Research on Commuting Mode Split Model Based on Dominant Transportation Distance

Abstract

Conventional commuting mode split models are characterized by inherent limitations in dynamic adaptability, primarily due to persistent dependence on periodic survey data with significant temporal gaps. A dominant transportation distance-based modeling framework for commuting mode choice is proposed, formalizing a generalized cost function. Through the application of random utility theory, probability density curves are generated to quantify mode-specific dominant distance ranges across three demographic groups: car-owning households, non-car households, and collective households. Empirical validation was conducted using Dongguan as a case study, with model parameters calibrated against 2015 resident travel survey data. Parameter updates are dynamically executed through the integration of big data sources (e.g., mobile signaling and LBS). Successful implementation has been achieved in maintaining Dongguan’s transportation models during the 2021 and 2023 iterations.

1. Introduction

Commuting constitutes a fundamental component of urban transportation, characterized by distinct spatiotemporal stability and peak concentration. Its demand demonstrates pronounced regularity, with clearly identifiable temporal and spatial distribution peaks that impose rigorous operational efficiency and resource allocation requirements on transportation systems. Consequently, commuting represents a critical domain for transportation modeling research. Nevertheless, conventional commuting mode split models predominantly rely on periodic comprehensive transportation surveys (typically conducted at 5–10 year intervals), resulting in inherent limitations including data obsolescence and inadequate model adaptability to dynamic conditions.

Within urban transportation ecosystems, each transportation mode possesses unique technical attributes and economic efficiencies. Through modal competition and complementarity, distinct optimal usage ranges emerge for different transportation modes [1]. Driven by the rapid advancement of big data technologies, a pivotal research challenge has arisen: leveraging real-time, dynamic traffic data to determine mode-specific dominant transportation distances for establishing scientifically robust commuting mode split models. Addressing this imperative, this study proposes a dominant distance-based modeling method for commuting mode split. The research outcomes were operationally implemented in the maintenance and version updates of Dongguan’s transportation models (2021 and 2023 iterations).

We summarize the contributions of this study as follows: (1) the development of a distance-based commuting mode choice model that eliminates dependency on survey data for current model recalibration; (2) the research outcomes applied in the maintenance of Dongguan’s Comprehensive Transportation Model (2021, 2023), providing practical implementation experience for subsequent model updates and maintenance efforts.

The remainder of the paper is organized as follows: Section 2 presents a literature review. Section 3 is the generalized cost function model. Section 4 constructs the dominant transportation distance model. Section 5 provides an empirical analysis using Dongguan as a case study. Section 6 summarizes the conclusions.

2. Literature Review

2.1. Influencing Factors of Commuting Mode Split

Researchers primarily employ Revealed Preference (RP) and Stated Preference (SP) surveys to analyze factors influencing commuting mode choices. Individual socioeconomic attributes constitute a significant factor influencing commuting mode choice. Significant behavioral differences exist between male and female commuters across spatial locations [2]. For example, as working hours increase, males’ choices of commuting modes vary depending on their residential location, although cars remain the overall preferred option. In contrast, females predominantly rely on cars, with only a few exceptions in certain areas [2]. Both educational attainment and population density have a negative correlation with peoples’ decision to utilize mass transportation since the unreliability of the transportation system lessens the benefit of using public transport in comparison with that of private transport [3]. This suggested that residents with higher education levels residing in high-density areas show a stronger propensity toward private transport modes. Trips dominated by commuting purposes are more likely to involve motorized transport modes than pure commute trips [4]. Higher bicycle use is correlated to positive cycling self-concepts [5]. Similarly, the commute choice of driving is positively correlated with car self-concepts and negatively correlated with functional difficulties in car use. Respondents with a strong focus on functional travel needs are most likely to commute using a car and least likely to use public transport [5]. Although all commuters prioritized safety, higher-income individuals demonstrated a marked preference for private transportation [6].

The influence of the transportation built environment (BE) on commuting mode choice cannot be overlooked. Although many studies explore BE effects on commuting behavior, most overlook BE characteristics at workplace locations and their non-linear impacts. More importantly, limited effort is placed on the integrative effects of the BE and transportation policies. Research demonstrates that the BE characteristics of residential and workplace locations, along with commuting plans, significantly influence commuting mode choice, collectively accounting for 65% of the predictive power in mode selection. While workplace BE variables demonstrate greater significance than residential BE elements, this difference is largely attributable to the distance from the Central Business District (CBD) [7]. Built environment characteristics are more influential in people’s choices among different public transport sub-modes than in their choice between public transport and cars. Compared to older commuters, millennials’ choices of rail-based and mixed-mode public transport are more susceptible to built environment attributes, whereas their effects on road-based transit usage are stronger for older commutes. Workplace accessibility exerts a significant impact on commuting patterns, whereas BE factors in residential areas demonstrate comparatively weaker effects. This underscores the critical importance of optimizing workplace layouts and transportation infrastructure for promoting public transport adoption [8].

The transportation operating environment and policy interventions also exert substantial influence on commuting mode choice. During the COVID-19 pandemic, over 60% of Seattle’s public transport users switched to alternative modes, with high-income cohorts demonstrating a pronounced shift toward solo driving [9]. This evidences how disruptive events significantly alter commuting behaviors. Furthermore, policy instruments—including transit-priority development strategies and private vehicle restrictions—profoundly reshape mode selection. Empirical analyses confirm that (1) parking pricing and non-auto incentives effectively reduce private vehicle usage while increasing the passenger volume of public transport (bus and metro); (2) parking fees achieve measurable car-trip reduction only when reaching critical pricing thresholds; (3) complementary incentives effectively motivate modal shifts toward sustainable options [10]. Sometimes, the effectiveness of public transit investments may not align with initial expectations. For instance, investments in rail transit often fail to significantly alleviate traffic congestion. The resulting increase in ridership primarily stems from a shift away from non-car modes and induced demand. On the one hand, rail transit accelerates jobs–housing separation, lengthens travel distances, and induces more car trips. On the other hand, the majority of the ridership attracted by rail transit comes from a shift away from bus services.

Commuting duration and satisfaction constitute critical determinants in mode choice behavior. Travel time significantly influences commuter satisfaction, with mode selection acting as a moderating variable in this relationship. Extended commute durations demonstrably reduce overall satisfaction levels. Obviously, opting for more flexible or comfortable modes partially mitigates such dissatisfaction [11].

2.2. Commuting Mode Split Models

There is relatively limited research specifically dedicated to commuting mode choice models. Instead, research efforts have predominantly focused on mode choice models in general. Mode choice models include aggregate models and disaggregate models.

Commonly used aggregate models for forecasting mode choice include the diversion curve model, cross-classification model, and regression model, among others. The diversion curve method depicts plotting the relationship between modal share and its influencing factors based on survey and statistical data [12]. The cross-classification model takes the household as the unit of analysis, categorizing all households into distinct types based on the key determinants of mode choice. The regression model, on the other hand, employs regression equations to predict the relationship between modal share and influencing factors, as exemplified by Scheiner’s model for mode split across different travel distances [13]. These aggregate models are suitable for forecasting urban mode choice where sufficient datasets are available. However, they are generally not applicable in cities or newly developed areas with limited data availability.

Given the challenges in acquiring foundational data such as population size, labor force, and employment opportunities, coupled with advancements in probability theory and the refinement of travel units, researchers have shifted their focus from macroscopic structural patterns and modal shift trends to individual travel behavior characteristics. Pioneered by scholars including MF Train, disaggregate mode choice models were developed by integrating utility theory from economics into mode split analysis based on probabilistic frameworks [14]. These models treat individual travelers as the unit of analysis, leveraging intrinsic relationships within survey samples to derive behavioral choice probabilities, ultimately translating them into modal share predictions [15]. Well-established models include Logit, Probit, and MD models. The Logit model, grounded in random utility theory and utility maximization principles, constructs utility functions incorporating variables related to travel behavior characteristics, traveler attributes, and environmental factors to forecast mode choice [16]. Through continuous evolution, the Logit framework has developed into a comprehensive system encompassing the Multinomial Logit (MNL), Mixed Logit (ML), and Nested Logit (NL) models [17,18].

Discrete choice models represent the predominant methodological framework for analyzing commuting mode selection. Among these, Probit regression and the Analytic Hierarchy Process (AHP) are extensively applied. Pradonoputro and Kozo [3] employed Probit regression to quantify the impacts of educational attainment, population density, and related determinants on public transport utilization. Similarly, Mayo and Taboada [6] implemented AHP methodology to establish a hierarchical ranking of factors influencing public transport mode choice, identifying safety as the paramount consideration. These approaches provided scientifically rigorous foundations for policy formulation through the precise quantification of mode choice determinants. Evolutionary game theory introduces a novel paradigm for understanding longitudinal behavioral adaptation. Gong and Jin [19] simulated commuter decision evolution via replication dynamics mechanisms, demonstrating that public transport development and congestion pressures collectively drive partial modal shifts from private vehicles to transit alternatives. This modeling framework elucidates emergent behavioral trends, delivering forward-looking insights for sustainable urban transportation planning.

As alternatives and supplements to aggregate models, disaggregate models place greater emphasis on individual choice behavior. Compared with traditional aggregate approaches, disaggregate models demonstrate distinct advantages: (1) theoretically grounded in explicit behavioral hypotheses with rigorous logical frameworks; (2) parameter efficiency requiring smaller sample sizes for model calibration; (3) enhanced explanatory power through explanatory variables directly linked to individual decision-making processes, enabling more precise characterization of travel behavior; (4) superior spatiotemporal transferability across contexts; (5) versatile policy assessment capabilities for evaluating diverse transportation planning schemes and policy interventions.

Within China’s current rapid urbanization and infrastructure development phase, sustainable development and the optimal allocation of transportation resources have gained critical importance. Concurrently, transportation planning paradigms are shifting from physical infrastructure improvements toward smart mobility solutions. Consequently, researchers are increasingly leveraging big data analytics to drive innovations in urban transportation modeling.

Currently, big data analytics is applied for the study on activity–travel chain modeling and transport mode choice behavior. A representative case in Shanghai analyzed GPS trajectory data to examine the decision-making process of travelers regarding mode switching and trip chaining patterns [20]. Tu et al. [21] analyzed GPS trajectories and smart card data in Shenzhen, China, revealing how factors such as employment, land use, and road density influence public ridership across different modes. Their findings demonstrate the spatial variability in mode choice and the importance of localized data in transportation planning. Kumar et al. [22] proposed a fuzzy inference-enabled deep reinforcement learning approach for traffic light control. Their system dynamically adjusts signal timing based on real-time traffic data, illustrating how big data can optimize traffic flow and reduce congestion, thereby improving transportation mode efficiency and choice.

2.3. Dominant Transportation Distance for Transport Modes

Investigations on dominant transportation distances primarily focused on comparative analyses of transportation efficiency. Krygsman [23] examined inter-dependencies between transport modes and activity patterns. Müller [24] identified distance as the pivotal factor triggering shifts from low-cost modes to high-cost alternatives, noting that each mode exhibited distinct distance-specific efficiency optima.

2.4. Critical Research Assessment

Commuting mode split: Previous studies leveraged comprehensive travel surveys to holistically examine influencing factors. Contemporary models systematically incorporated individual attributes, socioeconomic characteristics, built environment features, and policy variables, progressively advancing toward integrated mode choice frameworks.

Dominant distance modeling: Current research remains constrained to pairwise comparisons of limited transport modes, failing to establish a generalizable modeling paradigm. This limitation impedes practical implementation across diverse urban contexts.

Research imperative: Amidst rapid urbanization, transportation models, as foundational urban planning instruments, require dynamic updating mechanisms to accommodate spatial transformations. Commuting behavior analysis, being a core module, necessitates enhanced temporal validity and precision to ensure evidence-based decision making. This study proposes a dominant trip distance-based commuting mode split model, delivering a methodological infrastructure for big data-driven transportation model maintenance.

3. Generalized Cost Function

The generalized cost function describes the cost of different transportation modes in a transportation network. It reflects travelers’ costs or resistance under various road sections, time periods, or trip purposes, including factors such as travel time, expenses, comfort level, or a combination of these elements. For computational convenience, this study selects travel time and cost as key components. Travel time encompasses driving time, waiting time, transfer time, etc., and travel costs include fuel expenses, toll fees, parking charges, etc.

From the perspective of commuters, this study calculates the time cost and monetary cost associated with each transportation mode. These costs are normalized using the value of time (VOT) [12] to establish a generalized cost function. To construct this generalized cost function model, the following assumptions are made: (1) the urban transportation system is stable and will not undergo significant changes in the short term; (2) road traffic assignment follows User Equilibrium (UE) [25] principles; (3) transfers between different transportation modes are neglected.

Under the aforementioned assumptions, the generalized cost function for each transportation mode is expressed as

$${GC}_{\lambda }=T_i+F_i/{VOT}_{\lambda }$$

(1)

where ${GC}_{\lambda }$ is the general cost of the commuting group $\lambda$, $T_i$ is the time of transportation mode i, $F_i$ is the cost of transportation mode i, and ${VOT}_{\lambda }$ is the time value of the commuting group $\lambda$.

Walk

Walking incurs no monetary cost. Its generalized cost is expressed as

$${GC}_{walk}=T_{walk}=L_{walk}/V_{walk}$$

(2)

where $T_{walk}$ is the walking transportation time, $V_{walk}$ is the walking speed, and $L_{walk}$ is the travel distance of walking.

The walking distance is constrained by the traveler’s physical endurance. Data from the Dongguan Resident Travel Survey indicated that walking commutes within 1 km account for 89% of all walking commutes, with the maximum feasible walking commute distance being 3 km.

Bicycle

Bicycle travel time includes cycling time and walking connection time (access/egress time). The monetary cost is the bicycle usage fee (which is zero for personal bicycles). Its generalized cost is expressed as

$${GC}_{bcycle}=L_{bicycle}/V_{bicycle}+T_{W,bicycle}+L_{bicycle}\ast f_{bicycle}/VOT$$

(3)

where $V_{bicycle}$ is the average cycling speed, $T_{w,bicycle}$ is the walking connection time (access/egress time) for the bicycle trip, $f_{bicycle}$ is the usage fee per unit of distance traveled by bicycle, and $L_{bicycle}$ is the transportation distance of the bicycle.

Motorcycle

Travel time by motorcycle includes motorcycle riding time and walking connection time (access/egress time). The monetary cost refers to fuel consumption cost. Its generalized cost is expressed as

$${GC}_{moto}=L_{moto}/V_{moto}+T_{W,moto}+L_{moto}\ast f_{moto}/VOT$$

(4)

where $V_{moto }$ is the average motorcycle transportation speed, $T_{W,moto}$ is the walking connection time (access/egress time) for the motorcycle trip, $f_{moto}$ is the usage cost per unit of distance traveled by motorcycle, and $L_{moto}$ is the transportation distance of motorcycle.

Car

Travel time by car comprises driving time, parking time, and walking connection time (access/egress time); costs include fuel/electricity expenses, parking fees, etc. Its generalized cost is expressed as

$${GC}_{car}=L_{car}/V_{car}+T_{W,car}+T_{P,car}+(L_{car}\ast f_{car}+F_{O,car})/VOT$$

(5)

where $V_{car}$ is the average car transportation speed, $T_{W,car}$ is the walking connection time (access/egress time) for the car trip, $T_{P,car}$ is the car parking time, $f_{car}$ is the usage cost per unit of distance traveled by car, $F_{O,car}$ is other costs associated with traveling by car, and $L_{car}$ is the transportation distance of the car.

Taxi

Travel time by taxi comprises transportation time and access/egress walking connection time (waiting time is explicitly excluded due to universal online reservations), with costs comprising base fare, congestion surcharges, highway tolls, etc. Its generalized cost is expressed as

$${GC}_{taxi}=\left\{\begin{array}{l}{L}_{taxi}/V_{taxi}+T_{w,taxi}+F_{0,taxi}/VOT L\le 3 \mathrm{km}\\ L_{taxi}/V_{taxi}+T_{w,taxi}+(F_{0,taxi}+(L_{taxi}-L_0)\ast f_{taxi})/VOT L>3 \mathrm{km}\end{array}\right.$$

(6)

where $V_{taxi}$ is the average taxi transportation speed, $T_{w,taxi}$ is the walking connection time (access/egress time) for the taxi trip, $F_{0,taxi}$ is the starting fare of the taxi, ${ f}_{taxi}$ is the usage cost per unit of distance traveled by taxi, $L_{0 }$ is the starting distance of the taxi, and $L_{taxi}$ is the transportation distance of the taxi.

PT

The generalized cost of PT is expressed as

$${GC}_{PT}=\alpha \ast T_{walk}+\beta \ast T_{IVT}+\gamma \ast T_{IWait}+\delta \ast T_{XWait}+\epsilon \ast T_{Xfer}+{ϵ\ast T}_{Board}+Penalty+F_{PT}/VOT$$

(7)

where $T_{walk}, T_{IVT}, T_{IWait}, T_{XWait}, T_{Xfer}, T_{Board}$, and $Penalty$ denote the walking time, in-PT time, initial waiting time, transfer waiting time, transfer time, boarding time, and transfer penalty, respectively. ${ F}_{PT}$ is public transit fare, and $\alpha ,\beta ,\gamma ,\delta ,\epsilon$ represent the weighting coefficients.

4. Dominant Transportation Distance Model of Mode Choice

The utility of a transportation mode refers to the value provided by different modes in satisfying travel demand, primarily comprising time cost and monetary cost. Its expression is given by

$$u_{i,L}=\frac{1}{{GC}_{i,L}}$$

(8)

where ${GC}_{i,L}$ is the generalized cost of transportation $i$ at distance L.

According to the random utility theory, the probability density function for different transportation modes at the same travel distance is expressed as

$$P_{i,L}=\frac{u_{i,L}}{{\sum }_i{u}_{i,L}}$$

(9)

5. Empirical Research

Using Dongguan City, Guangdong Province as a case study, this research validates the practicality of the proposed methodology based on 2015 resident travel survey data and SP data from Dongguan.

In 2015, Dongguan had a permanent population of 8.343 million, 1.651 million motor vehicles, a regional GDP of CNY 666.5 billion, and a per capita income of CNY 3221 per month [26]. Dongguan’s commuting population is categorized into three groups based on household registration type and car ownership status: home-based work trips by car-owning households (CA_HBW); home-based work trips by non-car-owning households (NC_HBW); and home-based work trips by collective households (CL_HBW). A total of 263,853 travel samples were collected from the Dongguan residents’ travel survey data, including 14,9825 commuting trips. Among these, there were 48,485 CA_HBW trips, 88,020 NC_HBW trips, and 13,320 CL_HBW trips, representing distinct categories of home-based work commuting patterns. The time values for each group are 26.4 CNY/h, 20.4 CNY/h, and 18.1 CNY/h, respectively [26].

Various types of time and cost of transportation modes in Dongguan for 2015 are shown in

Table 1. Various types of time and cost of transportation modes in Dongguan (2015).

Transportation Modes	Walking Connection Time (h)	Waiting Time (h)	Boarding Time (h)	Average Travel Speed (h)	Parking Time (h)	Average Cost Per Kilometer (CNY)
Walk	--	--	--	5	--	--
Bicycle	0.05	--	--	12	--	--
Motorcycle	0.05	--	--	18	--	0.2
Taxi	0.20	--	--	25	--	Base fare for the first 3 km is CNY 8; additional fare beyond 3 km is CNY 2.5/km
Car	0.15	--	--	25	0.05	0.8
PT	0.20	0.10	0.01	18	--	≤24 km: CNY 2/person/ride; 24 km–36 km: CNY 3/person/ride; >36 km: CNY 4/person/ride.

[26].

The weight coefficients for the indicators used in the generalized cost function of public transportation in Dongguan City (2015) are shown in

Table 2. Generalized cost function weighting factor of Dongguan (2015).

Categories		Intra-Town Bus	Inter-Town Bus
Walking Time Weight		1.5	1.5
Waiting Time Weight		1.5	1.5
In-PT Time Weight		1.0	1.0
Boarding Penalty (min)		0.2	0.2
Transfer time (min)	Intra-town Bus	3	3
	Inter-town Bus	3	3
Fare (CNY)		2	≤24 km: CNY 2/person/ride; 24 km–36 km: CNY 3/person/ride; >36 km: CNY 4 person/ride.
Transfer fare (CNY)	Intra-town Bus	2	2
	Inter-town Bus	2	2

[26].

Based on the dataset and model framework established in preceding sections, the generalized cost functions corresponding to transportation modes are summarized in

Table 3. Generalized cost of various transportation modes in Dongguan (2015).

Modes	Generalized Cost Function	Explanation
Walk	${GC}_{Walk}=L/5$	--
Bicycle	${GC}_{Bicycle}=L/12+0.05$	--
Motorcycle	${GC}_{Moto}=L/18+0.05+L\ast 0.2/VOT$	CA_HBW, CA_HBW, and CL_HBW have time values of 26.4 CNY/h, 20.4 CNY/h, and 18.1 CNY/h, respectively.
Taxi	$$\begin{gathered} { GC}_{Taxi}=L/25+0.2+8/VOT L\le 3 \\ {GC}_{Taxi}=\frac{L}{25}+0.2+\frac{8+\left(L-3\right)\ast 2.5}{21.9} L>3 \end{gathered}$$
Car	${GC}_{Car}=L/25+0.15+0.05+(L\ast 0.8)/VOT$
PT	${GC}_{PUT}=0.2+L/18+0.1+0.01+L/20/VOT$

.

Considering the urban scope of Dongguan, the transportation distance is set at 60 km (with a maximum distance of 27 km for CL_HBW) with a step size of 0.5 km. Using the aforementioned calculation methods and constraints, the probability density curves for different commuting modes (CA_HBW, NC_HBW, and CL_HBW) across various transportation distances are illustrated in Figure 1, Figure 2 and Figure 3.

The probability density curves were fitted using MATLAB R2020a, with the optimal functional form for each transportation mode determined to be either $Y=a\ast x^b\ast exp(c\ast x)$ or $Y=a1\ast exp(-((x-b1)/c1)^2)+a2\ast exp(-((x-b2)/c2)^2$), where x represents distance (in meters) and Y denotes probability density.

Table 4. Dominant transportation distance function of multi-transportation modes for CA_HBW.

Modes	Dominant Transportation Distance Function	R2	RMSE	Residual Analysis
				Mean	Standard Deviation
Walk	$Y=0.1555\ast L^{-0.4522}\ast exp(0.0023\ast L)$	0.8853	0.0085	0.0000	0.0080
Bicycle	$Y=0.1529\ast L^{0.1158}\ast exp(-0.1532\ast L)$	0.9962	0.0063	0.0000	0.0060
Motorcycle	$Y=0.1518 \ast L^{0.1673}\ast exp(-0.1122\ast L)$	0.9916	0.0051	0.0000	0.0051
Taxi	$Y=0.0028\ast L^{ 0.3583 }\ast exp(-0.1947\ast L)$	0.9932	0.0003	−0.0001	0.0003
Car	$Y=0.5010\ast L^{0.1988}\ast exp(-0.0026\ast L)$	0.9483	0.0072	0.0003	0.0071
PT	$Y=0.0653\ast L^{0.1639}\ast exp(-0.0400\ast L)$	0.8963	0.0081	−0.0002	0.0080

,

Table 5. Dominant transportation distance function of multi transportation modes for NC_HBW.

Modes	Dominant Transportation Distance Function	R2	RMSE	Residual Analysis
				Mean	Standard Deviation
Walk	$Y=0.4379\ast L^{-0.5197}\ast exp(-0.0469\ast L)$	0.8943	0.0092	−0.0004	0.0091
Bicycle	$Y=0.3339\ast L^{0.4957}\ast exp(-0.0907\ast L)$	0.9772	0.0112	−0.0003	0.0111
Motorcycle	$Y=0.0454 \ast L^{1.1660}\ast exp(-0.0686\ast L)$	0.8661	0.0135	−0.0004	0.0134
Taxi	$Y=0.0002\ast L^{ 2.365 }\ast exp(-0.2715\ast L)$	0.9490	0.0051	−0.0002	0.0075
Car	$Y=0.3061\ast exp(-((x-59.69)/5.191)^2)+0.3001\ast exp(-((x-47.21)/16.39)^2)$	0.9049	0.0296	0.0002	0.0037
PT	$Y=0.6719\ast exp(-((x-65.05)/3.066)^2)+0.4494\ast exp(-((x-40.77)/27.51)^2)$	0.8909	0.0185	−0.0016	0.0019

and

Table 6. Dominant transportation distance function of multi-transportation modes for CL_HBW.

Modes	Dominant Transportation Distance Function	R2	RMSE	Residual Analysis
				Mean	Standard Deviation
Walk	$Y=1.46\ast L^{0.4598}\ast exp(-0.5868\ast L)$	0.9706	0.0080	0.0000	0.0075
Bicycle	$Y=0.1846\ast L^{1.2010}\ast exp(-0.1895\ast L)$	0.9814	0.013	0.0000	0.0120
Motorcycle	$Y=0.0076 \ast L^{1.8110}\ast exp(-0.1866\ast L)$	0.9106	0.0070	0.0000	0.0068
Taxi	$Y=0.0004\ast L^{ 2.3240 }\ast exp(-0.2132\ast L)$	0.8959	0.0005	0.0003	0.0031
Car	$Y=0.3458\ast exp(-((x-30.32)/3.704)^2)+0.6042\ast exp(-((x-20.26)/11.96)^2)$	0.9397	0.027	0.0001	0.0025
PT	$Y=0.3563\ast exp(-((x-25.53)/1.543)^2)+0.1525\ast exp(-((x-14.24)/12.16)^2)$	0.9169	0.0110	0.0003	0.0011

present the optimal transportation distance functions and fitting indicators for each transportation mode under CA_HBW, NC_HBW, and CL_HBW commuting patterns, respectively. The commuting mode split in Dongguan’s transportation model for 2021 and 2023 was maintained and updated by modifying the parameters in Table 4, Table 5 and Table 6 based on multi-source transportation big data (including mobile signaling data, LBS data, etc.).

Table 7. Modeled vs. surveyed mode share proportions for commuting.

Commuting Mode	Data Source	Walk	Bicycle	Motorcycle	Car	Taxi	PT
CA_HBW	Survey	11%	9%	12%	61%	1%	6%
	Model	13%	9%	12%	62%	0%	4%
NC_HBW	Survey	39%	35%	16%	2%	1%	7%
	Model	40%	36%	17%	1%	0%	6%
CL_HBW	Survey	60%	28%	3%	6%	0%	3%
	Model	66%	23%	1%	6%	0%	4%

presents the modal split proportions for each commuting mode calculated from Table 4, Table 5 and Table 6 compared with the 2015 survey data. The comparative analysis confirms that the model achieves ≥90% accuracy in reproducing the primary transportation mode structure, meeting the established validation criterion [27].

Based on the assumption in Section 3 that the urban transportation system is stable and will not undergo significant changes in the short term, Table 4–6 are applicable for current model maintenance and short-term transportation forecasting but are not suitable for medium- and long-term transportation predictions.

By analyzing the probability density curves of transportation distance for various modes, the following conclusions can be drawn.

Private cars hold dominance in CA_HBW. Beyond 4 km, the share rate of private cars exceeds 0.6, and dominance becomes more pronounced over longer distances.

There is a phased modal shift in NC_HBW. Short-distance trips rely predominantly on non-motorized transport (1–5 km: dominated by walking and cycling); short-to-medium-distance trips rely on cycling, motorcycles, and public transport (5–13 km: bicycle dominance, with motorcycles and public transport showing a significant increase in modal share; 13–27 km: absolute dominance of motorcycles); medium-distance trips rely on public transport and private cars (27–55 km: public transport dominance, with private cars exhibiting a significant rise in modal share); and long-distance trips are dominated by private cars (>55 km: private car dominance).

Extreme and flexible commuting exists in CL HBW. Short-to-medium-distance trips exhibit high reliance on non-motorized modes (1–3 km: absolute predominance of walking, exceeding 60% of trips aligning with Dongguan’s “factory-residence proximity” industrial model; 3–13 km: bicycle dominance). Long-distance commuting is predominantly reliant on private cars. Due to the multi-center structure of Dongguan’s industrial structure, the average commuting distance between core towns (such as Nancheng and Dongcheng) and peripheral industrial areas (such as Chang’an and Humen) is about 20–25 km. Exceeding 25 km is a typical distance for cross-town commuting (such as from Liaobu to Songshan Lake), triggering a change in transportation mode. And these long-distance cars are becoming increasingly competitive, while the competitiveness of public transportation is gradually declining.

Walking commuting: Walking holds absolute dominance for trips under 1.5 km and retains a comparative advantage within the 1.5–3 km range. Beyond 3 km, its selection probability approaches zero. CL_HBW exhibits the highest dependence on walking for commuting, while CA_HBW shows the lowest dependence. While the act of walking has no financial costs, the opportunity cost of time spent, especially over long distances, and the potential loss of productivity should be considered.

Bicycle commuting: Bicycles exhibit absolute dominance for trips concentrated within the 3–7 km range. NC_HBW exhibits the highest dependence on bicycles within this dominant distance.

Motorcycle commuting: Motorcycle usage is predominantly concentrated among local residents. Within NC-HBW, motorcycles show clear dominance in the 13–26 km range and a comparative advantage in the 5–13 km and 26–60 km ranges. However, they demonstrate no significant advantage in CA_HBW and CL_HBW.

Public transport commuting: Public transport (PT) is primarily used by NC_HBW, exhibiting absolute dominance in the 25–55 km range. Dongguan’s rapid transition from industrialization to urbanization occurred with severely underdeveloped PT infrastructure. Coupled with the recent rapid growth of private cars, residents have developed a significant dependence on cars. Consequently, PT’s overall modal share dropped from 5.3% in 2015 to 2.2% in 2023, with the downward trend persisting [26,28,29,30,31].

Private car commuting: Private cars serve as the primary commuting mode in CA_HBW, achieving absolute dominance for trips exceeding 4 km. The private car overall modal share rose from 25.2% (2015) to 35.8% (2023) [26,28,29,30,31].

The results reveal that walking and cycling, especially for short distances, have a higher preference threshold than other modes. Furthermore, it was found that private car use exhibits decreasing sensitivity as distance increases, and this distance attenuation effect is most pronounced in CA_HBW. Rapid urbanization favors private cars, while PT struggles to compete, and non-motorized modes remain vital for short-distance trips.

The permanent resident population, motor vehicle ownership, GDP, and per capita income of Dongguan City from 2021 to 2023 are shown in

Table 8. Key socioeconomic indicators of Dongguan City in 2021–2023.

Year	Permanent Resident Population (1000 Persons)	GDP (CNY Trillion)	Per Capita Income (CNY)	Motor Vehicle Ownership (10,000 Units)
2021	1053.7	1.09	5175	365.4
2022	1043.7	1.14	5316	391.5
2023	1048.5	1.19	5475	415.2

.

According to Table 7, the population size of Dongguan is basically stable, but with the increase in income, the number of motor vehicles has increased significantly. Building upon preceding analytical conclusions, it is evident that car-owning households secure comprehensive mobility access across all distances through private vehicles. Conversely, non-car-owning households resort to cost-effective combinations of non-motorized transport and public transit modes, though this results in markedly low efficiency for medium-to-long-distance travel. Collective households demonstrate a polarized mobility pattern: achieving extreme compactness for short-distance trips while enduring passive, high-cost solutions for longer journeys.

6. Conclusions

This study establishes a commuting mode split model based on dominant transportation distance, achieving three fundamental breakthroughs in traditional transportation modeling paradigms. Theoretically, we propose an “Integrated Impedance-Utility Threshold” framework that quantifies distance–sensitivity functions for walking, cycling, motorcycles, private cars, ride-hailing, and public transport. This systematically reveals travel characteristics among Dongguan’s commuter groups: (1) CA_HBW secures “full-distance mobility privilege” through private cars; (2) NC_HBW and CL_HBW demonstrate distance-dependent modal shift characteristics. Technologically, parametric dynamic update architecture leverages real-time mobile signaling and LBS data. This approach successfully replaces conventional 5–10 year survey cycles, compressing model maintenance to annual updates and powering Dongguan’s 2021 and 2023 transportation model iterations. This study provides policy recommendations for urban transportation planning by explaining the distance sensitivity of different transportation modes with an integrated model. The findings can guide decision-makers in determining thresholds, especially for promoting sustainable transportation modes.

The empirical results expose structural contradictions in Dongguan’s mobility system: private car mode share soared from 25.2% (2015) to 35.8% (2023), while public transport collapsed from 5.3% to 2.2%. To address this, future research will advance along two dimensions: (1) continuously collecting modal share data to build Dongguan’s transportation structure evolution model; (2) developing group-specific policy interventions and providing model-derived evidence for transportation planning.