily emission distributions significantly deviate from normality. Consequently, non-parametric statistical approaches or bootstrapped intervals are more appropriate for subsequent uncertainty assessment.

Further analysis of the standard deviation and coefficient of variation (CV) confirmed that NO_x and CO displayed the highest volatility (CV = 0.30 and 0.29), while CO₂, PM, and VOCs showed comparatively stable daily emission levels. The 95% confidence intervals of the means ranged approximately within ±10–20%, consistent with the expected uncertainty range of emission estimates derived from the COPERT V model and GPS-based activity data. Overall, these results reveal distinct temporal variability patterns among pollutants and provide a quantitative basis for assessing the reliability and uncertainty of the estimated HDDT emissions.

3.2. Results of Temporal Distribution of Emissions

3.2.1. Results of Daily Emission Variation

This study examined the temporal variation in CO₂ and air pollutant emissions from HDDTs across public holidays, weekdays, and weekends. The period from 1 May to 3 May 2023, coincided with the “Labor Day” holiday, commonly referred to as Golden Week, followed by 21 weekdays and 7 weekend days. Figure 7 depicts the daily emission patterns for each pollutant.

The results demonstrate similar temporal trends across the five pollutants—CO₂, CO, PM, VOCs, and NO_x—characterized by the subsequent features: (1) Emissions during Golden Week were significantly lower compared to those recorded in other periods. This reduction reflects decreased HDDT activity during the holiday period, as inferred from GPS trajectory data. (2) Weekday emissions were generally higher than weekend levels, with peaks typically occurring from Tuesday to Thursday, forming a mid-week hump-shaped pattern. These mid-week peaks may be related to port logistics scheduling and freight demand, although direct traffic volume and operational data were not available to quantify this effect. (3) Weekend emissions were consistently lower on Sundays compared to Saturdays. Apart from the first weekend, emissions remained relatively stable across subsequent weekends.

Quantitative analysis further revealed a clear holiday effect. Across all measured pollutants (CO₂, CO, PM, NO_x, and VOCs), mean daily emissions during non-holiday periods were approximately 140–155% higher than those recorded during Golden Week. For example, CO₂ emissions were 138.5% higher during non-holiday periods compared to Golden Week. A similar pattern was observed when comparing weekdays to weekends, with average emissions being 60–75% higher across pollutants. For instance, weekday CO₂ emissions were 59.6% greater than those on weekends. These findings highlight temporal variability in HDDT emissions and suggest that logistics operations—such as port handling schedules and freight flow patterns—likely influence emission peaks. Future studies should aim to integrate traffic volume and operational data to establish a more direct quantitative link between emission patterns and logistics policy constraints.

Taking CO₂ emissions as an example, a further statistical analysis of weekday and weekend differences was conducted. The results show that the mean daily CO₂ emissions on weekdays were 552.01 tons, significantly higher than the weekend mean of 384.57 tons, indicating that freight activities are more frequent on weekdays, leading to greater CO₂ emissions. In terms of data dispersion, the standard deviation of daily CO₂ emissions on weekdays was 134.93 tons, larger than the weekend standard deviation of 89.72 tons, suggesting greater variability in emissions during weekdays. A two-sample t-test revealed a t-statistic of −3.25 with a p-value of 0.003, indicating that the difference in mean daily CO₂ emissions between weekdays and weekends is statistically significant, with weekdays exhibiting substantially higher emissions.

Although daily traffic counts for the studied expressways are not available, previous studies have shown that vehicle emissions, especially from heavy-duty trucks, are positively correlated with traffic volume. For example, Mommens et al. [47] quantified the dynamic impact of freight transport on urban air quality and found that higher truck traffic levels lead to elevated emissions of CO₂, NO_x, and other pollutants. Similarly, studies assessing short-term traffic restriction policies indicate that reductions in freight traffic can substantially decrease traffic-related emissions [48]. These findings support the qualitative interpretation that observed reductions in emissions during holidays and non-working days are consistent with expected decreases in freight truck activity.

3.2.2. Results of Hourly Emission Variation

To further investigate the hourly emission patterns of various pollutants, the results are presented in Figure 8. Overall, the average hourly emissions of CO₂, NO_x, CO, PM, and VOCs exhibited marked differences across weekdays, weekends, and the Golden Week holiday, following a consistent trend of “weekdays > weekends > Golden Week” with periodic fluctuations.

Emissions across all pollutants began to rise at around 4:00, peaking between 6:00 and 7:00, and then gradually declined with intermittent variations, reaching minimum levels between 2:00 and 3:00 the following day. On weekdays, emissions between 6:00 and 16:00 displayed a distinctive “W-shaped” curve, with three major peaks occurring approximately at 5:00–7:00, 10:00–12:00, and 14:00–16:00 These findings confirm that daytime is the primary operating period for HDDTs, accounting for approximately 67% of total daily emissions, with consistent temporal patterns across different days.

Specifically, three emission peaks occur on weekdays between 6:00 and 16:00. The first peak, observed between 5:00 and 7:00, recorded average hourly emissions of 29,815.7 kg (CO₂), 103.8 kg (NO_x), 61.1 kg (CO), 1.2 kg (PM), and 0.8 kg (VOCs). This surge likely reflects early truck departures to avoid morning congestion, coinciding with the initiation of daily logistics operations. The second peak, from 10:00 to 12:00, shows increased emissions—33,186.8 kg (CO₂), 129.2 kg (NO_x), 69.0 kg (CO), 1.3 kg (PM), and 1.0 kg (VOCs)—possibly driven by merchants’ midday restocking activities. The third peak, from 14:00 to 16:00, reaches 35,529.1 kg (CO₂), 141.8 kg (NO_x), 73.8 kg (CO), 1.4 kg (PM), and 1.0 kg (VOCs), which may be associated with trucks entering urban areas ahead of evening traffic restrictions. In contrast, weekend emissions are generally lower but still exhibit a bimodal pattern, with peaks between 5:00–7:00 and 10:00–12:00. For example, average hourly CO₂ emissions during these periods were 18,168.3 kg and 20,998.1 kg, representing reductions of 39.1% and 36.7%, respectively, compared to corresponding weekday values. This decline likely results from partial business closures and reduced freight demand on weekends, leading to decreased HDDT operations.

During Golden Week, emission patterns flattened significantly, with minimal temporal variation and the absence of pronounced peaks. This uniformity likely reflects widespread business closures and a substantial drop in logistics demand during the holiday, resulting in reduced vehicle activity and more dispersed travel times, which in turn led to a consistently low and stable emission profile throughout the day. It is worth noting that the peak periods of all pollutants closely correspond to the observed peak traffic times of heavy-duty trucks, which is reasonable given that emission estimates are directly derived from vehicle activity and instantaneous speeds. This temporal alignment indicates that the emission peaks accurately reflect the underlying freight traffic dynamics.

3.3. Results of Spatial Distribution of Emissions

This study further explores the spatial distribution of CO₂ and other key air pollutants along the six major freight expressways in Shanghai. To quantify hourly emissions relative to the length of each route, the metric of emission intensity is introduced, calculated as follows:

$$I_e={\displaystyle \frac{M_p}{L_r\cdot H_t}}$$

(9)

where $I_e$ represents the emission intensity, expressed in kg/(km·h); $M_p$ denotes the total pollutant emissions accumulated over one month, in kilograms (kg); $L_r$ is the length of the road segment, in kilometers (km); $H_t$ refers to the total duration over the month, in hours (h). The emission intensity metric of kg/(km·h) was chosen to quantify pollutant emissions relative to both road length and time. This unit allows for the assessment of temporal variations (e.g., peak vs. off-peak hours) and spatial differences across road segments, facilitating more targeted emission control strategies.

Figure 9 presents both the total emissions and emission intensities of five major pollutants—CO₂, CO, NO_x, PM, and VOCs—emitted by heavy-duty diesel trucks (HDDTs) operating along six major freight expressways in Shanghai. As previously defined, emission intensities quantify emissions normalized by route length and time. Total emissions refer to the cumulative amount of each pollutant emitted over the course of one month on each expressway. The results reveal substantial spatial variation, with the G1503 and S20 expressways showing the highest total emissions, while S6 exhibits the lowest. Specifically, total emissions on G1503 reached 6129.1 tons (CO₂), 12.45 tons (CO), 19.59 tons (NO_x), 241.3 kg (PM), and 167.5 kg (VOCs). In contrast, S6 recorded just 324.8 tons, 0.68 tons, 1.46 tons, 12.8 kg, and 9.4 kg for the respective pollutants—representing reductions of approximately 94.7%, 94.5%, 92.5%, 94.7%, and 94.4% compared to G1503.

In addition, the results show that S20 had the highest emission intensities across all pollutants, with values of 36.7 kg/(km·h) for CO₂, 0.078 kg/(km·h) for CO, 0.184 kg/(km·h) for NO_x, 0.001 kg/(km·h) for PM, and 0.001 kg/(km·h) for VOCs. While CO₂ showed significantly higher intensities, PM and VOCs remained comparatively low. In contrast, S32 had the lowest emission intensities, at 6.88 kg/(km·h), 0.0139 kg/(km·h), 0.0216 kg/(km·h), 0.0003 kg/(km·h), and 0.0002 kg/(km·h) for the respective pollutants—amounting to roughly 81–84% lower intensities compared to S20.

Although G1503 recorded the highest total emissions, its emission intensity was not proportionally high. Conversely, S6—despite its low total emissions—had emission intensities comparable to G1503. This contrast highlights the critical role of road length in shaping emission intensity. G1503 spans 391.24 km, allowing for broader geographic dispersion of emissions and a lower per-unit pollution burden. In contrast, the shorter length of S6 (21.75 km) concentrates emissions spatially, resulting in higher intensity despite lower absolute totals.

To comprehensively examine the spatial distribution of total emissions from heavy-duty diesel trucks (HDDTs), this study employed ArcGIS to generate spatial maps of monthly cumulative emissions for five key pollutants across unit-length segments of six major freight expressways in Shanghai (Figure 10). The results reveal consistent spatial patterns across all pollutants, with high-emission zones primarily concentrated in the eastern and northern segments of G1503 and S20. Notably, significant emission hotspots were identified near the Dating Interchange—where S2 connects Luchao Port to G1503—and in the northeastern overlapping section between G1503 and S20.

These areas exhibit the highest pollutant concentrations observed throughout the month, marking them as critical emission hotspots. In contrast, lower emission levels were recorded along the western segment of G1503 and the entirety of S32, indicating comparatively reduced pollution loads in these regions. Although the spatial distribution patterns of the five pollutants are largely consistent, NO_x displays distinct variability, with more concentrated hotspots along S20. This underscores the importance of targeted NO_x mitigation strategies along that corridor.

3.4. Results of Spatiotemporal Distribution of Emissions

To further clarify the spatiotemporal distribution patterns of emissions, average daily heatmaps for weekdays and weekends were constructed (Figure 11) based on the previously analyzed temporal and spatial characteristics of five pollutant categories: CO₂, CO, NO_x, PM, and VOCs. Specifically, the weekend heatmaps were averaged over four weekend days in May 2023 (two Saturdays and two Sundays), while the weekday heatmaps were averaged over twenty weekdays in the same month, ensuring that the presented patterns are representative of typical weekday and weekend conditions.

The weekend emission heatmaps reveal regional distribution patterns that generally align with the monthly cumulative emission patterns shown in Figure 10. High-emission zones are primarily concentrated in the eastern and northern sections of G1503 and S20, particularly around the Dating Interchange—which connects Luchao Port via S2 (Shanghai–Luchao Port Expressway) to G1503—and the northeastern segment linking G1503 and S20. CO₂ emissions closely follow the overall spatial trend, while CO and NO_x show more localized variations. Specifically, the high-emission zones for CO and NO_x along the northern and eastern segments of S20 are more limited in extent, and their intensities are lower than those of CO₂ in the eastern G1503 segment and the S2 corridor from Luchao Port to Dating Interchange. For pollutants with relatively low total emissions, particularly PM and VOCs, spatial distribution patterns differ more noticeably. PM hotspots are largely consistent with those of CO₂ and are especially prominent along S20, the southern section of G1503, and G15. In contrast, the spatial distribution of VOCs more closely resembles that of CO and NO_x.

The weekday emission heatmaps display distinct characteristics compared to the cumulative monthly distribution in Figure 10, primarily due to a general intensification of pollutant emissions during weekdays. This is evidenced by a widespread increase in emission levels across various road segments, as reflected by a color shift in the heatmaps from green to red, indicating an approximate one-level rise in emission intensity. CO₂ emissions, in particular, exhibit significant weekday increases across most road segments—excluding S32, the western section of G1503, and the central portion of S2. Notable intensification is observed along the Luchao Port-to-Dating Interchange corridor and the northeastern overlapping segment of G1503 and S20, illustrating typical weekday emission hotspots. Although the spatial distributions of CO and NO_x partially overlap with those of CO₂, discrepancies remain in both the extent and magnitude of high-emission zones. Similarly to weekend patterns, the high-emission segments of CO and NO_x in the northern and eastern portions of S20 are shorter and less intense than those of CO₂, particularly in the eastern G1503 section and the S2 segment from Luchao Port to Dating Interchange, indicating spatial inconsistencies among pollutants. For PM and VOCs, their weekday spatial distribution characteristics remain broadly consistent with weekend trends and are therefore not reiterated here.

A comparison of pollutant spatial distributions between weekdays and weekends reveals substantial differences in both emission intensity and spatial extent. Emission levels are generally higher on weekdays, with pronounced hotspots concentrated around the northeastern intersection of G1503 and S20, as well as along the S2 corridor from Luchao Port to the Dating Interchange, exhibiting clear spatial clustering. In contrast, weekend emissions are markedly lower in both intensity and coverage, although the locations of major emission zones remain relatively consistent. This pattern reflects a significant decline in freight transportation activity during weekends. Furthermore, the contraction of emission spread along key freight corridors on weekends highlights the direct influence of traffic volume fluctuations on the spatiotemporal dynamics of pollutant emissions.

This study further examines the spatiotemporal distribution patterns of five major pollutants during three representative peak periods: 05:00–07:00, 10:00–12:00, and 14:00–16:00. The results indicate that the spatial emission patterns during these periods are consistent with the overall distribution, exhibiting significant clustering. These findings align with those reported by Zhou et al. [49], further confirming the presence of emission hotspots along key freight corridors. While all pollutant categories show spatial concentration along similar road segments, notable differences in emission intensity are evident: CO₂ emissions are the highest, followed by CO and NO_x, with PM and VOCs showing comparatively lower levels.

As illustrated in Figure 12, during the 05:00–07:00 period, CO₂ emission hotspots are primarily located in the northeastern section of the S20 expressway and along the S2 corridor from Luchao Port to the Dating Interchange, reflecting early-morning freight departures. Between 10:00 and 12:00, the hotspots shift slightly: the northeastern segment of S20 remains a high-emission area, while emissions increase along the eastern section of G1503 and the central portion of S32. In contrast, emissions along S2 decrease compared to the morning peak. During the 14:00–16:00 period, emission hotspots expand further, with the eastern segment of G1503, northeastern S20, and S2 once again emerging as key emission zones, indicating heightened freight activity in the afternoon.

Figure 13, Figure 14, Figure 15 and Figure 16 demonstrate that the spatiotemporal distribution patterns of CO closely mirror those of CO₂, with both pollutants exhibiting substantial regional overlap in their emission hotspots. In contrast, NO_x emissions remain relatively stable, consistently concentrated in the northeastern segment of the S20 expressway throughout all three peak periods, showing limited spatial variability over time. PM and VOC emissions exhibit more pronounced temporal fluctuations. Their high-emission zones during the three time periods are mainly concentrated along the northeastern section of S20, the eastern section of G1503, and the southeastern portion of S2. Notably, during the 10:00–12:00 period, emissions in the southeastern segment of S2 decline slightly, followed by a marked rebound from 14:00 to 16:00, likely reflecting increased return freight traffic in the afternoon.

4. Discussion

4.1. Analysis of Pollutant Emission from HDDTs

Among the pollutants estimated in this study, CO₂ was identified as the dominant emission from HDDTs. In May 2023, HDDTs operating on six major freight expressways in Shanghai emitted approximately 15,772.81 tons of CO₂. According to the Carbon Monitor dataset [46], Shanghai’s total CO₂ emissions and ground transportation-related CO₂ emissions in the same month were 10.212 million tons and 2.62 million tons, respectively. This indicates that ground transportation accounted for about 25.65% of the city’s total emissions, with HDDTs on the six expressways contributing around 6% of the ground transportation emissions and 1.54% of the city-wide total. Although our analysis did not include emissions from connecting secondary roads or from trucks operating at speeds below 5 km/h and above 100 km/h, the contribution of HDDTs remains considerable, underscoring their significant role in shaping both ground transportation emissions and overall CO₂ emissions in Shanghai.

In addition to CO₂, our results reveal that NO_x and CO also emerge as major pollutants from HDDT operations. Elevated NO_x emissions are particularly concerning, as they not only exacerbate respiratory illnesses but also act as precursors to ground-level ozone through photochemical reactions with VOCs. Meanwhile, CO emissions impair oxygen transport in the bloodstream, posing direct risks to cardiovascular and neurological health, particularly for sensitive populations such as children, the elderly, and individuals with pre-existing conditions. These health-related consequences highlight that mitigating HDDT emissions would not only reduce greenhouse gas contributions but also deliver substantial co-benefits for urban air quality and public health protection.

Estimates of small emissions, such as PM and VOCs, are particularly sensitive to GPS inaccuracies due to their relatively low absolute values and high spatial variability. In this study, considering typical GPS positioning errors under open expressway conditions (approximately 3–10 m), the resulting relative uncertainty in estimated vehicle speed is about ±2%, which translates into an estimated ±15–30% impact on minor emissions, while major emissions such as CO₂ and NO_x are less affected. In future work, the influence of GPS uncertainties on emission estimates will be systematically assessed using Monte Carlo simulations and sensitivity analysis to provide a more robust quantification of uncertainty and its effect on model predictions.

Although the COPERT V model was not directly calibrated using local measurement data due to the unavailability of portable emission measurement system (PEMS) or remote sensing data in Shanghai, previous studies have demonstrated that COPERT V can provide reasonably accurate estimates for HDDVs in China. For example, Deng et al. [50] reported relative errors of approximately 15% between PEMS-measured emissions and model estimates for logistics vehicles in Beijing, while Zhao et al. [51] provided measured CO₂ and NO_x emissions from container trucks to serve as a benchmark for model estimation. Based on these comparisons and in combination with the localization of key parameters, our estimation using COPERT V is expected to have an uncertainty range of ±10–20%. Therefore, although some uncertainty remains due to the absence of direct calibration, the estimated emissions can still be regarded as representative of the overall spatiotemporal patterns of HDDT-related emissions in Shanghai. Future work will focus on integrating local on-road emission data to refine model parameters and further improve estimation accuracy.

4.2. Discussion on Spatiotemporal Characteristics of Emissions

Research has found that among the emissions from HDDTs on six major freight expressways in Shanghai, CO₂ is the primary pollutant, accounting for a significant proportion, followed by NO_x and CO. This indicates that carbon emissions and nitrogen oxide emissions are the core pollution characteristics of the freight system in typical port cities.

The temporal distribution analysis reveals pronounced diurnal and weekly variability in HDDT emissions. On weekdays, emissions of all major pollutants are substantially higher than those on weekends and public holidays. This difference is mainly due to increased freight demand and intensified logistics operations during, while weekends experience reduced business activity and fewer deliveries; additionally, regulatory restrictions commonly limit heavy-duty vehicle traffic on weekends, all of which contribute to lower emissions. Daily emission peaks occur primarily during three-time intervals: 5:00–7:00, 10:00–12:00, and 14:00–16:00. These peak periods reflect freight vehicle operational behavior that avoids urban commuting hours, indicating a strong alignment between freight emission patterns and the broader rhythm of transportation activities. The three weekday peaks are likely influenced by freight transport management and port-related operational schedules. For instance, nighttime delivery restrictions and designated freight corridors may shift truck activities toward early morning and late evening hours. In addition, port dispatch and cargo handling schedules often lead to concentrated truck flows during mid-day periods. These institutional factors provide a plausible explanation for the temporal emission peaks identified in this study and highlight the importance of integrating traffic management policies into emission reduction strategies.

Notably, daytime hours (6:00 to 16:00) account for approximately 67% of total daily emissions, highlighting the high intensity of freight operations during this period. In contrast, emissions decline significantly on weekends and during the Golden Week holiday, with two smaller peaks observed at 5:00–7:00 and 10:00–12:00, mainly due to business closures and reduced freight demand. During Golden Week, emissions exhibit a marked “flattening” effect, characterized by more evenly distributed levels throughout the day, likely reflecting widespread business shutdowns and substantially decreased logistics activity. These temporal patterns are consistent with findings from the Shenzhen Port study, which also observed increased freight activity and pollutant emissions on weekdays in port cities [52].

Spatial distribution characteristic analysis shows that the high emission areas of five types of HDDTs emissions (CO₂, NO_x, CO, PM, VOCs) on six expressways are highly consistent in space, mainly concentrated in the eastern and northern parts of G1503 and S20, as well as the S2 section from Luchao Port to Dating Interchange, reflecting the emission aggregation characteristics of freight corridors around the port. These segments emerged as the most critical emission hotspot, contributing approximately 13.74% of total CO₂ emissions across the six studied expressways. Similar patterns were observed for NO_x and CO, indicating that mitigation efforts targeting this corridor could yield substantial reductions in overall emissions. In addition, a preliminary sensitivity assessment indicates that if the truck volume on S20 were increased by 10% relative to G1503, CO₂ emissions would increase by approximately 17.2%, assuming other factors remain constant. These quantifications provide a clearer basis for prioritizing emission control strategies along critical freight corridors in Shanghai.

Especially in the northeastern area where S20 intersects with G1503, the emission intensity in this region is significantly higher than in other sections due to the single channel for incoming and outgoing port traffic and frequent transfers. The emission intensity per unit length of S20 is about five times that of S32, while G1503, despite having the highest total emissions, has its unit intensity diluted by its longer road segments. This contrast between total emissions and unit intensity reveals the combined effect of freight traffic density and road geometric characteristics on emission distribution. These findings suggest that targeted mitigation strategies—such as traffic rerouting, optimization of port dispatch schedules, or speed regulation—could effectively reduce high-intensity emissions on shorter segments like S20 while also lowering overall emissions on longer corridors such as G1503. Quantifying both total emissions and emission intensity thus provides a clear basis for prioritizing interventions along critical freight corridors in Shanghai.

Further spatiotemporal cross-analysis indicates that high-emission segments on weekdays are more widely distributed and exhibit higher intensity, particularly concentrated in the northeastern key hub area. This pattern is closely linked to persistent traffic congestion along critical corridors such as the S20 Northeast segment, which connects the G1503 expressway with Shanghai Port. Congestion in this area arises from limited access routes and high freight vehicle volume, causing vehicles to operate at low speeds or engage in frequent stop-and-go movements. These conditions reduce engine efficiency and increase fuel consumption, resulting in elevated pollutant emissions per unit distance [53]. Consequently, NO_x, PM, and VOCs show relatively stable spatial concentrations, accumulating over time in the northeastern section of S20, reflecting the cumulative effects of congestion. In contrast, CO₂ and CO emission hotspots exhibit noticeable temporal migration during the three peak periods, with outbound transportation dominating in the early morning and return trips intensifying in the afternoon. On weekends, due to reduced freight demand, high-emission segments are shorter, and their emission intensity decreases. It worth notes that the spatiotemporal distribution patterns of CO₂ emissions obtained in this study are consistent with the findings of [54], where major freight corridors and port-related access roads were identified as key emission hotspots. Such consistencies further validate the plausibility of our results.

However, a detailed analysis of migration zones was not conducted in this study due to the unavailability of high-resolution mobility and meteorological data. Nevertheless, preliminary spatial patterns of emissions and truck activity suggest potential areas of high vehicle movement, which may correspond to migration zones. For a comprehensive assessment, future studies would require datasets such as spatiotemporal mobility patterns, high-resolution vehicle trajectory data, and local meteorological conditions. Incorporating such data would enable more accurate characterization of migration zones and their temporal variability, providing deeper insights into traffic-induced atmospheric emissions.

4.3. Policy Implications

Based on the spatiotemporal characteristics of HDDT emissions identified in this study, the following policy recommendations are proposed to support more refined management strategies for port cities, particularly in the context of coordinated “carbon reduction” and “pollution reduction” governance efforts.

(1) Promote nighttime freight operations to balance traffic load and distribute emissions more evenly. This study reveals that HDDT emissions are most concentrated during daytime hours on weekdays (6:00 to 16:00), accounting for approximately 67% of total daily emissions. To mitigate the environmental pressure during these peak periods, policies should encourage freight companies to optimize logistics schedules and promote nighttime transportation. Implementing “staggered freight scheduling” can help alleviate congestion and reduce pollutant emissions, similar traffic-demand control policies and ITS signal optimization in other studies have achieved CO and NO_x emission reductions in the range of 10–25% [55]. Nevertheless, the promotion of nighttime operations must proceed cautiously, with adequate provisions for road lighting, speed control, and rest area infrastructure to ensure the safety of drivers and overall transport operations.

(2) Promote the implementation of the “dynamic traffic restriction” and “staggered transportation” policies. Given the three distinct emission peaks—morning (5:00–7:00), midday (10:00–12:00), and afternoon (14:00–16:00)—which closely coincide with intensive freight activity, there is a strong rationale for adopting dynamic traffic control measures based on high-resolution emission data. By leveraging intelligent transportation systems (ITS), real-time traffic segment regulation and time-window guidance for freight vehicles can be implemented. Such adaptive strategies would help redistribute freight flows more efficiently throughout the day, thereby optimizing the temporal structure of emissions. Case studies of traffic demand control and ITS signal control have shown CO reductions of ~20–30% and NO_x reductions of ~10–20% under similar measures [56].

(3) Promote the adoption of new energy heavy-duty trucks and low-emission vehicles based on local conditions. Targeted emission control should be prioritized in high-intensity areas, such as the northeastern segment of S20 and the S2 corridor near the Dating Interchange, which exhibits the highest levels of HDDT emissions. In these zones, pilot programs for clean energy heavy-duty trucks, including LNG, hydrogen fuel cells, and electric vehicles—can be implemented, supported by the development of essential infrastructure such as charging stations, battery swapping facilities, and clean fuel refueling points. Concurrently, fleets still operating traditional diesel vehicles, progressively tightening emission access standards can serve as a regulatory mechanism to incentivize timely upgrades and accelerate fleet modernization.

To further evaluate the potential effectiveness of clean-fuel strategies, a simplified scenario analysis was performed based on our estimated emissions. In May 2023, HDDTs operating on the six studied expressways emitted approximately 15,773 tons of CO₂ and 61,318 tons of NO_x. Literature suggests that LNG trucks can achieve lower emissions than their diesel counterparts under comparable operating conditions. Specifically, Sütheö and Háry [57] reported that LNG heavy-duty trucks emit approximately 3–15% less CO₂, while Chhugani et al. [58] found that NO_x emissions can be reduced by around 31–33% compared with low-sulfur diesel trucks. Assuming that 20% of HDDTs were replaced by LNG trucks, the total CO₂ reduction would range from 95 to 473 tons, while NO_x emissions could decrease by 3800–4050 tons during the study month. For comparison, replacing the same proportion of HDDTs with electric trucks—characterized by near-zero tailpipe emissions—would reduce approximately 3155 tons of CO₂ and 12,264 tons of NO_x. These results highlight that LNG substitution offers modest CO₂ reductions but substantial NO_x mitigation, whereas electric trucks present near-zero tailpipe emissions. Although these estimates are preliminary and do not account for upstream emissions, fuel supply chains, or operational constraints, they underscore the importance of alternative-fuel adoption in reducing the environmental footprint of HDDT operations in Shanghai’s port-related transport corridors.

(4) Optimize traffic organization and infrastructure layout in port-adjacent areas. The study reveals that high-emission zones are predominantly concentrated along primary access routes to and from the port, highlighting the emission aggregation effect caused by a single-channel transport structure. To address this, a systematic optimization of the port’s collection and distribution network is needed, with consideration for road hierarchy and functional coordination. Measures such as the construction of dedicated freight lanes and the strategic diversion of container trucks across multiple exit routes can enhance traffic flow efficiency while mitigating localized emission accumulation. In practice, similar freight lane or signal optimization schemes in other urban corridors have been shown to reduce CO₂ or NO_x emissions by 5–15% [59].

(5) Promote regional logistics coordination to mitigate emission concentration along long-distance corridors. Although the G1503 corridor exhibits the highest total emissions, its lower emission intensity per unit length underscores its function as a major transit route for regional freight flows. To reduce the over-reliance on road transport for medium- and long-distance logistics, the development and enhancement of multimodal transport nodes—such as rail and inland waterway transfer hubs—should be prioritized along key sections of G1503. Facilitating road-to-rail and road-to-water modal shifts can help redistribute freight volumes more sustainably and alleviate emission burdens on primary highway corridors. Under emission-cap or modal-shift scenarios, the literature suggests that converting a modest share (~10–20%) of long-haul freight to rail/inland waterways may yield 10–30% reductions in transport CO₂ emissions, depending on regional conditions [60].

(6) Establish an emission monitoring system supported by multi-source data to enable precise control. Building on the integration of high-frequency GPS trajectories, detailed road network data, and localized emission factors as utilized in this study, it is recommended that government agencies develop a dynamic emission monitoring and early warning platform. A vehicle emissions monitoring system may use NDIR CO₂ sensors, chemiluminescence NO_x sensors, and optical PM sensors, with a measurement frequency of 1–5 min and real-time data transmission via IoT networks, enabling timely and accurate data acquisition. Such a system would provide robust technical support for the coordinated governance of traffic and environmental quality. By enabling real-time data collection and analysis, this approach facilitates a shift in the regulatory model from post hoc “emission assessment” to proactive “real-time control,” thereby enhancing the responsiveness and effectiveness of emission management strategies. Some pilot projects employing integrated traffic–emission monitoring and adaptive traffic control have demonstrated measurable air-quality benefits. Field and case studies report reductions in smog-forming pollutants on the order of single digits to a few tens of percent depending on context: for example, intersection-level adaptive control achieved a ~7.7% reduction in CO in one field trial [61], while signal-timing optimization in mixed traffic scenarios including heavy vehicles yielded ~12–24% reductions in NO_x in simulation/field assessments [62]. These results indicate that corridor-level adaptive measures can produce meaningful local improvements, although outcomes are highly sensitive to traffic composition, control algorithms, and local implementation.

5. Conclusions and Future Work

This study focuses on six major port collection and distribution expressways in Shanghai, utilizing GPS trajectory data from heavy-duty diesel trucks (HDDTs) collected in May 2023, along with road attribute data extracted from OpenStreetMap. Based on the COPERT V model—localized and calibrated for regional fuel characteristics—a high spatiotemporal resolution emission inventory was developed, covering five key pollutants: CO₂, NO_x, CO, PM, and VOCs. The study systematically reveals the temporal variation patterns, spatial distribution characteristics, and spatiotemporal coupling mechanisms of HDDT emissions. The main conclusions are as follows:

(1) There are marked differences in both the total volume and composition of emissions. CO₂ accounts for the largest share, followed by NO_x and CO, while VOCs and PM contribute relatively smaller amounts. Notably, daytime emissions of NO_x and CO show significant fluctuations, indicating high sensitivity to traffic flow variations. In contrast, CO₂, PM, and VOCs display more stable temporal emission patterns. To mitigate peak-time emissions, logistics schedules can be optimized through nighttime freight operations and staggered transport arrangements, thereby easing congestion and reducing pollution.

(2) Temporal analysis reveals that reduced freight activity during holidays leads to a significant drop (approximately 40–45%) in emissions, whereas emissions on weekdays rise markedly, forming a characteristic “convex” pattern. On an hourly scale, emission trends exhibit a distinct “W-shaped” or bimodal structure, reflecting the influence of commuting peaks and freight scheduling practices. Dynamic traffic management measures, such as time-window restrictions and adaptive scheduling guided by intelligent transportation systems, should be promoted to flatten emission peaks and balance weekday freight flows.

(3) Spatially, the G1503 and S20 expressways emerge as the primary contributors to HDDT emissions, with S20 showing the highest emission intensity per unit length. Emission hotspots are predominantly located in the eastern and northern segments of G1503 and S20, as well as the corridor stretching from S2 Luchao Port to the Dating Interchange, indicating pronounced spatial clustering. Targeted mitigation in these hotspots should prioritize the introduction of clean-energy heavy-duty trucks (e.g., LNG, hydrogen, electric) and the construction of supporting infrastructure, while progressively tightening emission standards for conventional diesel fleets.

(4) Spatiotemporal coupling analysis demonstrates that while the spatial distribution of emissions remains relatively stable across weekdays and weekends, there are notable differences in emission intensity and spatial extent. High-emission zones are primarily concentrated near port entrances/exits and expressway interchanges, exhibiting a temporal evolution pattern characterized by early departures and midday returns. Management strategies should focus on optimizing freight flows at port gateways and interchanges, including staggered departure schemes, priority lanes for clean trucks, and digital platforms for coordinated port–road scheduling.

The high-resolution emission inventory constructed in this study can provide foundational support for road traffic management, pollution control, and differentiated regulation in port cities. The results reveal significant spatiotemporal clustering and structural differences in HDDTs emissions in the port area, emphasizing the importance of periodical and regional management, particularly the joint governance of traffic hubs and high-emission sensitive periods.

Despite the valuable insights gained from this study on the spatiotemporal characteristics of HDDT emissions, several limitations should be acknowledged. First, the vehicle classification adopted was relatively coarse. The absence of finer distinctions regarding vehicle models, fuel types, and cargo attributes may obscure important variations in emission profiles, thereby reducing the representativeness of the results. Second, the analysis relied on GPS trajectory data collected during a single month (May 2023), which primarily reflects springtime logistics activities. This temporal limitation restricts the ability to capture seasonal or annual variations in emission distribution and intensity. Third, uncertainties associated with both emission factors and GPS data quality were not explicitly quantified. Such uncertainties may affect the accuracy of total emission estimates and the identification of emission hotspots, particularly under conditions of traffic congestion. Fourth, although the COPERT V model provided a robust basis for emission factor estimation, its applicability may be limited when confronted with heterogeneous and high-resolution datasets. Fifth, this study concentrated on constructing a high-resolution emission inventory and analyzing its spatiotemporal patterns but did not extend to assessing the subsequent impacts on air quality and human health. In addition, the bottom-up approach provides high-resolution emission estimates but is subject to potential errors. These include GPS positional inaccuracies, assumptions regarding vehicle type and load, fixed road segment divisions, and data processing steps such as filtering and smoothing, which may inadvertently bias emission estimates. Moreover, the map-matching method employed in this study was based on the nearest neighbor principle, which may lead to location misallocation in dense road networks or complex junctions.

To address these limitations, several directions for future research are proposed. A more detailed vehicle classification framework, incorporating parameters such as fuel type and cargo characteristics, should be adopted to improve the accuracy of emission characterization. Expanding the temporal scope of data collection to multiple seasons or a full year will enable the assessment of seasonal cycles and inter-annual variability in HDDT emissions. Incorporating uncertainty and sensitivity analyses will be essential to better evaluate the robustness of emission estimates. To enhance emission factor estimation under diverse operating conditions, the integration of machine learning techniques with traditional models is recommended. Furthermore, local validation of emission models should be conducted to increase the reliability of the results. Future studies should also incorporate detailed traffic composition, road geometry, and traffic density data to better control for factors influencing emission differences across expressways. Beyond emission inventory development, future studies should link HDDT emissions with advanced atmospheric chemistry and transport models, such as CMAQ and WRF-Chem, to simulate pollutant dispersion, chemical transformation, and population exposure. Additionally, future studies could mitigate the limitations of bottom-up approach by incorporating higher-resolution GPS data, more detailed vehicle classifications, and validation against on-road measurements. In particular, the adoption of more advanced map-matching algorithms (e.g., probabilistic or topology-based methods) should be considered to reduce positional errors in complex urban road environments. By integrating multi-source traffic and road data with improved modeling methods, future research can provide a more comprehensive and accurate assessment of the environmental and health impacts of HDDT emissions, thereby supporting more targeted and effective policy decisions.