Nothing Special   »   [go: up one dir, main page]
Next Article in Journal
TLSynth: A Novel Blender Add-On for Real-Time Point Cloud Generation from 3D Models
Previous Article in Journal
Distinguishing Lithofacies of Flysch Formations Using Deep Learning Models: Integrating Remote Sensing Data with Morphological Indexes
Previous Article in Special Issue
HoloGaussian Digital Twin: Reconstructing 3D Scenes with Gaussian Splatting for Tabletop Hologram Visualization of Real Environments
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 17
Issue 3
10.3390/rs17030423
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Analyzing Travel and Emission Characteristics of Hazardous Material Transportation Trucks Using BeiDou Satellite Navigation System Data

by
Yajie Zou
,
Qirui Hu
,
Wanbing Han
,
Siyang Zhang
and
Yubin Chen
*
Key Laboratory of Road and Traffic Engineering of Ministry of Education, Tongji University, Shanghai 201804, China
*
Author to whom correspondence should be addressed.
Remote Sens. 2025, 17(3), 423; https://doi.org/10.3390/rs17030423
Submission received: 1 December 2024 / Revised: 18 January 2025 / Accepted: 22 January 2025 / Published: 26 January 2025
(This article belongs to the Special Issue Application of Photogrammetry and Remote Sensing in Urban Areas)
Browse Figures
Figure 1
<p>Overall methodological framework.</p> "> Figure 2
<p>Vehicle speed distribution and fitted curve (<b>a</b>); comparison of static drift points and moving points direction change (<b>b</b>).</p> "> Figure 3
<p>Trip generation results (OD distribution) of HMTTs in Shanghai.</p> "> Figure 4
<p>Travel distance distribution (<b>a</b>) and time distribution (<b>b</b>) of HMTTs.</p> "> Figure 5
<p>Distribution of Vehicle Departure and Arrival Times.</p> "> Figure 6
<p>Daytime—Surrounding cities travel mode of HMTTs. Spatial distribution of travel mode 1 is shown as sub-figure (<b>a</b>), and temporal distribution of travel mode 1 is shown as sub-figure (<b>b</b>).</p> "> Figure 7
<p>Early morning—In-city travel mode of HMTTs. Spatial distribution of travel mode 2 is shown as sub-figure (<b>a</b>), and temporal distribution of travel mode 2 is shown as sub-figure (<b>b</b>).</p> "> Figure 8
<p>Midnight—scattered travel mode of HMTTs. Spatial distribution of travel mode 3 is shown as sub-figure (<b>a</b>), and temporal distribution of travel mode 3 is shown as sub-figure (<b>b</b>).</p> "> Figure 9
<p>Temporal (<b>a</b>) and spatial (<b>b</b>) emission characteristics of HMTTs.</p> "> Figure 10
<p>Comparison of road traffic flow (<b>a</b>), average speed (<b>b</b>), and emission level (<b>c</b>).</p> "> Figure 11
<p>Coefficients of GWR for different land use types: (<b>a</b>) Industrial-related POIs; (<b>b</b>) Company-related POIs; (<b>c</b>) Entertainment-related POIs; (<b>d</b>) Education-related POIs.</p> ">
Versions Notes

Abstract

:
Road hazardous material transportation plays a critical role in road traffic management. Due to the dangerous nature of the cargo, hazardous material transportation trucks (HMTTs) have different route selection and driving characteristics compared to traditional freight trucks. These differences lead to unique travel and emission patterns, which in turn affect traffic management strategies and emission control measures. However, existing research predominantly focuses on safety aspects related to individual vehicle behavior, with limited exploration of the broader travel and emission characteristics of HMTTs. To bridge this gap, this study develops a comprehensive framework for analyzing the travel patterns and emissions of HMTTs. The methodology begins by applying a Gaussian mixture distribution model to identify vehicle stop points, eliminating biases associated with subjective settings. Origin–destination (OD) pairs are then determined through stop time clustering, followed by the extraction of travel characteristics using non-negative matrix factorization. Emissions are subsequently calculated based on the identified trip data. The relationship between emissions and land use characteristics is further analyzed using geographically weighted regression (GWR). Crucially, this study leverages data from the BeiDou Satellite Navigation System, focusing on HMTTs operating within Shanghai. The processed data reveal three distinct travel modes of HMTTs, categorized by spatiotemporal patterns: Daytime—Surrounding cities, Early morning—In-city, and Midnight—Scattered. Moreover, unlike other road vehicles, HMTT emissions are heavily influenced by industrial and company-related points of interest (POIs). These findings highlight the significant role of BeiDou Satellite Navigation System data in optimizing HMTT management strategies to reduce emissions and improve overall safety.

1. Introduction

Hazardous material transportation trucks (HMTTs) are a critical component of logistics and supply chain management due to their significant implications for public safety, environmental health, and operational efficiency. As carriers of potentially dangerous substances, HMTTs require meticulous planning and management to minimize risks and ensure safe transportation. However, the challenges associated with HMTTs extend beyond safety concerns, encompassing environmental impacts such as heightened greenhouse gas emissions and fuel consumption.
Compared to standard freight vehicles, HMTTs face unique operational constraints due to the hazardous nature of their cargo. These constraints result in distinct travel and emission patterns. HMTTs are typically required to follow specific routes, avoid high-risk areas (e.g., densely populated zones, tunnels), and adhere to speed limits to minimize the risk of hazardous material spills. These operational factors not only influence travel behavior but also lead to higher fuel consumption and increased emissions. Additionally, HMTTs often carry heavier loads, further reducing fuel efficiency. As a result, HMTTs emit more greenhouse gases and pollutants than traditional freight vehicles [1,2,3,4,5,6], contributing to serious environmental issues such as climate change, air pollution, and acid rain. This underscores the importance of studying their emissions in greater detail. By identifying emission profiles and the factors that drive these emissions, effective strategies can be developed to reduce the carbon footprint of hazardous material transportation [7,8].
Despite the growing recognition of these challenges, existing research has primarily focused on risk prediction and route planning for HMTTs, leaving critical gaps in the analysis of their travel dynamics and environmental impacts [9,10,11,12]. Specifically, there is a lack of detailed exploration into the decomposition of HMTT travel patterns. Travel pattern decomposition is crucial for uncovering the underlying connections among HMTT trips, allowing for more precise and targeted management strategies tailored to different travel modes and risk levels [13,14]. Furthermore, the specific emission characteristics of HMTTs have not been sufficiently studied, despite their relevance to environmental sustainability and the urgent need to develop policies that address the ecological consequences of hazardous material transportation.
This study seeks to address these research gaps by focusing on three key objectives: extracting and analyzing the travel characteristics of HMTTs, decomposing their travel patterns, and evaluating their carbon emissions. By leveraging BeiDou satellite data, this research provides detailed insights into HMTT movements and emission profiles, enabling a more comprehensive understanding of their impact. By studying the relationship between HMTT emissions and building environmental factors, the research reveals distinct emission characteristics of HMTTs compared to regular vehicles. The findings of this study contribute to the development of targeted strategies for reducing HMTT-related emissions, offering guidance for policymakers, transportation planners, and environmental advocates.
The remainder of our paper is structured as follows. Section 2 provides a detailed review of the related literature and establishes the theoretical framework for our study. Section 3 introduces the methodology used. Section 4 introduces the BeiDou data utilized in this study, providing a detailed description of the data format and outlining the methods used for data processing. Section 5 presents the results, accompanied by an in-depth discussion of key findings. Section 6 discusses how our findings echo the current literature and the critical new knowledge from our study. Section 7 concludes the paper by summarizing the main contributions and outlining potential future research directions.

2. Literature Review

Trip generation methods can be divided into two methods: traditional travel survey methods like paper questionnaires, telephone interviews, and mail inquiry [15]; and positioning data such as bus IC card data, smartphone satellite data, and vehicle positioning data. Traditional travel survey methods are widely used as they can provide many detailed descriptions and catch subtle travel phenomena. However, disadvantages such as high cost, long update cycle, and collection bias make it difficult to collect and use in a macroscope travel characteristics analysis.
Various methodologies to identify trip ends from trajectory data have been proposed in previous studies and most of them consist of two parts: (1) stop points identification from trajectory data and (2) trip ends (ODs) identification from stop points. For the former part, spatial clustering and speed threshold setting are two of the most widely used methods employed. The density-based spatial clustering of applications with noise (DBSCAN) algorithm has been widely used for its capability to recognize high-density areas. Diana et al. [16] used the DBSCAN method to identify the loading and unloading locations of trucks in urban areas. Sun et al. [17] implemented a customized DBSCAN to identify stop points based on passive trajectory data. However, clustering-based methods are highly dependent on parameter selection and perform poor transferability across different datasets. Speed threshold setting is a simple and commonly used method but is often considered to be subjective. Yang et al. [18] used a Gaussian mixture distribution model to fit the speed distribution of trucks and obtained similar distribution characteristics to select a proper threshold in four cities. For the latter part, the methods include stop threshold setting [19,20,21] and stop time-based one-dimensional clustering [22,23,24,25,26].
Travel characteristics can be recognized based on the identification of trip ends. Previous studies have comprehensively investigated travel characteristics by various methods from temporal and spatial aspects. Yang et al. [27] estimated intercity heavy truck mobility flows using a deep gravity framework. Siripirote et al. [28] proposed a statistical approach to estimate truck activities. The negative matrix factorization method (NMF) is proposed and applied to traffic spatial–temporal characteristics analysis, which provides a way to combine temporal and spatial characteristics and catch the correlation [29,30,31,32].
Emission calculation and the relationship between land use and emission intensity have long been research hotspots. Research on carbon emission estimation primarily encompasses top-down and down-top methods. The top-down approach starts from energy consumption to estimate total emissions over large regions, but this method only provides low-resolution emission characteristics with administrative districts as units. Conversely, the bottom-up approach, represented by trajectory data-based calculations, offers a low-cost method for calculating road traffic carbon emissions and has gradually become a new research direction for estimating traffic carbon emissions over large scales [33,34,35,36,37].
Based on the high-resolution emission distribution map, factors that affect emission and their relationship can be estimated. Plenty of studies have investigated the built environment, land use, and their effects on traffic emissions [33,38,39,40,41,42]. Cui et al. [40] incorporated massive vehicle trajectory big data and vehicle type data to calculate all on-road emissions and estimated their relationship with the built environment in Shenzhen. Feng et al. [43] conducted similar research on a provincial level. The results show that different land use types, as well as cross-factors (i.e., job–home balance), significantly impact the emissions. Though such research has been conducted comprehensively, few studies have considered the travel and emission characteristics of HMTTs.
The travel patterns of HMTTs are influenced by time-based restrictions and the need to avoid peak traffic hours, leading to specific travel windows such as early mornings or late nights [44]. These time-based patterns are critical for understanding the efficiency and environmental impact of hazardous materials transport. Despite recognition of these unique travel patterns, research on the operational constraints and travel dynamics of HMTTs remains limited. Additionally, studies related to HMTTs often involve route models that focus on accident prevention and spill mitigation, as accidents involving hazardous materials can have significant environmental consequences [45,46]. These risk factors make HMTT route models more complex and safety-sensitive compared to standard freight operations [2,47]. Some studies have incorporated carbon emissions into HMTT route planning, aiming to minimize both costs and hazard risks while limiting emissions [48]. However, macro-level research on the operational and emission characteristics of HMTTs is still insufficient. In this study, a framework from trip generation based on trajectory data for the analysis of the travel and emissions characteristics is proposed and applied to HMTTs. The travel modes are decomposed and their characteristics are analyzed, and relationships between HMTTs’ emissions and land use are estimated separately.

3. Materials and Methods

This study proposes a trip generation method using trajectory data and applies it to HMTTs. Then, based on the trips of HMTTs, trip and emission characteristics are analyzed relatively. The research framework for this study is shown in Figure 1. The trip generation method consists of two parts: (1) stop identification using a speed threshold given by the saddle point of the Gaussian mixture distribution model, and (2) OD extract by stop time K-medoid clustering. Travel characteristics are recognized as three modes by non-negative matrix factorization. For emission characteristics, firstly, HMTTs’ emissions are estimated considering mileage, speed, environment, and traffic conditions as Equations (3) and (4). Secondly, the relationships between emission and land use are caught by GWR as Equation (5), through which different land use types’ effects on emission are quantified. The methodologies used are introduced in sequence.

3.1. Gaussian Mixture Distribution Model

This method corresponds to the ① step in Figure 1, where a Gaussian Mixture Model (GMM) is used to fit the velocity distribution in the trajectory data to determine the threshold velocity separating driving points from drifting points. Among all locating points of HMTTs, the distribution of speed is composed of a combination of drifting points with lower speed and driving points with higher speed. The speed threshold for identifying vehicle stop points can be determined by identifying the breakpoint between the two distributions.
The Gaussian mixture distribution model is used to determine the threshold of stop speed to avoid the inaccuracy caused by manual setting. In this distribution, the minimum value between the two peaks is referred to as the stationary point of the distribution function, which is a critical feature for distinguishing and identifying the two components of the mixed distribution and marks the boundary between the two distribution intervals. Therefore, the speed corresponding to the stationary point can be used as the threshold to distinguish between a static drift state and a normal driving state. When the vehicle speed is below this threshold, the vehicle is considered to be in a stationary state. The Gaussian mixture distribution is shown in Equation (1):
f x = ω 1 1 2 π σ 1 exp ⁡ − x − x 1 2 2 σ 1 2 + ω 2 1 2 π σ 2 exp ⁡ − x − x 2 2 2 σ 2 2
where x denotes the logarithm of speed to avoid losing information of low-speed interval since drifting points concentrate on this interval; ω 1 , ω 2 represent weights of the two Gaussian distributions; x 1 , x 2 represents the means of the two Gaussian distributions (mean of the logarithm of drifting and driving speed); and σ 1 , σ 2 represent standard deviations of the two Gaussian distributions.

3.2. K-Medoid Clustering

The K-medoid Clustering method corresponds to the ② step in Figure 1. After the identification of stop points, it is essential to determine which of them are OD points and which are not. Stop time is a common index since HMTTs need time to load and unload. The K-Medoids clustering algorithm is used to distinguish OD stops from other stops by stop time clustering. Similar to the K-Means clustering algorithm, it weakens the influence of abnormal outliers to a certain extent and has stronger robustness. The objective function K-Medoids clustering algorithm is shown in Equation (2):
J = ∑ i = 1 k ∑ t ∈ C i d t ,   m i
where k is the number of clusters; C i denotes the set of data points assigned to the i cluster; t denotes the stop time of each stop point and m i is the medoid point of the i cluster; the distance d t ,   m i is calculated by Euclidean distance. The K-Medoids clustering algorithm minimizes the objective function J iteratively, aiming to reduce intra-cluster variance. In the context of this study, this corresponds to minimizing the difference in stop time within the OD stop point cluster and the non-OD stop point cluster.

3.3. Non-Negative Matrix Factorization

Matrix Factorization methods correspond to the ③ step in Figure 1, which is widely used in the study of travel mode decomposing in the transportation field. By decomposing a travel spatiotemporal matrix, it can be broken down into multiple matrices composed of simple patterns. Travel mode information can be represented with several recapitulative and focused spatiotemporal patterns. For any given non-negative matrix M m × n (Original-Destination Matrix, with m and n denote the number of origins and destinations relatively), the Non-negative Matrix Factorization (NMF) algorithm can find two non-negative matrices W and V that satisfy Equation (3), which are more in line with the actual situation of travel.
M m × n ≈ W m × r V r × n = ∑ i = 1 r w i v i r
This algorithm decomposes the original non-negative matrix into two non-negative matrices. Where W is the basis matrix, with each column representing a latent travel pattern; V is the coefficient matrix, indicating the contribution of each latent pattern to specific destinations; r is a critical super parameter that denotes the number of latent travel patterns. Each ω i v i T can be considered as a “travel pattern” and represents a distinct latent feature or travel pattern. The decomposition tends to minimize redundancy so the travel patterns can capture unique features in the data with minimal overlap. All ω i v i T form a complementary decomposition of the original Original-Destination Matrix. These components capture different aspects of the travel behavior, ensuring that the reconstructed matrix provides a comprehensive and more focused representation of the observed data.

3.4. Trip Carbon Emissions Estimation

Carbon emissions of HMTTs are estimated considering mileage, speed, environment, and traffic conditions based on the identified trips. High-resolution emissions maps as precise as 0.005° × 0.005° are drawn to capture the characteristics and accurately estimate HMTTs’ emissions. Carbon emissions of each trip are systematically calculated as Equation (4):
E i = L i × E F i ( v ) × 10 − 3
where E i denotes the carbon emissions of trip i , L i denotes the distance of trip i , and E F i ( v ) denotes the emission factor under v speed and can be calculated as Equation (5):
E F i v = B E F i × φ i × γ i
where B E F i denotes the base emission factor of the vehicle’s type of trip i ; φ i and γ i represent the correction factors for environmental and traffic conditions, respectively. These factors are calibrated according to the actual situation and vehicle characteristics.

3.5. Geographically Weighted Regression

After the calculation of HMTTs’ emissions, we investigate the effect factors of HMTTs’ emissions and the difference with other on-road vehicles. The relationship between land use and HMTTs’ emissions is modeled from a spatial aspect. In spatial statistics, the coefficients obtained from a multiple linear regression model do not take into account the spatial relationships between samples. In contrast, in a GWR model, the regression coefficients for the same independent variable are not fixed values, but instead change with geographic zones as in Equation (6) [49,50], according to which the relationship between land use and HMTTs’ emissions can be estimated more precisely.
y i = β i 0 + ∑ k = 1 p β i k x i k + ε i
where y i is the estimated emissions of zone i , p denotes the number of independent variables (land use types), x i k denotes the land use types k for geographic i ; and β i k denotes the regression coefficient of the geographic zone i to the independent variable k .

4. Data Description and Data Preprocessing

The data used in this study are from the BeiDou Satellite Navigation System (BDS) data of hazardous materials transport vehicles in Shanghai from 1–14 November 2021. This dataset includes about 7800 road hazardous materials transport vehicles. Each record represents a successful upload of information by the Beidou vehicle-mounted terminal, with a data sampling interval of about 30 s. The dataset includes data fields such as vehicle license plate number, positioning time, vehicle latitude, vehicle longitude, vehicle travel azimuth, and vehicle transport industry code, among other attributes. This study conducts an analysis of the original data and identifies four major categories of data issues (i.e., data redundancy, data anomalies, data drift, and invalid travel data).
For each category, specific data-cleaning principles have been established to address these issues effectively. First, the data contains a large number of duplicate entries, where fields such as the positioning time and latitude/longitude for the same vehicle are exactly the same. This situation occurs because the vehicle’s data collection terminal uploads the same record multiple times during data transmission. In the processing stage, duplicate data needs to be removed, keeping only the first entry from the duplicates. Second, data drift refers to sudden and significant abnormal fluctuations in vehicle positioning data over a short period of time. Even if a vehicle is traveling at its maximum speed, significant position changes are unlikely to occur in a short time when the vehicle’s positioning data sampling frequency is stable. To address the issue of data drift, it can be detected and resolved by calculating the changes in position and speed between two trajectory points. We define two consecutive trajectory points in the vehicle’s trajectory as P i and P i + 1 . The distance Δ s between the two points can be calculated using their latitude and longitude coordinates. Then, using the time difference Δ t between the two points, the average speed v between the two points can be computed. If v > v m a x , delete point P i + 1 , and proceed to calculate the average speed between points P i and P i + 2 . If v ≤ v m a x , calculate the average speed between points P i + 1 and P i + 2 , and the speed limit for large buses and freight vehicles, v m a x , is set to 100 km/h. Third, in cases where duplicate vehicle data points with minor differences in latitude and longitude are collected at the same time, only the first record is retained to maintain analysis integrity. Additionally, discrepancies between recorded speeds and unchanged positions, or zero speeds with changing positions, are corrected by recalculating speed. Fourth, travel typically requires minimum time and distance; this study classifies trips shorter than 10 min or under 500 m as invalid and excludes them from the data. Data description and samples are shown in Table 1.

5. Results

5.1. Trip Generation from Trajectory Data

The distribution of average vehicle speeds reveals a bimodal pattern: the left side corresponds to low speeds characteristic of static drift conditions, and the right side corresponds to normal speeds observed during driving conditions according to Figure 2a. The horizontal axis represents the logarithm of speed to ensure the retention of a substantial amount of data in the low-speed part. This indicates that the overall distribution is a composite of these two distinct states. The saddle point that separates the two sides is located at 0.96 km/h, so this speed is taken as the speed threshold to distinguish drift points from driving points. Change of direction is used to evaluate the performance of this threshold since drifting and driving points behave differently. Figure 2b shows the comparison of static drift points and moving points’ direction change, which shows a significant difference and suggests the reasonability of the threshold chosen. After identifying stop points by speed threshold, clustering is performed based on the duration of stops to obtain OD points. The travel OD distribution is shown in Figure 3.

5.2. Travel Characteristics of HMTTs

5.2.1. Temporal and Spatial Travel Characteristics

The travel distances of HMTTs exhibit minimal variation between weekdays and weekends (Figure 4a). The majority of these vehicles travel distances of less than 40 km. As the distance increases, there is a noticeable decline in the relative frequency of trips. In terms of travel time characteristics (Figure 4b), there is a pronounced tendency for HMTTs to undertake short trips lasting between 0–3 h on both weekdays and weekends. The minimal variation in travel distances between weekdays and weekends suggests a consistent demand for hazardous material transportation, likely driven by industrial and logistical requirements that remain stable throughout the week.
The temporal distribution of departures and arrivals for HMTTs is extensive (Figure 5). With the exception of there being fewer activities before dawn hours, departures and arrivals are observed throughout most of the day. It is noted that vehicles typically arrive within 1–3 h post-departure, aligning with the observed distribution of travel durations. This pattern can be attributable to the unique requirements of hazardous material transportation, which necessitates regular and timely deliveries, thereby lacking distinct peak times for departures and arrivals. Furthermore, due to Shanghai’s traffic control measurement on HMTTs, logistics companies may schedule nighttime transportation for these vehicles, leading to a notable number of departures and arrivals during night hours.

5.2.2. Spatiotemporal Pattern Analysis of Travel Analysis by NMF

To better identify the travel characteristics of HMTTs, their overall spatiotemporal distribution trends can be viewed as a composition of several distinct travel modes, each defined by unique and significant features. By identifying and decomposing these travel modes, more targeted and effective management strategies can be developed. The NMF method provides a powerful tool to facilitate this process. It is essential to determine the number of travel modes (i.e., the value of r ) before applying NMF to split the travel modes. Four scenarios are examined with r values set to 2, 3, 4, and 5, respectively, and the matrix decomposes best when r is set to 3. An excessively large r value results in an excess of travel modes but with minimal differences between them, which makes it difficult to explain each travel mode, while the small r value leads to incomplete identification of travel modes. Following this principle, travel modes of HMTTs are divided into three types accordingly (i.e., (1) Daytime—Surrounding cities, (2) Early morning—In-city, and (3) Midnight—Scattered).
Mode 1 has the most travel demand, and encompasses the majority of the daytime from the perspective of temporal distribution, and is mainly consistent with standard working hours (Figure 6). In addition to industrial hotspot areas within the city, there is a significant volume of cross-province travel towards other provinces (Zhejiang and Jiangsu) from the perspective of spatial distribution. This reflects a normative demand for daytime hazardous material transportation. Within this travel demand pattern, the daytime demand remains relatively stable, with a slight increase observed during the early evening peak hours and a decrease after 18:00. Since cross-province travels have extended distances and increased time requirements, companies may typically allocate substantial daytime hours for these trips.
Mode 2 signifies the morning peak travel demand for HMTTs in terms of temporal distribution analysis (Figure 7). In comparison to mode 1, the morning peak period exhibits a notably higher demand, with the peak concentrated between 5:00 and 7:00 a.m., which is earlier than the regular commuting morning peak. The spatial distribution corresponding to this temporal distribution tends to be more concentrated within the city, such as in the Chongming District in Shanghai. Compared with cross-province travel, HMTTs’ travels in Shanghai are under more strict traffic control. It is reasonable for HMTTs to schedule more concentrated transport during the early morning to adapt to urban traffic management and avoid disrupting regular commuting morning peak.
Mode 3 has the least travel demand compared with mode 1 and mode 2, which mainly distributes during the night from the temporal perspective (Figure 8). Due to the cessation of regular production activities and the inconveniences of night on-road transportation, travel demand for HMTTs decreases consequentially.

5.3. Emission Characteristics of HMTTs

5.3.1. Temporal–Spatial Emission Characteristics

To explore the temporal–spatial distribution of HMTTs’ emissions, the average emission amount for each hour and each grid is calculated and drawn separately (Figure 9). HMTTs have a high emission level between 6 a.m. and 6 p.m. and little fluctuation. This non-peak mode aligns with the intensity of travel and differs from the findings of previous studies on private cars [51,52,53], which typically exhibit distinct morning and evening peaks in emissions since they have distinct peaks in travel. However, HMTTs do not exhibit such pronounced peaks both in travel and emissions due to their specific transportation needs.
The spatial characteristics of HMTTs’ emission are shown in Figure 9b, with highways, expressways, and some main roads having the highest emission levels. Comparing traffic flows and average speed (Figure 10), areas where the traffic flows are high and average speeds are low, which represent more travel and more congested traffic, tend to have higher emission levels. However, it is worth noting that although the traffic flow is large in some areas (e.g., middle ring, outer ring, and some cross-province highways), the emission intensity is not high due to the high average road speed, which means good traffic conditions. In contrast, the emission level on the inner ring road is relatively high due to the routine traffic congestion. These observations are aligned with previous research analyzing the emission characteristics of private cars [54,55], specifically indicating that high-emission road networks extend from the city center to the suburbs. For heavy trucks like HMTTs, it is also not accurate to evaluate the environmental impact simply from the number of vehicles. Emission intensity often has a close relationship with traffic conditions and vehicle speed, and the emissions of vehicles at steady high speeds are usually lower than those at low speeds or frequent stop and start.

5.3.2. Relationship Between Emission and Land Use Analysis by GWR

Extensive research has been conducted to examine the impact of land use on emissions. However, few of them consider HMTTs separately. The GWR model is a commonly used geographical model that differs from Ordinary Least Squares (OLS) by providing estimated parameters that consider spatial heterogeneity [56,57]. The GWR model in this study analyzed the relationship between HMTTs’ emissions and land use, which is quantified by the number of certain types of POIs obtained from a map. Middle traffic analysis zones are divided according to the street division subdistricts, which reflect the environment and land use of the real world, and the zones are divided more precisely in downtown areas and reversed in the suburban areas. The descriptive statistics of POI numbers in different traffic analysis zones are shown in Table 2.
The statistics and parameter estimates are shown in Table 3, including the mean (Mean), standard deviation (Std), minimum (Min), median (Mid), and maximum (Max) of the estimated parameter for each variable. An adaptive bandwidth approach is employed in the GWR model. Estimated parameters of all variables have a proportion of significant sample points exceeding 80% at the 5% significance level. More detailed parameter comparisons are shown in Figure 11. We also conducted a multiple linear regression (MLR) using the ordinary least squares (OLS) method. The results indicate that GWR outperformed MLR across evaluation metrics such as R 2 and Akaike Information Criterion (AIC), as shown in Table 4.
Industrial and company-related POIs have a strong positive effect on HMTT emissions in almost all zones, as shown in Figure 11a,b. This indicates that areas with a higher density of these POIs are associated with higher emissions from HMTTs. This finding differs from previous studies on emissions from all on-road vehicles, where company-related POIs were found to have a negative correlation with emissions, and industrial-related POIs exhibited a smaller impact [58,59,60,61].
Figure 11c demonstrates that entertainment-related POIs have a negative correlation with HMTT emissions across all zones. This suggests that areas with more entertainment-related POIs tend to experience lower HMTT emissions. This result contrasts with findings from studies on private car emissions, where entertainment-related POIs were identified as attractors for private car trips, leading to increased emissions [58]. The geographical distribution of coefficients shows that the influence of entertainment-related POIs on HMTT emissions is stronger in western zones compared to eastern zones (primarily the Pudong district). This suggests spatial heterogeneity and possible cross-influences between different types of land use.
Education-related POIs, as illustrated in Figure 11d, exhibit minimal impact on HMTT emissions. This finding differs from previous studies on all on-road vehicle emissions, where education-related POIs were found to have a significant positive effect [53,58].

6. Discussion

The spatiotemporal travel characteristics of HMTTs in the Shanghai metropolitan area exhibit notable differences from other types of vehicles. These differences include: (1) minimal variation between weekdays and weekends, as well as (2) the absence of peak periods throughout the day, except for a significant reduction in travel during the early morning hours (1:00–4:00 a.m.).
Based on these patterns, HMTT trips can be categorized into three primary modes: Daytime—Surrounding cities, Early morning—In-city, and Midnight—Scattered. These modes are likely influenced by Shanghai’s regulatory restrictions on large vehicles during specific hours and the unique demands of hazardous material transportation. For instance, daytime restrictions on large vehicles in urban areas may compel HMTTs to commence trips earlier or operate during midnight hours. While this scheduling helps avoid conflicts with regular traffic and enhances overall road safety, it also introduces new challenges. Nighttime and early morning travel periods are associated with poor visibility and increased driver fatigue, elevating the risks for HMTT operations, to which more attention should be paid.
The strong positive correlation between industrial and company-related POIs and HMTT emissions aligns with the functional roles of these areas. Many chemical and industrial companies rely heavily on HMTTs for the transportation of hazardous materials, making these land uses critical hubs for HMTT activity and consequently higher emissions. This insight underscores the necessity of incorporating emissions mitigation strategies in industrial zones. As for entertainment-related POIs, the negative correlation between entertainment-related POIs and HMTT emissions highlights the unique travel demands characteristic of HMTTs. Unlike private cars, HMTTs rarely serve entertainment-related trips, which explains the observed decrease in emissions in areas with higher entertainment POI densities. The influence also demonstrates significant geographical disparities. Coefficients of entertainment-related POIs density on HMTTs’ emissions are greater for the western zones than for the eastern ones (mainly Pudong district). This may suggest the presence of cross-influences between different types of land use. In the predominantly industrial area of Pudong, the effect of entertainment-related POIs on emissions is relatively less significant compared to the non-industrial southwestern regions.
Education-related POIs are also considered in this research but show little effect on the HMTTs’ emissions. The reasons for this phenomenon can be two-fold. Firstly, education-related POIs such as schools attract trips by taxis and private cars inherently but have little traffic demand and trip generation of HMTTs. Besides, it may also indicate that the HMTTs’ administrators seem not to take educational spots into consideration when they plan the routes of HMTTs. Making route plans for HMTTs and taking vulnerable educational sites into consideration should be added to future government works.
Based on the current results, some suggestions are put forward accordingly: (1) In terms of the establishment of new facilities, it is essential to thoroughly assess the various types of traffic movements they generate and the impact of generated trips on adjacent areas. Specifically, for facilities such as chemical or industrial developments, careful consideration of potential hazardous material transport is necessary, which involves integrating transport routes and scheduling to mitigate impacts on adjacent areas; (2) Additionally, the existing traffic generation and transit patterns of one area should be accounted for in the planning of new developments and construction. For instance, it is not recommended to establish educational or other sensitive land uses in areas with a high density of hazardous material transportation trucks’ movements, not only high density of chemical or industrial POIs. (3) Lastly, in traffic route planning and navigation, special vehicles such as those transporting hazardous materials should carefully consider the impact these vehicles may have on the areas they pass through, and if necessary, reduce their exposure in densely populated areas such as entertainment and educational POIs.

7. Conclusions

This study developed a trajectory data-based framework for the analysis of HMTTs travel and emission characteristics and the exploration of how land use affects HMTTs’ emissions. In this framework, the vehicle stop points are first identified using a speed threshold determined by the saddle point of a Gaussian mixture distribution model. Subsequently, OD is identified through stop time clustering, and travel characteristics are classified into three modes using non-negative matrix factorization. Next, HMTTs’ emissions are estimated by considering mileage, speed, environment, and traffic conditions. Finally, GWR is employed to capture the relationship between emissions and land use. The results reveal that: (1) HMTT travel in Shanghai can be categorized into three modes: Daytime—Surrounding cities, Early morning—In-city, and Midnight—Scattered, each with distinct spatiotemporal characteristics, and similar travel mode decomposing can be done in other cities; (2) HMTTs’ emissions show a positive relationship with travel density and a negative one with travel speed; (3) Industrial-related POIs and company-related POIs exhibit a significant positive impact on HMTTs’ emissions; (4) Unlike private cars, entertainment-related POIs have a minimal negative effect on HMTTs’ emissions, while educational sites show a minimal impact. These findings can help planners develop more targeted traffic and environmental policies and provide crucial insights for optimizing HMTTs’ emissions.
However, this study still has limitations in the following aspects: (1) Due to data limitations, this study focuses only on the analysis of HMTTs in Shanghai. Future research could conduct similar studies in different cities to identify the characteristics and variations of HMTTs transportation; and (2) The number of different POI types is utilized to represent the land use, which can be extended in the future with the four Ds (socio-demographic density, land use diversity, urban design, and distance to the urban center) [57,58,59,60,61,62,63,64,65,66], which are often used to evaluate the built environment, for a more in-depth analysis. (3) This study focuses on the macro-level relationship between land use types and HMTT emissions in urban traffic zones. However, factors like vehicle types, cargo types, and road gradients, which are important in route planning, were not addressed. Future research could explore these factors to offer a more comprehensive understanding of HMTT emissions.

Author Contributions

Conceptualization, Y.Z. and W.H.; methodology, Y.Z. and Q.H.; software, Q.H.; validation, W.H., Y.Z. and S.Z.; formal analysis, W.H.; investigation, Q.H.; resources, Q.H.; data curation, Y.Z.; writing—original draft preparation, Y.Z., W.H. and Q.H.; writing—review and editing, Y.Z. and Y.C.; visualization, Q.H.; supervision, Y.Z.; project administration, Y.Z.; funding acquisition, Y.Z. and S.Z.; All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (Grant No.52472351), Shanghai Science and Technology Committee (Grant No. 24170742100) and the Fundamental Research Funds for the Central Universities (Grant No. 22120230310).

Data Availability Statement

The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Moghaddam, K.S.; Azadian, F. Chance-constrained multi-objective approach for hazardous materials routing and scheduling under demand and service time uncertainty. J. Multi-Criteria Decis. Anal. 2020, 27, 318–336. [Google Scholar] [CrossRef]
  2. Erkut, E.; Tjandra, S.A.; Verter, V. Hazardous materials transportation. Handb. Oper. Res. Manag. Sci. 2007, 14, 539–621. [Google Scholar]
  3. Holeczek, N. Hazardous materials truck transportation problems: A classification and state of the art literature review. Transp. Res. Part D Transp. Environ. 2019, 69, 305–328. [Google Scholar] [CrossRef]
  4. Lomotko, D.; Ohar, O.; Kozodoi, D.; Barbashyn, V.; Lomotko, M. Efficiency of “green” logistics technologies in multimodal transportation of dangerous goods. In Proceedings of the International Conference on Smart Technologies in Urban Engineering (STUE-2022), Kharkiv, Ukraine, 9–11 June 2022; Springer International Publishing: Cham, Switzerland; pp. 831–841. [Google Scholar]
  5. Covello, V.T.; Merkhoher, M.W. Risk Assessment Methods: Approaches for Assessing Health and Environmental Risks; Plenum Press: New York, NY, USA, 1993. [Google Scholar]
  6. Ziaei, Z.; Jabbarzadeh, A. A multi-objective robust optimization approach for green location-routing planning of multi-modal transportation systems under uncertainty. J. Clean. Prod. 2021, 291, 125293. [Google Scholar] [CrossRef]
  7. Sun, M.; Zhou, R.; Jiao, C. Exploring HAZMAT Truck Risk Factors with Driver Monitoring Systems Data Using Random Parameter Ordered Probit Model. In Proceedings of the 22nd COTA International Conference of Transportation Professionals (CICTP 2022), Changsha, China, 8–11 July 2022; pp. 1704–1714. [Google Scholar]
  8. Sun, M.; Zhou, R. Investigation on Hazardous Material Truck Involved Fatal Crashes Using Cluster Correspondence Analysis. Sustainability 2023, 15, 9369. [Google Scholar] [CrossRef]
  9. Hong, J.; Tamakloe, R.; Park, D. Application of association rules mining algorithm for hazardous materials transportation crashes on expressway. Accid. Anal. Prev. 2020, 142, 105497. [Google Scholar] [CrossRef]
  10. Iranitalab, A.; Khattak, A.; Bahouth, G. Statistical modeling of cargo tank truck crashes: Rollover and release of hazardous materials. J. Saf. Res. 2020, 74, 71–79. [Google Scholar] [CrossRef]
  11. Weng, J.; Gan, X.; Zhang, Z. A quantitative risk assessment model for evaluating hazmat transportation accident risk. Saf. Sci. 2021, 137, 105198. [Google Scholar] [CrossRef]
  12. Kong, D.; Zhao, X.; Chen, J.; Yang, H.; Du, J. Study on hazard characteristics and safety distance of small-scale boilover fire. Int. J. Therm. Sci. 2021, 164, 106888. [Google Scholar] [CrossRef]
  13. Chen, J.; Pei, T.; Li, M.; Song, C.; Ma, T.; Lu, F.; Shaw, S.-L. An enhanced model for evacuation vulnerability assessment in urban areas. Comput. Environ. Urban Syst. 2020, 84, 101540. [Google Scholar] [CrossRef]
  14. Yilmaz, O.; Kara, B.Y.; Yetis, U. Hazardous waste management system design under population and environmental impact considerations. J. Environ. Manag. 2017, 203, 720–731. [Google Scholar] [CrossRef] [PubMed]
  15. Yao, Z.; Yang, F.; Guo, Y.; Jin, P.J.; Li, Y. Trip end identification based on spatial-temporal clustering algorithm using smartphone positioning data. Expert Syst. Appl. 2022, 197, 116734. [Google Scholar] [CrossRef]
  16. Diana, M.; Pirra, M.; Woodcock, A. Freight distribution in urban areas: A method to select the most important loading and unloading areas and a survey tool to investigate related demand patterns. Eur. Transp. Res. Rev. 2020, 12, 40. [Google Scholar] [CrossRef]
  17. Sun, D.; Leurent, F.; Xie, X. Mining vehicle trajectories to discover individual significant places: Case study using floating car data in the Paris region. Transp. Res. Rec. 2021, 2675, 1–9. [Google Scholar] [CrossRef]
  18. Yang, Y.; Jia, B.; Yan, X.-Y.; Li, J.; Yang, Z.; Gao, Z. Identifying intercity freight trip ends of heavy trucks from GPS data. Transp. Res. Part E Logist. Transp. Rev. 2022, 157, 102590. [Google Scholar] [CrossRef]
  19. Hadavi, S.; Verlinde, S.; Verbeke, W.; Macharis, C.; Guns, T. Monitoring urban-freight transport based on GPS trajectories of heavy-goods vehicles. IEEE Trans. Intell. Transp. Syst. 2018, 20, 3747–3758. [Google Scholar] [CrossRef]
  20. Yang, Y.; Jia, B.; Yan, X.-Y.; Chen, Y.; Tavasszy, L.; de Bok, M.; Bai, Z.; Liu, E.; Gao, Z. Structure and dynamics of urban freight truck movements: A complex network theory perspective. Transp. Res. Part C Emerg. Technol. 2024, 158, 104442. [Google Scholar] [CrossRef]
  21. Zhao, Y.; Cheng, S.; Lu, F. Seasonal characteristics of agricultural product circulation network: A case study in Beijing, China. Agronomy 2022, 12, 2827. [Google Scholar] [CrossRef]
  22. Sharman, B.W.; Roorda, M.J. Analysis of freight global positioning system data: Clustering approach for identifying trip destinations. Transp. Res. Rec. 2011, 2246, 83–91. [Google Scholar] [CrossRef]
  23. Laranjeiro, P.F.; Merchán, D.; Godoy, L.A.; Giannotti, M.; Yoshizaki, H.T.; Winkenbach, M.; Cunha, C.B. Using GPS data to explore speed patterns and temporal fluctuations in urban logistics: The case of São Paulo, Brazil. J. Transp. Geogr. 2019, 76, 114–129. [Google Scholar] [CrossRef]
  24. Mesa-Arango, R.; Ukkusuri, S.V. Demand clustering in freight logistics networks. Transp. Res. Part E Logist. Transp. Rev. 2015, 81, 36–51. [Google Scholar] [CrossRef]
  25. Gingerich, K.; Maoh, H.; Anderson, W. Classifying the purpose of stopped truck events: An application of entropy to GPS data. Transp. Res. Part C Emerg. Technol. 2016, 64, 17–27. [Google Scholar] [CrossRef]
  26. Kinjarapu, A.; Demissie, M.G.; Kattan, L.; Duckworth, R. Applications of passive GPS data to characterize the movement of freight trucks—A case study in the Calgary region of Canada. IEEE Trans. Intell. Transp. Syst. 2021, 23, 9210–9225. [Google Scholar] [CrossRef]
  27. Yang, Y.; Jia, B.; Yan, X.-Y.; Chen, Y.; Song, D.; Zhi, D.; Wang, Y.; Gao, Z. Estimating intercity heavy truck mobility flows using the deep gravity framework. Transp. Res. Part E Logist. Transp. Rev. 2023, 179, 103320. [Google Scholar] [CrossRef]
  28. Siripirote, T.; Sumalee, A.; Ho, H. Statistical estimation of freight activity analytics from Global Positioning System data of trucks. Transp. Res. Part E Logist. Transp. Rev. 2020, 140, 101986. [Google Scholar] [CrossRef]
  29. Lee, D.D.; Seung, H.S. Learning the parts of objects by non-negative matrix factorization. Nature 1999, 401, 788–791. [Google Scholar] [CrossRef]
  30. Yong, N.; Ni, S.; Shen, S.; Chen, P.; Ji, X. Uncovering stable and occasional human mobility patterns: A case study of the Beijing subway. Phys. A Stat. Mech. Its Appl. 2018, 492, 28–38. [Google Scholar] [CrossRef]
  31. Cazabet, R.; Jensen, P.; Borgnat, P. Tracking the evolution of temporal patterns of usage in bicycle-sharing systems using nonnegative matrix factorization on multiple sliding windows. Int. J. Urban Sci. 2018, 22, 147–161. [Google Scholar] [CrossRef]
  32. Maeda, T.N.; Shiode, N.; Zhong, C.; Mori, J.; Sakimoto, T. Detecting and understanding urban changes through decomposing the numbers of visitors’ arrivals using human mobility data. J. Big Data 2019, 6, 4. [Google Scholar] [CrossRef]
  33. Cheng, S.; Zhang, B.; Peng, P.; Yang, Z.; Lu, F. Spatiotemporal evolution pattern detection for heavy-duty diesel truck emissions using trajectory mining: A case study of Tianjin, China. J. Clean. Prod. 2020, 244, 118654. [Google Scholar] [CrossRef]
  34. Lv, Y.-Y.; Yan, X.-Y.; Jia, B.; Yang, Y.; Liu, E. Quantifying the overall spatial distribution characteristics of urban heavy truck trips: The case of China. J. Transp. Geogr. 2024, 115, 103822. [Google Scholar] [CrossRef]
  35. Cheng, S.; Lu, F.; Peng, P. A high-resolution emissions inventory and its spatiotemporal pattern variations for heavy-duty diesel trucks in Beijing, China. J. Clean. Prod. 2020, 250, 119445. [Google Scholar] [CrossRef]
  36. Ağbulut, Ü. Forecasting of transportation-related energy demand and CO2 emissions in Turkey with different machine learning algorithms. Sustain. Prod. Consum. 2022, 29, 141–157. [Google Scholar] [CrossRef]
  37. Deng, F.; Lv, Z.; Qi, L.; Wang, X.; Shi, M.; Liu, H. A big data approach to improving the vehicle emission inventory in China. Nat. Commun. 2020, 11, 2801. [Google Scholar] [CrossRef]
  38. Lin, B.; Wang, M. The role of socio-economic factors in China’s CO2 emissions from production activities. Sustain. Prod. Consum. 2021, 27, 217–227. [Google Scholar] [CrossRef]
  39. Liu, M.; Zhang, X.; Zhang, M.; Feng, Y.; Liu, Y.; Wen, J.; Liu, L. Influencing factors of carbon emissions in transportation industry based on CD function and LMDI decomposition model: China as an example. Environ. Impact Assess. Rev. 2021, 90, 106623. [Google Scholar] [CrossRef]
  40. Cui, X.; Zhuang, C.; Jiao, Z.; Tan, Z.; Li, S. How can urban built environment (BE) influence on-road (OR) carbon emissions? A road segment scale quantification based on massive vehicle trajectory big data. J. Transp. Geogr. 2023, 111, 103669. [Google Scholar] [CrossRef]
  41. Wang, G.; Feng, Y. Analysis of carbon emission drivers and peak carbon forecasts for island economies. Ecol. Model. 2024, 489, 110611. [Google Scholar] [CrossRef]
  42. Solarin, S.A.; Al-Mulali, U. Transportation fuel subsidies and CO2 emissions: The roles of economic complexity and uncertainty. Environ. Prog. Sustain. Energy 2022, 41, e13797. [Google Scholar]
  43. Feng, C.; Xia, Y.-S.; Sun, L.-X. Structural and social-economic determinants of China’s transport low-carbon development under the background of aging and industrial migration. Environ. Res. 2020, 188, 109701. [Google Scholar] [CrossRef]
  44. Park, H.-S.; Jun, C.-H. A simple and fast algorithm for K-medoids clustering. Expert Syst. Appl. 2009, 36, 3336–3341. [Google Scholar] [CrossRef]
  45. Inanloo, B.; Tansel, B. A transportation network assessment tool for hazardous material cargo routing: Weighing exposure health risks, proximity to vulnerable areas, delay costs, and trucking expenses. J. Loss Prev. Process Ind. 2016, 40, 266–276. [Google Scholar] [CrossRef]
  46. Bula, G.A.; Afsar, H.M.; González, F.A.; Prodhon, C.; Velasco, N. Bi-objective vehicle routing problem for hazardous materials transportation. J. Clean. Prod. 2019, 206, 976–986. [Google Scholar] [CrossRef]
  47. Zou, Y. Modeling highly dispersed crash data with Sichel GAMLSS: An alternative approach to traditional methods. Multidiscip. Sci. J. 2024. accepted. [Google Scholar] [CrossRef]
  48. Yuan, W.; Niu, X.; Yan, B.; Liang, M.; Wang, Q.; Wu, J.; Santibanez Gonzalez, E.D. A novel variant of hazardous materials vehicle routing problem considering carbon emission and its hybrid heuristic algorithm. SSRN 2024, 5064415. [Google Scholar] [CrossRef]
  49. Brunsdon, C.; Fotheringham, A.S.; Charlton, M.E. Geographically weighted regression: A method for exploring spatial nonstationarity. Geogr. Anal. 1996, 28, 281–298. [Google Scholar] [CrossRef]
  50. Fotheringham, A.S.; Brunsdon, C.; Charlton, M.E. Geographically Weighted Regression: The Analysis of Spatially Varying Relationships; Wiley: Hoboken, NJ, USA, 2002. [Google Scholar]
  51. Wen, Y.; Wu, R.; Zhou, Z.; Zhang, S.; Yang, S.; Wallington, T.J.; Shen, W.; Tan, Q.; Deng, Y.; Wu, Y. A data-driven method of traffic emissions mapping with land use random forest models. Appl. Energy 2022, 305, 117916. [Google Scholar] [CrossRef]
  52. Liu, X.; Liu, X.C.; Liu, Z.; Shi, R.; Ma, X. A solar-powered bus charging infrastructure location problem under charging service degradation. Transp. Res. Part D Transp. Environ. 2023, 119, 103770. [Google Scholar] [CrossRef]
  53. Liu, B.; Li, F.; Hou, Y.; Biancardo, S.A.; Ma, X. Unveiling built environment impacts on traffic CO2 emissions using Geo-CNN weighted regression. Transp. Res. Part D Transp. Environ. 2024, 132, 104266. [Google Scholar] [CrossRef]
  54. Oanh, T.T.K.; Ha, N.T.H. Impact of income inequality on climate change in Asia: The role of human capital. Humanit. Soc. Sci. Commun. 2023, 10, 461. [Google Scholar] [CrossRef]
  55. Wu, J.; Jia, P.; Feng, T.; Li, H.; Kuang, H.; Zhang, J. Uncovering the spatiotemporal impacts of built environment on traffic carbon emissions using multi-source big data. Land Use Policy 2023, 129, 106621. [Google Scholar] [CrossRef]
  56. Dai, Z.; Wu, S.; Wang, Y.; Zhou, H.; Zhang, F.; Huang, B.; Du, Z. Geographically convolutional neural network weighted regression: A method for modeling spatially non-stationary relationships based on a global spatial proximity grid. Int. J. Geogr. Inf. Sci. 2022, 36, 2248–2269. [Google Scholar] [CrossRef]
  57. Siqi, J.; Yuhong, W.; Ling, C.; Xiaowen, B. A novel approach to estimating urban land surface temperature by the combination of geographically weighted regression and deep neural network models. Urban Clim. 2023, 47, 101390. [Google Scholar] [CrossRef]
  58. Zhao, C.; Tang, J.; Zeng, Y.; Li, Z.; Gao, F. Understanding the spatio-temporally heterogeneous effects of built environment on urban travel emissions. J. Transp. Geogr. 2023, 112, 103689. [Google Scholar] [CrossRef]
  59. Zhou, L.; He, Q.; Tu, W.; Du, J.; Zhang, S.; Li, Q.; Guan, D.; Zhang, X. A heterogeneous streaming vehicle data access model for diverse IoT sensor monitoring network management. IEEE Internet Things J. 2024, 11, 26929–26943. [Google Scholar] [CrossRef]
  60. Zhou, L.; Tu, W.; Wang, C.; Li, Q. A heterogeneous access metamodel for efficient IoT remote sensing observation management: Taking precision agriculture as an example. IEEE Internet Things J. 2021, 9, 8616–8632. [Google Scholar] [CrossRef]
  61. Zhou, L.; Li, Q.; Tu, W.; Wang, C. A heterogeneous key performance indicator metadata model for air quality monitoring in sustainable cities. Environ. Model. Softw. 2021, 136, 104955. [Google Scholar] [CrossRef]
  62. Ding, C.; Cao, X.; Wang, Y. Synergistic effects of the built environment and commuting programs on commute mode choice. Transp. Res. Part A Policy Pract. 2018, 118, 104–118. [Google Scholar] [CrossRef]
  63. Ding, C.; Cao, X.; Wang, Y. Multi-Depot Pickup and Delivery Problem with Resource Sharing. J. Adv. Transp. 2021, 2021, 5182989. [Google Scholar]
  64. Wang, Y.; Jin, Y.; Mishra, S.; Wu, B.; Zou, Y. Integrated travel demand and accessibility model to examine the impact of new infrastructures using travel behavior responses. J. Transp. Eng. Part A Syst. 2022, 148, 05021009. [Google Scholar] [CrossRef]
  65. Zhang, H.; Zou, Y.; Yang, X.; Yang, H. A temporal fusion transformer for short-term freeway traffic speed multistep prediction. Neurocomputing 2022, 500, 329–340. [Google Scholar] [CrossRef]
  66. Peng, Y.; Liu, D.; Wu, S.; Yang, X.; Wang, Y.; Zou, Y. Investigating traffic operation under freeway mixed traffic flow conditioned on platoon control embedding dedicated lane management. Sensors, 2025; accepted. [Google Scholar]
Remotesensing 17 00423 g001
Figure 1. Overall methodological framework.
Figure 1. Overall methodological framework.
Remotesensing 17 00423 g001
Remotesensing 17 00423 g002
Figure 2. Vehicle speed distribution and fitted curve (a); comparison of static drift points and moving points direction change (b).
Figure 2. Vehicle speed distribution and fitted curve (a); comparison of static drift points and moving points direction change (b).
Remotesensing 17 00423 g002
Remotesensing 17 00423 g003
Figure 3. Trip generation results (OD distribution) of HMTTs in Shanghai.
Figure 3. Trip generation results (OD distribution) of HMTTs in Shanghai.
Remotesensing 17 00423 g003
Remotesensing 17 00423 g004
Figure 4. Travel distance distribution (a) and time distribution (b) of HMTTs.
Figure 4. Travel distance distribution (a) and time distribution (b) of HMTTs.
Remotesensing 17 00423 g004
Remotesensing 17 00423 g005
Figure 5. Distribution of Vehicle Departure and Arrival Times.
Figure 5. Distribution of Vehicle Departure and Arrival Times.
Remotesensing 17 00423 g005
Remotesensing 17 00423 g006
Figure 6. Daytime—Surrounding cities travel mode of HMTTs. Spatial distribution of travel mode 1 is shown as sub-figure (a), and temporal distribution of travel mode 1 is shown as sub-figure (b).
Figure 6. Daytime—Surrounding cities travel mode of HMTTs. Spatial distribution of travel mode 1 is shown as sub-figure (a), and temporal distribution of travel mode 1 is shown as sub-figure (b).
Remotesensing 17 00423 g006
Remotesensing 17 00423 g007
Figure 7. Early morning—In-city travel mode of HMTTs. Spatial distribution of travel mode 2 is shown as sub-figure (a), and temporal distribution of travel mode 2 is shown as sub-figure (b).
Figure 7. Early morning—In-city travel mode of HMTTs. Spatial distribution of travel mode 2 is shown as sub-figure (a), and temporal distribution of travel mode 2 is shown as sub-figure (b).
Remotesensing 17 00423 g007
Remotesensing 17 00423 g008
Figure 8. Midnight—scattered travel mode of HMTTs. Spatial distribution of travel mode 3 is shown as sub-figure (a), and temporal distribution of travel mode 3 is shown as sub-figure (b).
Figure 8. Midnight—scattered travel mode of HMTTs. Spatial distribution of travel mode 3 is shown as sub-figure (a), and temporal distribution of travel mode 3 is shown as sub-figure (b).
Remotesensing 17 00423 g008
Remotesensing 17 00423 g009
Figure 9. Temporal (a) and spatial (b) emission characteristics of HMTTs.
Figure 9. Temporal (a) and spatial (b) emission characteristics of HMTTs.
Remotesensing 17 00423 g009
Remotesensing 17 00423 g010
Figure 10. Comparison of road traffic flow (a), average speed (b), and emission level (c).
Figure 10. Comparison of road traffic flow (a), average speed (b), and emission level (c).
Remotesensing 17 00423 g010
Remotesensing 17 00423 g011
Figure 11. Coefficients of GWR for different land use types: (a) Industrial-related POIs; (b) Company-related POIs; (c) Entertainment-related POIs; (d) Education-related POIs.
Figure 11. Coefficients of GWR for different land use types: (a) Industrial-related POIs; (b) Company-related POIs; (c) Entertainment-related POIs; (d) Education-related POIs.
Remotesensing 17 00423 g011
Table 1. Examples and meanings of BDS data.
Table 1. Examples and meanings of BDS data.
VariablesSampleMeanings
Vehicle IDDZ2345Identical vehicle ID
Direction75Range from 0–359 with 0 for north and incrementing clockwise
Longitude120.894300Accurate to 0.000001 degree
Latitude31.050563Accurate to 0.000001 degree
Timestamp11 November 2021 08:34:41Timestamp accurate to second-degree
Speed86Instantaneous speed (km/h)
Table 2. Descriptive statistics of POI numbers in different traffic analysis zones.
Table 2. Descriptive statistics of POI numbers in different traffic analysis zones.
VariablesMeanStd.Min.Max.
Industrial-related POIs3.929.03088
Company-related POIs370.62437.4702473
Entertainment-related POIs44.6743.640328
Education-related POIs7.959.90077
Table 3. Coefficients summary of GWR for different land use types.
Table 3. Coefficients summary of GWR for different land use types.
VariablesMeanStd.Min.Mid.Max.Band
Width
Industrial-related POIs0.0400.035−0.0100.0330.103427
Company-related POIs0.2360.0020.2310.2360.242427
Entertainment-related POIs−0.2010.017−0.252−0.197−0.178427
Education-related POIs−0.0140.004−0.025−0.013−0.009427
Table 4. Goodness of fit comparison of MLR and GWR.
Table 4. Goodness of fit comparison of MLR and GWR.
Goodness-of-FitMLRGWR
Log-likelihood−412.903−314.362
AIC849.805745.278
R20.6270.760
Adjust R20.6180.725
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zou, Y.; Hu, Q.; Han, W.; Zhang, S.; Chen, Y. Analyzing Travel and Emission Characteristics of Hazardous Material Transportation Trucks Using BeiDou Satellite Navigation System Data. Remote Sens. 2025, 17, 423. https://doi.org/10.3390/rs17030423

AMA Style

Zou Y, Hu Q, Han W, Zhang S, Chen Y. Analyzing Travel and Emission Characteristics of Hazardous Material Transportation Trucks Using BeiDou Satellite Navigation System Data. Remote Sensing. 2025; 17(3):423. https://doi.org/10.3390/rs17030423

Chicago/Turabian Style

Zou, Yajie, Qirui Hu, Wanbing Han, Siyang Zhang, and Yubin Chen. 2025. "Analyzing Travel and Emission Characteristics of Hazardous Material Transportation Trucks Using BeiDou Satellite Navigation System Data" Remote Sensing 17, no. 3: 423. https://doi.org/10.3390/rs17030423

APA Style

Zou, Y., Hu, Q., Han, W., Zhang, S., & Chen, Y. (2025). Analyzing Travel and Emission Characteristics of Hazardous Material Transportation Trucks Using BeiDou Satellite Navigation System Data. Remote Sensing, 17(3), 423. https://doi.org/10.3390/rs17030423

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop