A Study on Motorists’ P&R Use Behavior in a River–Valley-Type City Considering the Moderating Effect of Terrain Spatial Perception

Abstract

Traffic problems in China’s river–valley-type cities are prominent. In order to solve these traffic problems, this paper presents four aspects of motorists’ terrain spatial perception according to the characteristics of river–valley-type cities. Based on the structural equation model (SEM), four-dimensional terrain spatial perception is reduced by second-order confirmatory factor analysis. The SEM–Logit model was constructed to analyze the influences of individual socioeconomic attributes, objective built-environment attributes, travel characteristics, and terrain spatial perception on Park and Ride (P&R) use behavior, as well as the moderating effects of terrain spatial perception. The results show that the four latent variables can explain the terrain spatial perception of motorists in river–valley-type cities well. Objective built environments and motorists’ subjective terrain spatial perception have significant impacts on the use of P&R facilities in river–valley-type cities. The terrain spatial perception of motorists in a river–valley-type city moderates the impact of work–residence distance, road slope, river crossing demand, group travel, departure time, and the time saved on P&R use behavior. The research results can provide some reference for solving traffic problems in river–valley-type cities.

1. Introduction

China has a large population base and particularly inadequate land resources. With the advancement of urbanization, the number of private cars continues to grow. According to statistics, the number of motor vehicles in the country is projected to reach 435 million in 2023, with 24.56 million newly registered cars, marking a 5.73% increase from 2022 [1]. As a result, most cities in China face varying degrees of traffic congestion. The fundamental reason is that the speed of road construction is far less than the growth rate of traffic demand, leading to an imbalance between traffic supply and demand. China is a mountainous country, and valley cities account for approximately half of the cities in China [2]. The term river–valley-type cities refers to cities that are formed and developed between valleys and rivers, such as Lanzhou and Xining. Due to the limitations of terrain and the division of the river, the road resources in river–valley-type cities are more limited than those in plain cities, making the contradiction between supply and demand particularly prominent. However, new roads alone cannot solve the urban traffic issue [3].

Park and Ride (P&R) is an effective method to reduce the number of private cars entering the city center by implementing a low-cost or even toll-free management policy to encourage motorists to transfer to the subway at the city’s periphery bottleneck [4]. This helps to reduce travel costs and promotes environmental protection. However, it also faces some challenges, such as inconvenient transfers, reliance on the public transportation system, parking-lot management issues, and the possibility of worsening congestion near the stations. Therefore, when constructing Park and Ride facilities, it is necessary to balance their advantages and disadvantages and take corresponding measures according to the traffic environment of different cities to optimize the user experience of Park and Ride. The comparison between P&R and driving all the way is shown in Figure 1. In river–valley-type cities, P&R can not only alleviate the issue of car congestion and parking difficulty in the narrow junction between the urban groups and within the group center but also strengthen the connectivity of subway stations. This is one of the effective traffic management strategies to reduce urban carbon emissions and alleviate traffic pressure in the central area. In addition, the axial distance of transportation corridors in river–valley-type cities is long, which presents a clear agglomeration effect that is highly beneficial for the development of subways [5]. Commuter travel, as the primary travel demand in urban transportation, reflects the efficiency of urban travel. Therefore, it is of great significance to study the commuting behavior of motorists who use P&R in river–valley-type cities to alleviate traffic congestion issues in such areas. P&R has been successfully applied in London, Singapore, Tokyo, Shanghai, and other cities, but some cities have failed [6]. This highlights the importance of adjusting P&R based on the requirements of motorists in various traffic environments, particularly in river–valley-type cities, which differ significantly from plain cities in terms of terrain, spatial layout, and road network planning and development.

The purpose of this paper is to identify the factors that influence various aspects of motorists’ behavior of using P&R in a river–valley-type city and to assess the extent of their impact. We conducted a questionnaire survey in Lanzhou and analyzed the collected data using the structural equation model (SEM) and the Logit model. The results of this study could offer valuable insights for accurately formulating traffic policies and optimizing the P&R service environment to encourage motorists to embrace P&R as a reliable mode of travel. It also helps to improve the use of P&R in river–valley-type cities and provides a certain degree of reference for other river–valley-type cities.

This paper is mainly divided into five parts. In the second part, previous research on travel behavior using P&R is reviewed to gain insights and understand the limitations of current studies. In the third part, the models and methods used in this paper are described. In the fourth part, the research area is determined, and the collected data are described, tested, and used to solve the model. In the fifth part, the model results are analyzed and discussed. The last part provides a summary.

2. Literature Review

In the years that P&R have been in operation, there has been a marked disparity in development among different cities. After years of practice, P&R’s planning has gradually shifted from being demand-oriented to resource-oriented, where limited resources are utilized to address parking needs. Building P&R in various traffic contexts is crucial for enhancing P&R utilization rates. Therefore, it is also the core of the current research to identify the key factors that affect motorists’ P&R use behavior under different traffic conditions and to examine the moderating effects of context-specific factors on the factors influencing P&R use behavior.

Motorists’ willingness to use P&R will vary depending on their individual circumstances and the travel environment. In terms of the individual socio-economic attributes of motorists, in addition to common factors such as gender, income, private car ownership, and degree of freedom in working time, which have been proven to have a significant impact on the travel mode choice [7], the driving age of motorists is also considered to have a significant impact on the use of P&R in the context of special urban terrains [8]. Due to the economic benefits of using P&R for commuting compared to driving all the way, some researchers have also considered transportation payment methods [9]. Other scholars point out that the experience of using P&R also significantly influences the choice of using P&R [10]. In terms of travel characteristics and attributes, compared with driving all the way, the time and cost saved by P&R commuting also significantly influence motorists’ behavior of using P&R [11,12,13,14].

Scholars also found that travelers’ behavior will change with changes in the built environment [15]. The built environment of a river–valley-type city differs significantly from that of a plain city of the same scale. River–valley-type cities exhibit a significant separation between workplace and residence, distinct group characteristics, narrow connections between groups, and limited road infrastructure [16]. These characteristics make the peak-hour duration of river–valley-type cities longer [17]. Therefore, the travel time during peak periods [18] and the necessity of cross-group travel [19] may influence the selection of the travel mode. The flow direction of the river in a river–valley-type area is consistent with the main axis of urban areas [17]. Therefore, compared with plain cities, motorists in river–valley-type cities have to take more detours across rivers [20]. River–valley-type cities have steep road slopes, and Xiong et al. have confirmed the significant impact of road slopes on the choice of travel mode [21]. Studies have shown that, in addition to objective built-environment factors, travelers’ subjective perception of the built environment also significantly affects travel behavior [22]. Xiong et al. proposed the subjective perception of road slope in mountainous cities, which confirmed the significant impact of terrain perception on travel mode choices [23]. Basu pointed out that pedestrians’ perception of security would affect their intention to walk [24]. For P&R users, the aspects of security perception are important, such as surveillance, lighting facilities, and pedestrian-crossing facilities [14,25,26]. In addition, for motorists in river–valley-type cities with long peak hours and narrow group junctions, P&R can also provide users with a sense of psychological security, reducing the fear of being late [27]. It has been confirmed that the motorist’s perception of convenience and comfort when using P&R will influence the motorist’s choice of travel mode [8]. Research has shown that both the availability of parking spaces when using P&R [28,29] and the additional services provided [30,31,32] will promote the use of P&R. Congestion times [18], driving pressure [10], and uneven experiences [33] of driving all the way can also encourage motorists to choose P&R. In addition, the research results of Guo et al. show that the motorist’s perception of accessibility impacts the choice of travel mode [34]. With good traffic condition around P&R, motorists can easily choose P&R and reach destinations across different groups more quickly. In addition, commuting using P&R is less affected by the large number of one-way roads and dead-end roads in river–valley-type cities compared to driving all the way [17].

After reviewing the existing literature, it is evident that, while there are numerous studies on P&R use behavior, few scholars have taken the subjective and objective built-environment factors into account to analyze the behavior of using P&R in river–valley-type cities with unique terrains. Particularly, the moderating effect of motorists’ subjective terrain spatial perception on various factors affecting P&R use behavior has not been thoroughly explored. The travel space for motorists in a river–valley-type city is different from that in a plain city. The study of motorists’ behavior of using P&R in a river–valley-type city cannot simply rely on the research results of other cities. Therefore, it is necessary to carefully understand whether the key factors affecting motorists’ use of P&R in river–valley-type cities differs and to what extent compared with the results that do not consider the terrain spatial perception of river–valley-type cities. In this study, security perception (SEC), convenience perception (CON), comfort perception (COM), and accessibility perception (ACC) are introduced to explain the terrain spatial perception (TSP) of motorists in a river–valley-type city using P&R commuting. The SEM–Logit model was developed to analyze the impact of factors on motorists’ P&R use behavior in river–valley-type cities. Furthermore, the model examined the moderating effect of terrain spatial perception on the influence of various factors of P&R use behavior.

3. Research Methodology

3.1. Research Area

Lanzhou, the second-largest city in Northwest China, is located at a crossroads in the northwest. It has greatly benefited from the Silk Road, evolving into an important tourist destination and transportation hub. In the region, the north and south mountains face each other, and the Yellow River flows from the northwest to southeast, dividing the entire area. This formation creates a beaded pattern of alternating rivers and valleys, characteristic of a river–valley-type city [3]. Compared with plain cities of the same scale, river–valley-type cities exhibit distinct group characteristics, greater separation between workplace and residence, and steeper road slopes. In addition, there is a river in the same direction as the main axis of the city in the region. The travel characteristics are quite different from those of ordinary plain cities [17].

In 2023, the total passenger volume of Lanzhou subway exceeded 100 million. By the end of 2023, the number of motor vehicles in Lanzhou had reached 1.316 million, showing a 9.89% increase compared to the previous year. Additionally, the newly registered motor vehicles had surged by 43.94% compared to the same period last year [35]. It can be seen that, although the subway carries a larger portion of the passenger flow, the travel of motor vehicles is less affected. In order to alleviate road congestion, two P&R parking lots were established in December 2023 in Lanzhou City. They are located at Donggang Station of Subway Line 1 and Yanbai Bridge Station of Subway Line 2 [36].

3.2. Questionnaire Design and Data Collection

Due to significant differences in traffic characteristics and backgrounds, the index system of plain cities of the same scale has little reference significance in the study of P&R use behavior of motorists in river–valley-type cities. Therefore, based on the characteristics of river–valley-type cities mentioned above, this paper introduces the latent variables of terrain spatial perception in four aspects: security perception, convenience perception, comfort perception, and accessibility perception. The question description of the corresponding observation variables is shown in

Table 1. Description of observed variable.

Variable Symbol		Question Description of the Corresponding Observed Variables
Security perception	SEC1	The P&R parking lot has complete surveillance lighting facilities.
	SEC2	An underpass between parking and pick-up points makes P&R safer.
	SEC3	Using P&R, which has less risk of being early or late, provides more psychological security.
Convenience perception	CON1	It is less affected by the blockage at the narrow junction of urban groups when using P&R.
	CON2	It is less affected by rivers in river–valley-type cities when using P&R.
	CON3	It is easier to find a parking space when using P&R.
	CON4	Other services are provided near the P&R parking lot.
Comfort perception	COM1	The stability of using P&R is better.
	COM2	Using P&R is not affected by peak hour congestion and does not require a long ride time.
	COM3	There is no driving pressure caused by complex road conditions when using P&R.
Accessibility perception	ACC1	Using P&R travel between urban groups can enable users to reach the destination faster.
	ACC2	The traffic around P&R parking lots is good.
	ACC3	It is less affected by the one-way street and the broken road when using P&R.
	ACC4	It is less affected by all aspects of the spatial terrain than driving all the way.

. These question descriptions are aimed at the mode that respondents use most during the survey month.

Taking Lanzhou, a typical river–valley-type city in China, as an example, this study collected commute data from motorists around P&R parking lots of Donggang subway station and Yanbai Bridge subway station through an offline questionnaire survey. The study focused on the commuting peak period from 18 March to 29 March 2024.

The questionnaire is divided into two parts. The first part is divided into four question sets and is conducted by an RP survey. The first set includes individual socio-economic attributes of the motorist, such as gender, driving age, monthly income, private car ownership, transportation payment methods, degree of freedom in working time, and experience of using P&R. The second set is the motorists’ typical commuting travel attribute, which includes travel distance, road slope, river crossing demand, group characteristics between the OD, and departure time. The third set is the time and cost saved by using P&R compared to driving all the way. The fourth set is whether P&R is the most commonly used mode of commuting in the last month. The second part involves examining the potential variables influencing motorists’ psychological perception of using P&R through a survey conducted using the SP method. Fourteen questions were asked from five aspects: perception of security, perception of convenience, perception of comfort, perception of accessibility, and behavioral intention. They were evaluated using a five-level Likert scale. The description of psychological latent variables is presented in Table 1.

3.3. Research Method

Compared with the traditional linear regression model, the Logit model is more suitable for studying problems where the dependent variable is a binary variable. Therefore, the Logit model is well-suited for analyzing issues with binary outcomes: 1 (using public transportation) and 0 (driving all the way). However, it cannot accurately explain the influence of drivers’ perception, attitude, and other psychological variables on travel behavior [37]. By comparing the existing literature, it is predicted that including psychological factors can more effectively explain individual travel patterns compared to when psychological factors are absent [38,39]. SEM is an important method for multivariate analysis. It can quantify the psychological underlying factors of drivers in river–valley-type cities. Its advantage lies in its ability to estimate the relationship between multiple variables simultaneously and test the relationship between multiple variables in a single model [40]. However, if only the influence of a single SEM is analyzed, and the relationship between the influencing factors of multiple aspects considered in this paper is analyzed simultaneously, the model will become more complex. The structural relationship between each factor becomes difficult to clarify, and the calculations become too extensive. Additionally, the relationship between different factors may be overlooked, leading to redundancy or duplication in factor construction [41].

Therefore, in this paper, based on the Logit model, SEM is introduced to describe the impact of potential psychological perception variables on motorists’ P&R decision-making. Four latent variables, namely, security perception (SEC), convenience perception (CON), comfort perception (COM), and accessibility perception (ACC), are utilized to elucidate the terrain spatial perception of motorists in river–valley-type cities using P&R. The Logit model examined the effects of the psychological attributes mentioned above, as well as the motorists’ individual socio-economic attributes, objective built-environment attributes, and travel characteristics on P&R use behavior. It also analyzed the moderating effects of terrain spatial perception on other factors influencing P&R use behavior (refer to Figure 2).

SEM mainly consists of two parts: the measurement model and the structural model.

The measurement model describes the latent variables through the corresponding observed variables.

$$x={\Lambda }_x\xi +\epsilon$$

(1)

$$y={\Lambda }_y\eta +\epsilon$$

(2)

where $x$ and $y$ are the vectors of observed variables of exogenous latent variables and endogenous latent variables, respectively; ${\Lambda }_x$ is the factor loading matrix of $\xi$ to $x$; ${\Lambda }_y$ is the factor loading matrix of $\eta$ to $y$; $\xi$ and $\eta$ are the vector composed of exogenous latent variables and endogenous latent variables, respectively; $\epsilon$ is the vector of observation error.

The structural model utilizes exogenous latent variables to elucidate the endogenous latent variables.

$$\eta =\mathrm{{\rm B}}\eta +\Gamma \xi +\zeta$$

(3)

where Β is the structural coefficient matrix of endogenous latent variables; $\xi$ is exogenous, the structural coefficient matrix of latent variables; $\zeta$ is the residual vector of the model.

In this paper, four aspects of terrain spatial perception are proposed. When analyzing the moderating effect of terrain spatial perception on the influence of other factors on P&R use behavior, if the interaction terms are directly constructed for modeling and analysis, the model structure will be very redundant and the correlation between variables will increase, ultimately leading to deviations in the model results [42]. In a confirmatory factor model with multiple factors, if the lower-order factors are closely related and can collectively represent a broader concept in theory, then these lower-order factors can be considered as new measurement indicators to estimate the higher-order factor structure. Therefore, before constructing the SEM–Logit model, a second-order confirmatory factor analysis is used to reduce the dimensionality of the four subjective perception variables, resulting in the extraction of a terrain spatial perception factor. The four subjective perception variables are normalized to the load coefficient of the terrain spatial perception factor. The calculation formula is shown in Equation (4):

$${\lambda }_r={\displaystyle \frac{{\lambda }_r}{{\displaystyle \sum _{r=1}^4{\lambda }_r}}},r=1,2,3,4.$$

(4)

where ${\lambda }_r$ is the confirmatory factor model coefficient of the latent variable $r$.

This coefficient is then used as the weight for each factor to calculate the corresponding observation value of the terrain spatial perception. The calculation formula is shown in Equation (5):

$$TSP={\lambda }_{1}SEC+{\lambda }_{2}CON+{\lambda }_{3}COM+{\lambda }_{4}ACC$$

(5)

The utility $U_{in}$ of motorist $n$’s choice of mode $i$ can be expressed by Equation (6):

$$U_{in}=V_{in}+{\epsilon }_{in}=\beta x_{in}+{\epsilon }_{in}$$

(6)

where $V_{in}$ is the observable utility part, that is, the definite term; ${\epsilon }_{in}$ is the unobservable portion of utility. $x_{in}$ is the characteristic variable when motorist $n$ chooses scheme $i$. $\beta$ is the parameter to be estimated for the characteristic variable.

According to the theory of utility maximization, when a motorist commutes, he will choose the commuting mode with the greatest utility according to his own situation and various factors. Suppose that the set of all travel mode selection schemes that the motorist can choose is $A$, where the utility of the scheme $j$ is $U_{jn}$, then the conditions for the motorist to choose the scheme are:

$$U_{in}>U_{jn}, i\ne j, j\in A$$

(7)

According to the research object in this paper, it is defined that the travel mode selection set of motorists consists of only 1 (using P&R) and 0 (driving all the way), so the binary Logit model is adopted for analysis. The probability that the motorist chooses to commute using P&R and driving all the way can be expressed by Equation (8):

$$P_{in}={\displaystyle \frac{\exp (V_{in})}{\exp (V_{1n}+V_{0n})}}, i=0,1.$$

(8)

where $V_{in}$ is the utility fixed term of commuting scheme $i$ selected by car motorist $n$.

4. Results

4.1. Descriptive Statistics of Questionnaire Data

Prior to participating in the study, all participants were informed about the study’s details. Upon completing and submitting the study questionnaire, their consent to participate in the study and the release of their data would be assumed. A total of 500 questionnaires were sent out. Invalid questionnaires containing missing values and extreme continuous values were excluded [43]. A total of 448 valid questionnaires were obtained, achieving an effective rate of 89.60%, which met the sample size requirement [44]. The sample characteristics of motorists are shown in

Table 2. Variable explanation and descriptive statistics.

Variable	Definition	Mean	Std.
Gender	1 male, 0 female	0.480	0.500
Monthly income (CNY)	1 [0, 5k), 2 [5k, 10k), 3 [10k, 15k), 4 [15k, +∞)	2.536	1.123
Drive age	1 [0, 5 years), 2 [5 years, 10 years), 3 [10 years, 15 years), 4 [15 years, +∞)	2.469	1.135
Degree of freedom in working time	1 high, 0 low	0.507	0.506
The number of private cars owned	1–2 and above, 0–1	0.469	0.499
The payment method of traffic cost	1 at public expense, 0 at your own expense	0.516	0.498
Have the experience of using P&R	1 no, 0 yes	0.478	0.460

and Figure 3.

4.2. Terrain Spatial Perception

In order to ensure the effectiveness of the subsequent modeling research, IBM SPSS 26 software should be used to test the scale data in the questionnaire. As shown in

Table 3. Data validity tests.

Latent Variable	Cronbach’s Alpha	KMO	Bartlett’s Test of Sphericity	Total Variance Explained Cumulative
Security perception	0.829	0.900	0.000	73.953%
Convenience perception	0.865
Comfort perception	0.799
Accessibility perception	0.851

, the questionnaire data successfully passed the Cronbach and Kaiser–Meyer–Olkin (KMO) tests [45], indicating that the data met the requirements of multivariate normality and sampling adequacy, thus ensuring reliable data quality [40]. In addition, the cumulative variance explanation rate is greater than 40% [45].

As shown in

Table 4. Reliability and validity tests.

Latent Variable		Factor Load Coefficient	AVE	CR
Security perception	SEC1	0.804	0.627	0.834
	SEC2	0.821
	SEC3	0.749
Convenience perception	CON1	0.823	0.620	0.867
	CON2	0.803
	CON3	0.752
	CON4	0.769
Comfort perception	COM1	0.679	0.617	0.826
	COM2	0.924
	COM3	0.732
Accessibility perception	ACC1	0.762	0.608	0.861
	ACC2	0.877
	ACC3	0.751
	ACC4	0.720

, composite reliability (CR) and average variance extracted (AVE) were used to assess the reliability and validity. The CR and AVE values for all dimensions are above 0.7 and 0.4, respectively. This suggests that the measurement indicators of latent variables presented in Table 1 exhibit strong reliability and validity [46].

The result of the two-order confirmatory factor analysis is shown in Figure 4. AMOS 28.0 was used to estimate the parameters of SEM. Chi-square degrees of freedom (CMIN/DF), approximate root mean square error (RMSEA), root mean square residual (RMR), goodness of fit index (GFI), comparative fit index (CFI), normalized fit index (NFI), Tucker–Lewis index (TLI), incremental fit index (IFI), and standardized root mean square residual (SRMR) were used to assess the fitness of model. The model has a high degree of fitness and meets the standard requirements [40], as depicted in

Table 5. Goodness of fit for structural equation models.

Indicators	CMIN/DF	GFI	CFI	NFI	TLI	IFI	RMSEA	RMR	SRMR
Criteria	1~5	>0.8	>0.8	>0.8	>0.8	>0.8	<0.08	<0.05	<0.05
Actual	3.267	0.934	0.962	0.946	0.950	0.962	0.071	0.026	0.038

.

The path coefficients were standardized, as shown in

Table 6. Coefficient of second-order confirmatory factor analysis.

Model Path	Std Estimate	Adapt Estimate	p
Terrain spatial perception → Security perception	0.862	0.253	<0.001
Terrain spatial perception → Convenience perception	0.868	0.254	<0.001
Terrain spatial perception → Comfort perception	0.731	0.214	<0.001
Terrain spatial perception → Accessibility perception	0.950	0.279	<0.001

[47].

According to Table 5, the four-part perception outlined in this paper—security perception, convenience perception, comfort perception, and accessibility perception—can effectively elucidate travelers’ terrain spatial perception. Furthermore, we can draw an interesting conclusion from Table 6 that the significance of accessibility, convenience, security, and comfort in P&R travel diminishes successively in travelers’ perception of terrain space in river–valley-type cities. This conclusion points to the priority of the river–valley-type cities’ traffic management departments when optimizing P&R.

4.3. Terrain Spatial Perception

To verify the necessity of considering the built environment and the moderating effects of the built environment when analyzing motorists’ P&R behavior in river–valley-type cities, we constructed three models in

Table 7. The results of three models.

Variable	Model 1		Model 2		Model 3
	Coef.	p	Coeff.	p	Coef.	p
con_	2.808	0.000 ***	−2.284	0.036 **	8.366	0.000 ***
Individual socioeconomic attributes
Gender	−0.371	0.083 *	−0.308	0.167	−0.401	0.092 *
Monthly income	0.087	0.363	0.110	0.269	0.074	0.484
Drive age	−0.453	0.034 **	−0.519	0.020 **	−0.593	0.012 **
Degree of freedom in working time	−0.170	0.072 *	−0.167	0.088 *	−0.164	0.113
The number of private cars owned	0.513	0.018 **	0.528	0.019 **	0.498	0.036 **
The payment method of traffic cost	−0.409	0.056 *	−0.386	0.084 *	−0.440	0.065 *
The experience of using P&R	−0.466	0.030 **	−0.479	0.032 **	−0.594	0.012 **
Travel characteristics
Departure time	0.272	0.028 **	0.235	0.073 *	−0.326	0.300
The time saved	0.403	0.001 ***	0.323	0.013 **	−0.319	0.315
The cost saved	0.449	0.000 ***	0.307	0.020 **	−0.405	0.195
Objective built-environment attributes between ODs
The distance			0.313	0.024 **	−0.183	0.577
The rode slope			0.371	0.079 *	−1.025	0.050 *
The demand of river crossing			0.397	0.065 *	−0.531	0.323
Group characteristic			0.474	0.021 **	−0.844	0.112
Terrain spatial perception attributes
Terrain spatial perception			0.188	0.029 **	−3.072	0.000 ***
Interactions
Terrain spatial perception × The distance					0.208	0.066 *
Terrain spatial perception × The demand of river crossing					0.220	0.056*
Terrain spatial perception × The rode slope					0.204	0.086 *
Terrain spatial perception × Group characteristic					0.422	0.025 **
Terrain spatial perception × Departure time					0.337	0.085 *
Terrain spatial perception × The time saved					0.395	0.037 **
Terrain spatial perception × The cost saved					0.099	0.703
N	448.000		448.000		448.000
Prob > chi2	0.000		0.000		0.000
AIC	542.447		520.546		492.347
Pseudo R2	0.114		0.168		0.240

Note: * p < 0.1, ** p < 0.05, *** p < 0.01.

.

Model 1 only incorporates personal socioeconomic attributes and travel characteristics into the model to analyze motorists’ P&R usage behavior. Model 2 covers the objective built environment and subjective terrain spatial perception of river–valley-type cities based on Model 1. Model 3 considers the moderating effect of terrain spatial perception on other factors influencing motorists’ P&R behavior. The final results are presented in Table 7. According to the model results, the significance values of the likelihood ratio test are all less than 0.001, indicating that the three established models are significant [48]. The goodness of fit of Model 1, Model 2, and Model 3 increases successively, and the AIC value of the models decreases successively. This suggests that Model 3 has more explanatory power for motorists’ P&R usage behavior in river–valley-type cities [49].

In Model 1, travelers’ gender, drive age, degree of freedom in working time, the number of private cars owned, the payment method of traffic cost, the experience of using P&R, departure time, time saved, and cost saved by using P&R all have significant effects on P&R use. Among them, gender, driver age, degree of freedom in working time, the payment method of traffic cost, and the experience of using P&R have negative effects on the use of P&R. The number of cars owned, departure time, time saved, and cost saved by P&R have positive effects on P&R usage.

After including the subjective and objective built-environment attributes in Model 2, gender becomes relatively insignificant. However, the significance and influence direction of other factors are consistent with Model 1. Additionally, the objective built environment of river–valley-type cities, such as commute distance, the road slope, the demand of river crossing, group characteristic, and travelers’ subjective terrain spatial perception, has significant positive effects on the use of P&R.

Model 3 is the model to analyze the moderating effect. After Model 1 and Model 2 both prove that the main effect has a significant impact on the use of P&R, we only need to focus on the significance of the moderating effect. The results show that there is no moderating effect of the cost saved by P&R on terrain spatial perception influencing the use of P&R; the other six moderating effects exist.

In order to further investigate how terrain spatial perception moderates the impact of significant influencing factors on P&R usage behavior, moderating effect figures are created based on high and low terrain spatial perception [50]. Figure 5 illustrates how the influence of different factors on P&R usage behavior is moderated by motorists’ terrain spatial perception in river–valley-type cities.

5. Discussion

The purpose of this study is to determine the various factors influencing P&R use in urban areas and to investigate how motorists’ spatial perception of special terrains moderates the factors affecting P&R usage behavior in a river–valley-type city.

Our research led to some interesting conclusions. According to the results of Model 1 and Model 2, motorists’ driving age, degree of freedom in working time, ownership of private car, transportation payment, and the experience of using P&R have significant effects on motorists’ P&R usage behavior in river–valley-type cities. Transportation costs paid by companies, including fuel and parking fees, serve as a significant disincentive for motorists to use P&R to commute [51]. Unlike some studies, monthly income does not have a significant effect on P&R usage behavior. We speculate that companies with higher-monthly-income motorists may offer higher transportation subsidies. Even with the use of P&R, there may not be a significant transportation fare gap to entice higher income motorists to transfer to P&R commuting. Compared with driving all the way, the time and cost saved by using P&R have a significant impact, which aligns with the findings of previous studies [11,12,13,14]. In Model 1, gender has a significant effect on P&R use behavior. Women play a more important role in the family. Therefore, female motorists are more likely than men to choose P&R for commuting after dropping their children off at school. The peak period of river–valley-type cities is longer; motorists with less degree of freedom in working time and departure during peak hours will bear a greater risk of being late, so they are more inclined to commute by P&R [52]. Different from previous studies, we believe that more private car ownership reflects motorists’ travel preferences to some extent, so they still choose to commute by car even in the terrain background of river–valley-type cities [33]. Due to the complex terrain of river–valley-type cities, motorists with a lower driving age obviously face higher driving pressure and turn to P&R commuting [29,53]. Similarly, the experience of using P&R clearly has a significant impact on motorists’ P&R usage behavior [29,53].

The subjective and objective built environment proposed for river–valley-type cities in this paper has a significant positive impact on the motorists’ use of P&R, which is also the part to be emphasized of our study. In cities of the same scale, river–valley-type cities exhibit a greater degree of separation between residence and workplace, steeper road slopes, and more distinct group characteristics. Additionally, there is an imbalance in traffic development between the outskirts and the center of the urban group. Moreover, rivers flow parallel to the main axis of the city. As a result, travelers using ground traffic mode in these cities will inevitably need to find bridges to cross the river, leading to longer detours. Therefore, motorists in river–valley-type cities are more inclined to choose P&R.

As far as the security perception of motorists using P&R is concerned, parking spaces in river–valley-type cities are scarce, and motorists find it difficult to locate parking spots. Parking lots of P&R have comprehensive lighting and monitoring facilities, which enhance security for vehicles [54]. In addition, the unique terrain of the river–valley-type city increases the risk of accidents. To address this issue, a pedestrian-friendly parking lot is established around the P&R parking lot. An underground express street crossing connects the parking lot to the subway ride point, mitigating these risks and ensuring the security of individuals traveling between the parking area and the subway station. Compared to driving all the way, P&R can help passengers arrive at their destination more punctually, providing the motorist with greater peace of mind [10].

Peak hours in river–valley-type cities are perceived to last longer than in plain cities. Additionally, P&R is less affected by peak-hour congestion compared to driving all the way. In river–valley-type cities, road traffic motorists have more detours to cross the river compared to motorists in plain cities of the same scale, but P&R is not restricted by this. River–valley-type cities face restricted road resources due to terrain and challenges with parking in the city center. P&R reduces the time for car owners to search for parking spaces [55]. The group characteristics of river–valley-type cities are obvious, and there is low mixed land use. Routes may require detours if additional services are needed when driving all the way. Having additional services like restaurants and convenience stores near P&R transfer points can better cater to the needs of motorists.

In terms of accessibility, river–valley-type cities exhibit an imbalance in traffic development compared to plain cities. The transportation network and facility development at each group center are superior to those at each group edge. This imbalance results in variations in the convenience of cross-group and intra-group road travel. In contrast, the subway is not limited by the development of ground transportation facilities and has higher speed, allowing it to reach the destination faster. There is a problem of parking difficulty in the group centers of the river–valley-type city, and the P&R reduces the car owners’ long search time for parking spaces by parking the car in the subway P&R parking lot away from the group center, and the traffic around the parking space is better. The complex network of river–valley-type cities, with numerous broken roads and one-way roads, and the necessity for residents to pass through the city center to travel between the two ends of the city contribute to an increase in unnecessary detours. Therefore, using P&R is obviously more accessible than driving all the way. In addition, subways in river–valley-type cities are not affected by road traffic congestion, detours, or slopes in roads. They operate at a higher speed, allowing passengers to reach their destinations faster.

In river–valley-type cities, commuting via P&R is less affected by the steep slopes compared to driving all the way. River–valley-type cities have more intricate road conditions and longer average commuting distances compared to plain cities of similar size. Therefore, driving in these cities can impose greater pressure on motorists. In addition, the peak hours in river–valley-type cities tend to last longer. As a result, motorists may spend more time in their cars, increasing the risk of experiencing psychological pressure due to boredom and impatience. This can ultimately lead to a less comfortable commuting experience.

In addition to the fact that terrain spatial perception has no moderating effect on the impact of cost saved by P&R, it does have a moderating effect on the impact of workplace and residence distance, river crossing demand, road slope, group characteristics, departure time, and time saved on motorists’ behavior of using P&R. According to the coefficients of the interaction terms, terrain spatial perception has a significant positive moderating effect on the impact of work–residence distance, river crossing demand, road slope, group characteristics, departure time, and time saved on promoting motorists’ P&R behavior. In other words, motorists’ understanding of terrain spatial perception will enhance the impact of these factors on P&R behavior.

6. Conclusions

6.1. Contributions

In general, this research contributes in the following ways. Firstly, both the objectively built environment factors and motorists’ subjectively perceived terrain spatial factors proposed for the characteristics of river–valley-type cities have significant impacts on P&R use behavior to varying degrees. This highlights the importance of considering both the subjective and objective built environments of cities when studying P&R use behavior. Secondly, there are few studies on the use of P&R behavior in river–valley-type cities, especially those considering motorists’ perception of terrain space. Therefore, this part of the study lacks a mature scale for reference. We demonstrate that the latent variables representing the four aspects of terrain spatial perception, as constructed in this study, exhibit strong explanatory power. Each question effectively reflects the corresponding subjective terrain-spatial-perception latent variables, offering a scale reference and theoretical foundation for other river–valley-type cities to study P&R use behavior. In addition, we combined the advantages of SEM and Logit models to establish a SEM–Logit model that comprehensively considers the subjective and objective built environments of river–valley-type cities. Compared with models that do not consider both built environments, this model can more accurately explain and predict the usage behavior of P&R. Finally, we established a model with interactive terms to verify the significant moderating effect of motorists’ subjective terrain spatial perception on the other factors of P&R use behavior in river–valley-type cities.

6.2. Suggestion

The research results of this paper have significant implications for the planning and construction of built environments in valley-type cities. We advocate for enhancing the mixed use of communal land and decreasing the distance between workplaces and residences in the short and medium term. This approach can effectively lessen motorists’ reliance on cars. Rationally plan subway stations in areas with high road slopes and enhance the service level of underground walkways and overpasses connecting parking points to subway boarding points and exit stations. Design parallel subway lines on both sides of the river along the main axis of a river–valley-type city and construct fishbone branch lines perpendicular to the main axis on the two parallel lines; this can minimize the need for detours after exiting the subway station. Placing P&R parking lots at the junction of the main line and branch line may also be helpful. In contrast to plain cities, a city’s outskirts in a river–valley-type city are distinct from its city center. Therefore, the placement of P&R sites could be contemplated at the intersection of each city group, and additional entrances of subway stations could be strategically added at the junctions and the center of the city group. Implementing preferential Park and Ride rates during peak commuting hours will enhance the competitiveness of P&R, decrease the volume of cars entering the city center during peak hours, and reduce peak congestion duration in river–valley-type cities. In addition, enhancing the perception of the actual P&R experience in terms of security, convenience, comfort, and accessibility can encourage motorists to use P&R effectively. This approach maximizes motorists’ needs and preferences, enhances motorists’ subjective well-being, and improves travel quality. Finally, the use of P&R in river–valley-type cities will be less affected by ground traffic conditions compared with driving all the way. Factors such as road slopes, detours to cross bridges, less developed traffic infrastructure on the outskirts of the city’s center, and peak congestion at major intersections will enhance the attractiveness of P&R. Individual terrain spatial perception not only significantly influences P&R use behavior directly but also moderates the impact of other factors on P&R use behavior. Therefore, enhancing motorists’ positive perception of P&R in river–valley-type cities is crucial for improving motorists’ travel experience and promoting the use of P&R.

6.3. Limitations

The study was limited in two ways. Few scholars have studied the terrain spatial perception of motorists using P&R in river–valley-type cities. Although the terrain-spatial-perception index proposed in this paper has been proven to be applicable to this survey data, there is no reference for a mature scale. Therefore, based on the terrain-spatial-perception-index system proposed in this study, it is essential to adjust the index according to the specific investigation circumstances of various river–valley-type cities, so as to obtain the most targeted results and helpful conclusions. In addition, the construction of P&R in river–valley-type cities generally started late, and the P&R system in the study area is in the initial stage of continuous development and evolution. Therefore, when the P&R system reaches a mature stage, it is essential to replicate this experiment with fresh research data to guarantee the real-time validity of the results and conclusions. This is what we will do in the future.