Access the full text.

Sign up today, get DeepDyve free for 14 days.

Journal of Advanced Transportation
, Volume 2023 – Feb 28, 2023

/lp/hindawi-publishing-corporation/exploring-the-spatial-variation-of-access-egress-distances-of-subway-oFG0wEoKFW

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

- Publisher
- Hindawi Publishing Corporation
- ISSN
- 0197-6729
- eISSN
- 2042-3195
- DOI
- 10.1155/2023/8008667
- Publisher site
- See Article on Publisher Site

Hindawi Journal of Advanced Transportation Volume 2023, Article ID 8008667, 17 pages https://doi.org/10.1155/2023/8008667 Research Article Exploring the Spatial Variation of Access/Egress Distances of Subway Stations Using Mobile Phone Positioning Data in Chengdu, China 1 1 2 1 1 Chao Wang , Zhongquan Qiu , Renbin Pan , Xiaojian Wang , and Yusong Yan School of Transportation and Logistics, Southwest Jiaotong University, Chengdu 611756, China School of Business Administration, Southwestern University of Finance and Economics, Chengdu 611130, China Correspondence should be addressed to Zhongquan Qiu; qiuzhongquan@swjtu.edu.cn Received 3 June 2022; Revised 26 January 2023; Accepted 13 February 2023; Published 28 February 2023 Academic Editor: Maria Vittoria Corazza Copyright © 2023 Chao Wang et al. Tis is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Te distance from the origin or destination to or from the subway station is defned as the access or egress distance, which determines the service coverage of the subway station. However, little literature studies the distances at the station level, and they may vary from station to station. Terefore, this study aims to explore the infuencing factors and spatial variation of the distances at the station level by using the mobile phone positioning data of more than 1.2 million anonymous users in Chengdu, China. First, this study proposes a method to extract the access and egress trips of the subway. Next, the ordinary least squares (OLS) regression models are carried out to select the signifcant explanatory variables. Finally, the geographically weighted regression (GWR) models are used to model the spatial variation relationship between the 85th percentile access/egress distances and the selected explanatory variables. Te results show that diferent stations’ access/egress distances vary signifcantly in space. Hotel, residence, life, fnance, road density, and mixed land use are found to be negatively correlated with distances, while education, 36–45 years old, male, and high education are positively correlated. In addition, the GWR model reveals that the infuence of explanatory variables on access/egress distance varies from space to space. Te results further promote the understanding of the existing system and provide a relevant reference for planners and transportation departments to optimize land use and public transportation planning. from automobile to subway [3]. Naturally, the construction 1.Introduction of subway stations must be within a reasonable distance to be Te subway in China is experiencing rapid growth to cope convenient for residents to use. However, an important issue with new problems related to urban expansion, trafc is how to defne an acceptable and practical distance to walk, congestion, and air pollution [1–3]. At the same time, the bicycle, and other modes of transportation to conveniently service quality of the subway is constantly improving, with access/egress the station for most subway passengers and potential passengers from their homes, workplaces, schools, longer service time and higher service reliability [4]. In addition to eforts to improve the service quality of the and other locations. Te answer to this question can help subway itself, providing a high-quality travel environment provide information for planning and design decisions of around stations is another way to increase its attractiveness. TOD’s scale and geographical scope. Terefore, some cities vigorously develop transit-oriented In recent years, more and more scholars have begun to development (TOD) projects based on compact, mixed-use, pay attention to studying the access/egress distances of pedestrian- and bicycle-friendly urban construction con- subway stations. Te proportion of the population served by cepts to realize the efective strategy of changing the mode of the transportation system is a crucial indicator of system short-distance into walking and bicycle and long distance performance. Terefore, when determining the service area 2 Journal of Advanced Transportation egress distances of subway stations at the station-level around the station, most studies usually use the over- simplifed method that most people walk 800 meters to the change in space? (2) What are the infuencing factors of the access/egress distances? (3) How does the infuence of station to defne it [5, 6]. However, the accuracy and ap- propriateness of this one-size-fts-all approach are often infuencing factors on the access/egress distances change in questioned in other studies [7]. Because people usually walk space? Tese results can provide a helpful reference for to the subway station, also by bicycle, automobile, and bus. planners and city managers to optimize new subway stations In Sydney, Australia, only 50% walk, 34% by automobile, and transportation systems. In particular, the Chengdu rail and 14% by bus [5]. In Beijing, China, 65.25% walk, 19.91% transit group is vigorously developing the TOD project, and by bus, 14.85% by bicycle, and 5.28% by automobile [8]. In a large-scale analysis of urban residents’ activities in Chengdu can determine the applicable distance threshold Toronto, Canada, buses and trams connected to the subway account for more than one-third of all passengers [9]. Most for the urban environment with a unique background. In addition, this study also shows the potential of mobile phone studies also admit that walking is a better choice for short distances, but within the acceptable travel time, using bi- positioning data in exploring residents’ subway access/egress distances, which can provide a certain degree of explanation cycles as the connection mode makes the subway more attractive and expands the subway station’s service area. For for the activity intentions of diferent groups of people. example, Zuo et al. [10] indicated that the bicycle distance is Next, the literature on access and egress distances of the 1.7 to 2.3 times walking in Shanghai. Lee, et al. [11] found subway is reviewed. Te third section describes the study that bicycles can expand urban service areas from 29.9% to area and data. Te fourth section introduces the method. Te 93.6% in Seoul. Other modes of transportation also increase ffth section is the explanation and discussion of the results. the accessibility of the subway system, but there are apparent Finally, the last section is the conclusion. variations in diferent ways. Terefore, the planning of subway stations needs to consider the connection of various 2.Literature Review trafc modes. Te lack of appropriate data is one of the main bottle- Tis section reviews the research related to this topic in necks in transportation research, which signifcantly afects recent years, mainly divided into two aspects: the access/ the accuracy of the results. Existing research often relies on egress distances of the subway for diferent modes of the overall inference from census data or estimates the transportation and the infuencing factors of access/egress distance at the level of individual trips based on small sample distances. surveys and trajectory data. However, the low diversity and quantity of data are still challenging. More importantly, subway stations’ access and egress distances may vary from 2.1. Access/Egress Distances of the Subway. As the most station to station and from space to space. Tis has not been critical factor afecting the quality of the mode exchange, fully revealed in the existing research. At present, new data researchers have perfomed much work in this area, espe- sources related to information and communication tech- cially for walking into subway stations. However, the dis- nology have emerged: mobile phone positioning data. Tese tance thresholds obtained from diferent studies vary due to data are based on the records generated by the interaction the diferences in study areas and data sources. For example, between the mobile phone and the points of interest (POI) on the one hand, El-Geneidy et al. [7] found that the average and the timing report every half hour under a continuous walking distance of the subway from home is 0.564 km, and bright screen condition. Mobile phone positioning data can the 85th percentile distance is 0.873 km through the Mon- record users’ arrival and departure times at specifc places treal OD survey in Montreal, Canada. On the other hand, (such as subway stations). More importantly, the basic Daniels and Mulley [5] determined the longer walking demographic information is stored in these data, which can distance through the family travel survey in Sydney, Aus- be used to understand the travel characteristics of various tralia, with an average of 0.805 km and the 75th percentile of user groups and reveal the observed behaviors. In this way, 1.018 km. Given that the distances between access and egress we can realize the complete end-to-end travel inference to are not clearly distinguished in previous studies, Wang and detect human movement behavior’s macro and micro levels. Cao [12] analyzed the walking egress distances through the Based on this background, this paper explores the access 2010 Transit Onboard Survey in the Minneapolis and St. and egress distances of subway passengers for the frst time Paul Metropolitan Area. Tey concluded that the average using mobile phone positioning data of 1.2 million users in distance is 0.494 km, and the 85th percentile distance is Chengdu, China, which has a larger data sample and is closer 0.845 km. Later, Tao et al. [13] examined the distances be- to the real travel conditions of residents than the ques- tween home and subway stations in the same city using the tionnaires and other data used in previous studies. In fact, 2016 transit on-board survey and found a shorter average the 85th percentile access and egress distance of each subway walking distance of 0.317 km. Under the Chinese back- station is an important indicator to evaluate the accessibility ground, He et al. [14]’s questionnaire survey of Nanjing of subway services. Terefore, this paper proposes using the Metro showed a longer distance and walking distances to the geographically weighted regression (GWR) model at the subway station range from 1.050 to 1.2 km. station level to reveal the infuencing factors of the spatial Te bicycles have the potential to promote the use of variation of subway stations’ access/egress distances. It at- subways by connecting stations with origins or destinations. tempts to answer three questions: (1) How do the access and In recent years, the transfer between bicycles and subways Journal of Advanced Transportation 3 and more academic studies have shown the potential of has become a research hotspot. Recent studies also show that bike-sharing has expanded the catchment area of subway large-scale data as an alternative source of travel behavior information, which can be used to derive the origin- stations, but the extent is various. For example, Rastogi and Krishna Rao [15] found that the access distances of subway destination matrix. For example, some studies have ana- stations are 1.8–4.05 km by investigating the operation lyzed the transfer distances in bike-sharing utilizing the GPS system in Mumbai, India. However, Pan et al. [16] examined and order data about Mobike in Shanghai. Te results the access distances using questionnaire surveys in Shanghai showed that bicycle distances have increased compared with and concluded that more than 70% of bicycles travel within walking distances [19, 20]. However, these current studies 1.5 km, and only about 5% are more than 2.5 km. By mainly focus on the distance of a single mode of trans- portation. More importantly, it is not known whether users comparing the distance variations between diferent cities, Hochmair [17] proposed that the service area of stations in of bike-sharing really transfer to/from the subway. Tus, the usage pattern lacks a comprehensive exploration based on Los Angeles is 2-3 times that of Atlanta and the Twin Cities. In addition, the median access distances observed are within sufcient data covering diferent modes of transportation. the bufer radius of the proposed community hub (1.609 km) and the gateway hub (3.218 km). As an essential part of 2.2. Factors Infuencing the Access/Egress Distances. Many residents’ mobility, commuter access can signifcantly refect factors may afect people’s use of the subway, which leads to the service quality of transportation. According to surveys in signifcant diferences in the access/egress distances of the the Seoul Metropolitan and Deajeon Metropolitan Areas, subway in various environments. Terefore, the relation- Lee et al. [11] found that the distances from home to the ships between distances and critical factors such as the built station and from the station to work are estimated to be environment and user characteristics must be fully 1.96 km and 2.13 km, respectively. To get a more accurate understood. value, Zuo et al. [10] used Cincinnati GPS-based household travel survey data and examined the distance threshold of bicycles, which was more than twice that of walking (4.36 km 2.2.1. Built Environment. Te built environment refers to vs. 1.30 km). the artifcial environment provided for human activities, Besides walking and bicycle, the distances of other including various forms of buildings (such as residential, modes of transportation are also compared. For example, industrial, and commercial), infrastructure (such as trans- Wang et al. [8] compared several main modes of trans- portation and parks), and urban space [21]. It is generally portation in Beijing, indicating that the average distance of believed that the built environment has a signifcant infu- walking is the lowest, which is 0.43 km, while the others are ence on shaping the mode of human mobility and activities, in order: the bicycle is 1.452 km, the bus is 6.262 km, and the which may be directly related to their accessibility to subway automobile is 9.115 km. Xi et al. [9] analyzed the Trans- stations and perceived convenience [22, 23]. In particular, portation Tomorrow Survey in Toronto, Canada, indicating when there is no environment suitable for pedestrians, that subways’ service areas connected with buses and trams people’s decision to drive instead of walking to the station are critical because they account for more than one-third of can be afected [24]. Some studies have found that high all passengers. In space, the radius of the pedestrian service density at intersections and roads positively correlated with area is generally less than 1.609 km, while the bus, tram, and walking distances. Tey also indicated that population automobile are often many times larger. density negatively correlated with walking distances [7, 13]. Most of the above studies are based on the data collected In addition, these factors are negatively correlated with from traditional travel surveys, such as questionnaire-based bicycle distances [17]. Unlike previous studies that take interviews and travel OD surveys. Access/egress distances are individuals as the analysis unit, Lin et al. [20] established usually obtained from the participants’ own reports, and are a regression model at the station level. Teir results showed assumed to be capable of remembering (or willing to share) that the subway stations’ catchment area is positively cor- the actual activity route and movement distance. However, related with the distance to the city center but negatively Weinstein Agrawal et al. [18] indicated that only half of the correlated with the density of subway stations. Later, Li et al. people’s actual distances are similar to those they remember. [19] used the same level to investigate the relationships Although GPS data can provide the most accurate spatial between the 85th percentile distances of diferent subway trajectory of personal movement patterns, they cannot be stations and the built environment. Many built environment used on a larger scale [13]. Both travel surveys and personal factors are related to distances, but their relationships show GPS are small samples of nonpopulation data. Tese data some variations in space. Generally speaking, these studies limit the sample’s geographic and demographic coverage, reveal the infuence of the built environment on the walking making them challenging to refect distance patterns fully. or bicycle distances of subway stations, which provide ex- With the development of information and communi- tensive enlightenment for the regional planning of the cation technology, large-scale data about human spatio- station. temporal motion trajectories can be obtained from many sources, such as transportation network companies and social media data. Using new data, researchers have the 2.2.2. User Characteristics. Scholars generally believe that opportunity to solve the traditional problem of distance demographic characteristics signifcantly impact the dis- calculation caused by limited samples. Furthermore, more tance to the subway station. However, there are apparent 4 Journal of Advanced Transportation diferences in the direction and degree of personal char- acteristics (including age, gender, income, and other social factors). Te representative point of view is that there are diferences in walking distances in terms of age. Young people are more likely to walk to the station and have a longer walking distance [7, 14, 24, 25]. In terms of gender, 025 50 County km males tend to walk or bicycle for a longer distance to the station than females [7, 25], but He et al. [14] thought that there is no diference between genders. Since car ownership, family income, and family size also have a negative impact on walking distances because families with more vehicles, higher income, and more members are more likely to choose to travel by car but less likely to live near the subway [7, 24, 25]. When considering the travel purpose, compared with shopping travelers, working travelers have the most Legend extended walking distances and the highest possibility of Subway station walking to the subway station [14]. For users in bike-sharing, Ma et al. [26] found that the distances between males are Subway line Road higher than that of females, and urban residents are shorter than suburban residents. In addition, they also analyzed the possible infuence of time on distance. Considering the Figure 1: Study areas and the subway system in Chengdu. impact of travel habits on distances, Lin et al. [20] proposed that the higher the proportion of a single user of the subway Longquanyi, Pidu, Qingbaijiang, Qingyang, Shuangliu, station, the greater the service distance. Wenjiang, Wuhou, Xindu, and Xinjin. At the end of 2019, the resident population of Chengdu has reached 16.581 million (https://gk.chengdu.gov.cn/govInfo/detail.action? 2.3. Summary. Although more and more evidence shows id=2576335&tn=2). With the vast population, rail transit that the distance between diferent modes of travel is dif- in Chengdu has to develop rapidly to relieve the travel ferent, due to the limitation of data sources, the existing pressure of the metropolis. Since September 27, 2010, literature mainly studies the distance of a single connection Chengdu metro line 1 has been put into operation. By the mode or the distance of individual travel levels. Although end of 2020, it has a relatively developed rail transit network. some existing literature has studied access and egress dis- Tere are seven lines and 202 subway stations in operation, tances at the station level, very little literature has explored with a total length of about 518 km and an average daily how distances vary spatially at the station level and the passenger fow of 3.75 million. At the same time, to bring existence of spatial correlations of distances. In addition, a better living and traveling environment to residents, the there are signifcant diferences in the direction and degree Chengdu rail transit group is now vigorously developing of key factors afecting distance in diferent studies. TOD projects with high density, multifunction, pedestrian- Terefore, the discussion on policies or plans to measure the friendly environments, and high quality (https://www. scope of subway services is limited. In order to fll these chengdutod.com/#home1). research gaps, this paper tries to use the geographically weighted regression (GWR) model to explore the spatial change of subway access and egress distances at the station 3.2. Mobile Phone Positioning Data. Mobile phone posi- level. Te variable coefcient in the GWR model is allowed tioning data are obtained from Jike (https://www.isjike.com/). to change with space, while the variable coefcient in general Tere are 4.994 million users in Chengdu, accounting for ordinary least squares is fxed, eliminating the spatial au- 30.12% of the 16.581 million permanent residents at the end tocorrelation of variables [22, 27–29]. Terefore, the GWR of 2019. Te average number of active users per day is 2.228 model is more suitable for analyzing the spatial change of the million. Te users used in this paper are 1,210,252 users distances in diferent subway stations. randomly selected from the total users. Tere are about 301,420,967 records, which are the continuous trajectory of sample users from October 15 to November 15, 2020. Te 3.Study Areas and Data data for this time were chosen because the weather during this 3.1. Study Areas. Tis study focuses on Chengdu, the capital period was relatively mild and more suitable for travel, and of Sichuan Province, a high-tech industrial base, a com- the results of the study would be more representative. Mobile mercial logistics center, and a comprehensive transportation phone positioning data are divided into two parts. Te frst hub in western China. Te whole city consists of 12 mu- part is scene data with POI records, with a total of 88,940,033 nicipal districts, three counties, and fve county-level cities, records. Te main felds of scene data include user id, arrival with 14,335 km . Te study areas are shown in Figure 1. Te time, longitude, latitude, departure time, scene classifcation, main urban areas of Chengdu are mainly composed of 12 and the exact name of the POI. Table 1 is an example of scene municipal districts, namely Chenghua, Jinniu, Jinjiang, data. Te data is generated by the interaction between the Journal of Advanced Transportation 5 Table 1: Example of scene data. User id Arrival time Longitude Latitude Departure time Classifcation POI name 3d5h### 2020/11/06 16 : 46 : 40 104. ### 30. ### 2020/11/06 16 : 49 :16 Subway station Dongpo road station 3d5h### 2020/11/06 16 : 49 :12 104. ### 30. ### 2020/11/06 16 : 51 :12 Subway station Cultural palace station 3d5h### 2020/11/06 16 : 51 : 28 104. ### 30. ### 2020/11/06 16 : 54 : 51 Subway station Jinsha museum station 3d5h### 2020/11/06 16 : 55 : 31 104. ### 30. ### 2020/11/06 16 : 59 : 54 Science education culture Jinsha site museum ### is used to replace more detailed information for privacy reasons. software development kit inside the mobile phone and the Table 2: Example of timing report data. POI scene through wireless fdelity, Bluetooth, and near feld communication and calibrated by innovative “intelligent User id Time Longitude Latitude scene recognition” technology (https://www.cdstats.chengdu. 3d5h### 2020/11/06 19 : 56 : 07 104. ### 30. ### gov.cn/htm/detail_180293.html). Te identifcation standard 3d5h### 2020/11/06 20 : 50 :16 104. ### 27. ### is to require identifcation at the front gate of the subway, and 3d5h### 2020/11/06 22 : 05 : 57 104. ### 30. ### 3d5h### 2020/11/06 22 : 39 : 51 104. ### 30. ### the farthest 30 meters is the identifcation range. Te shops are identifed based on the entrance of the shops, and the ### is used to replace more detailed information for privacy reasons. range within 5 meters of the entrance is the efective iden- tifcation range. According to the “intelligent scene recog- a complete subway trip is defned as arriving at a subway nition” technology, when a user enters or tends to enter a POI station from the origin through a certain mode of trans- scene (such as the subway station and shop), the entry records portation (including walking, cycling, self-driving, and so are generated, and when a user leaves the scene, which records on), taking the subway through at least two subway stations, the time, latitude, and longitude, POI name, and other user or leaving the station to reach the fnal destination through information at this moment. Terefore, the user’s stay time in another mode of transportation. It is important to note that a scene can be calculated. Te second part is the timing report because of data quality limitations, we are not able to identify data, with a total of 212,480,934 records. Te main felds the mode of transportation by which users arrive or leave the include user id, time, longitude, and latitude. An example of subway station. timing report data is shown in Table 2. Te generation Next, trip extraction and screening from Steps 1 to 3 in principle of timing report data is that a report record is Figure 2 are introduced. For a similar trip extraction pro- generated every half hour when the mobile phone continu- cedure, readers can refer to Wang et al. [29] ’s paper. ously lights up. (i) Step 1: Clean the timing report and scene data. First Mobile phone positioning data also contain information of all, some scene data cannot identify detailed POI, of each user, such as age, gender, education level, and in- so they are recorded as timing report data by come level, which is mainly judged by a combination of mistake. For this reason, the timing report data integrating real samples with Jike company’s in-depth whose time interval with the previous record is less partners, users’ APP online usage characteristics and of- than 30 minutes are deleted. Second, when the user line visiting behaviors. Te specifc sources of user char- approaches some subway stations built on the acteristics are as follows: one of them is from information ground, scene data may be generated, but users do such as the APP list. By analyzing the APP list installed on not actually enter the station and take the subway. mobile devices and the usage as well as the reference feature Terefore, when the distance between two adjacent labels, user characteristics are analyzed and judged. For recorded subway stations is more than 5 km, or the example, in the inference of gender, the typical applications time interval is more than 10 minutes, it is con- of the APP are the great aunt, male health care, and male sidered that the subway is not used to reach the private doctor. Te second comes from the type of mobile adjacent stations, so these subway records are phone information pushed, the frequency of pushing, and so deleted. on. Te third comes from the location information such as (ii) Step 2: Trip extraction under space-time constraints. the user’s residential address and ofce address resolved In this step, the origin and destination of the access based on the latitude and longitude of the user’s activities to and egress trip are judged according to the moving identify the income level. Te fourth comes from external and staying state of the continuous track of the data sources such as UnionPay and operators, as well as scene and the timing report data. Considering the public data on the Internet. data structure of this study (the principle of gen- erating data every 30 minutes by timing report 3.3. Mobile Phone Positioning Data Processing. Given that data), and referring to the related literature on travel the mobile phone positioning data only records the discrimination based on mobile phone cellular movement trajectory of the user in continuous time, it is network data [30, 31], this study assumes that users stay at a position for 30 minutes, and the position is impossible to know the actual origin and destination of the subway trips. Terefore, the trips from the origin to the regarded as a staying point (origin or destination of the trip). For scene data, when the time diference subway station and leaving the station to the destination need to extract according to some principles. In this study, between the user’s arrival and departure time in 6 Journal of Advanced Transportation a POI is more than 30 minutes, it is marked as a staying point. For timing report data, given that 1. Timing report data: adjacent records should be less than 30 minutes. these data only record the time, longitude, and latitude of every half hour when the screen is continuously illuminated but do not know the 2. Scene data: adjacent subway scene should be less than staying time of the user in this position. Terefore, 5 km or less than 10 minutes. a timing report record is assumed that the user stays in the position for half an hour. For the sake of travel safety, people rarely use their mobile phones for 1. Scene data: Te location of the POI whose stay time 30 minutes continuously under a bright screen exceeds 30 minutes is judged as a staying point. when they move continuously by walking or cycling. In the case of taking the subway, the records gen- 2. Timing report data: Te location recorded by a report erated during the subway trip are mainly in the ABA data is considered as a staying point. (If a report data is and AAB forms (A is the subway station, and B is judged to be moving, delete the data). the timing report record). If the time interval be- tween A and B is less than 30 minutes, B is regarded 3. Extraction trips: the trips from origin to the subway station and from the subway station to the destination. as a nonstay point and excluded. Ten, for the records marked as staying points, the trips from the staying point (origin) to the subway station and from the subway station to the staying point (destination) are extracted as access and egress trips. 1. Delete the non-Chengdu trips. (iii) Step 3: Trip screening. After extracting subway trips in step 2, the abnormal trips must be deleted 2. Delete the trips that are not in subway operation time according to the following rules. First, the trips or the time interval of the trips is greater than 12 hours. whose origin or destination does not belong to Chengdu are eliminated. Second, the trips that are 3. Use the third quantile of box chart to eliminate the not in the subway operation time (00 : 00–6: 00) and trips with abnormal travel distance. the time diference between the origin (or desti- nation) and the station of more than 12 hours are Figure 2: Te extraction process of access and egress trips. deleted. Tird, the outliers of the distance of access and egress trips are eliminated by using the three times quantile of the box diagram. data are obtained from OpenStreetMap (OSM) (https:// www.openstreetmap.org/). Te bus stop data are obtained According to the previous three steps, the number of from the Chengdu public platform, including 10,228 bus users extracted from the original data is 166,913, and the stops. Bus stop data record each station’s station name, access and egress trips are 840, 312 and 763, 086, re- longitude, latitude, line number, and line direction spectively. Tese trips are used as follow-up analysis. (https://www.cddata.gov.cn/oportal/index). Population data are obtained from WorldPop, counted at the grid 3.4. Variable Description. According to the relevant liter- level of 100 ×100 m in 2020 (https://www.worldpop.org/). ature and available data [7, 12, 13, 19, 20], we have selected Tis study takes Tianfu Square in Chengdu as the center two categories of independent variables that can explain and calculates the distance from each station to the city the dependent variables, namely the built environment center. Finally, user characteristics are the proportion of and user characteristics. Teir descriptive statistics are diferent users in each station according to access and shown in Table 3. Te built environment is the statistical egress trips. In order to reduce repeated displays, the value of various variables within the 1 km (approximately statistical values of user characteristic variables in the equal to the average distance of all the extracted trips in access trips are only presented in Table 3. Section 3.4) bufer zone of the subway station, mainly including land use characteristics, trafc-related facilities, 4.Method and other variables. Te land-use variables are calculated using POI data. POI data are collected from Amap (also In this study, the spatial regression model GWR was used to known as Gaode Map) through the application program explore the spatial variation relationship between the 85th interface (https://www.amap.com/). Te total number of percentile access and egress distance and travel-related POI is 416,459. Each POI record usually contains the POI variables and the above selected built environment vari- name, address, scene classifcation, longitude, and latitude ables. First, the ordinary least squares (OLS) regression of the specifc location. Some scenes are too few and have model was used to explore the relationship between ex- been deleted. Finally, according to the scene classifcation, planatory variables and the distances. Ten, to analyze the POI is mainly divided into 13 categories for research. spatial change of their relationship, we calculated Moran’s I Parking lot data are extracted from POI. Road network to test the existence of spatial correlation of variables. Step 3: Trip screening Step 2: Trip extraction Step 1: Data cleaning Journal of Advanced Transportation 7 Table 3: Description and statistics of variables. Variables Description Mean SD Built environment (within the 1 km bufer zone) Hotel Te proportion of hotel services (chain hotels, service apartments, and so on) 94.089 169.811 Sport Te proportion of sports facilities (gym center, playground, and so on) 64.470 60.775 Public Te proportion of public facilities and services (bridges, intersections, and so on) 16.436 16.698 Company Te proportion of companies (enterprise, factory, and so on) 16.436 16.698 Te proportion of residences (a residence is a place used as a home or dwelling, Residence where people reside, mainly including commercial residential buildings, and 120.267 147.511 residential communities) Te proportion of life services (life refers to the places or facilities that provide life services to ensure the normal operation of people’s daily activities such as express Life 158.386 148.263 delivery, laundry, mainly including post ofces, travel agencies, logistics and express delivery, and laundries) Healthcare Te proportion of medical services (general and special hospitals, clinics, and so on) 116.297 115.913 Te proportion of government (administrative agency, civil service, welfare Government 77.797 97.841 institution, and so on) Te proportion of science and education culture (college, university, high, middle Education 118.139 127.856 and primary school, and so on) Te proportion of shopping malls (retail, shopping mall, convenience store, and so Shopping 19.842 16.792 on) Te proportion of fnancial services (banks, insurance companies, fnancial Finance 58.535 68.825 corporations, and so on) Tourist Te proportion of tourist attractions (parks, museums, historical sites, and so on) 6.426 9.119 Te proportion of catering services (Chinese and foreign restaurants, fast food, Catering 280.460 253.100 cofee house, and so on) POIMix Shannon entropy of all POI categories 3.164 0.248 Road density Length of all roads (km) 37.065 13.775 Bus station Number of bus stations 33.480 23.420 Parking lots Number of parking lots 160.020 161.447 Population Te number of people 42868.607 39434.025 Distance Euclidean distance between each station and city center (km) 10.830 7.529 User characteristics Te proportion of users aged 26–35 years old (Te reference variable of age is Age 26–35 0.428 0.081 16–25 years old) Age 36–45 Te proportion of users aged 36–45 years old 0.053 0.017 Over 46 Te proportion of users over 46 years old 0.019 0.008 Male Te proportion of male 0.552 0.067 Median education Te proportion of college and undergraduate 0.364 0.029 High education Te proportion of graduate students or above 0.125 0.030 Middle income Te proportion of users with monthly income of 7,000 to 15,000 yuan 0.319 0.042 High income Te proportion of users with a monthly income greater than 15,000 yuan 0.124 0.024 Finally, the GWR model is used to quantitatively analyze the and 1. When Moran’s I is positive, the variables have positive local relationship between access/egress distances and ex- spatial autocorrelation; if Moran’s I is negative, the variable planatory variables. Te following sections briefy introduce has negative spatial autocorrelation; if Moran’s I is 0, it the principle and calculation of the model. means that the variable is random to some extent. Te Z-value of Moran’s I can be calculated by the fol- lowing equation: 4.1. Spatial Autocorrelation Test. Before using the spatial I − E[I] regression model, the spatial autocorrelation of variables Z � , (2) should be tested. Moran’s I is a widely used global spatial SD[I] autocorrelation measure. Moran’s I can be expressed as where E[I] and SD[I] are the expectation and standard follows: deviation of the global Moran’s I, respectively. A positive Z n n w y − y y − y n -value indicates that the variable has more spatial aggre- i�1 j�1 ij i j , (1) I � × n n gation, while a negative Z-value indicates that the variable w y − y ij i�1 j�1 i�1 i has more spatial dispersion. Generally, the signifcance of where n is the number of subway stations, y is the average Moran’s I is estimated by pseudo P value. If the pseudo P value is less than 0.05, the global Moran’s I is statistically value of y, and w is the spatial weight between station i and ij station j. Te value of global Moran’s I is usually between −1 signifcant at the confdence level of 95%, which means that 8 Journal of Advanced Transportation the variable is spatially correlated. On the other hand, if the 1 x · · · x 11 m1 ⎡ ⎢ ⎤ ⎥ ⎢ ⎥ pseudo P value is greater than or equal to 0.05, it means that ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 1 x · · · x ⎥ ⎢ ⎥ ⎢ 12 m2 ⎥ ⎢ ⎥ ⎢ ⎥ the variable is likely to be randomly and independently ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ X � ⎢ ⎥, ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ distributed in space. ⎢⋮ ⋮ ⋱ ⋮ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ ⎦ 1 x · · · x 1n mn 4.2. Geographically Weighted Regression Model. ⎡ ⎢ ⎤ ⎥ ⎢ ⎥ ⎢ ⎥ Geographically weighted regression (GWR) model is an ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ y ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 2 ⎥ ⎢ ⎥ ⎢ ⎥ extended form of the OLS model, which is used to model ⎢ ⎥ ⎢ ⎥ Y � ⎢ ⎥, (6) ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⋮ ⎥ ⎢ ⎥ spatial variation. Compared with the general OLS and ⎢ ⎥ ⎣ ⎦ GWR allows the coefcients of explanatory variables to change in space. In order to better understand GWR, this w 0 · · · 0 i1 paper frst explains the OLS model. ⎢ ⎥ ⎡ ⎢ ⎤ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ Assuming that the space surface is uniform, the tradi- ⎢ 0 w · · · 0 ⎥ ⎢ ⎥ ⎢ i2 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ W � ⎢ ⎥, ⎢ ⎥ ⎢ ⎥ tional global OLS model is often used to explore the re- i ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⋮ ⋮ ⋱ ⋮ ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ ⎦ lationship between dependent and independent variables. 0 0 · · · w Te model formula is as follows: in where w represents the spatial weight value between the in y � β + β x + ε , (3) i 0 k ik i station i and others. In this study, the commonly used k�1 adaptive bi-square kernel is used to calculate the spatial where y is the distance of access or egress trips at the station weighting matrix, and the adaptive distance decay simulates i, β is the intercept term, β is the estimation coefcient of the spatial efect of the surrounding station in the bandwidth 0 k the k th independent variable, x is the environmental range. It is worth noting that the bandwidth selection is also ik variable, and ε is the model error at station i. important, because it will greatly afect the coefcient es- Considering the global nature of the OLS model, the timation. Tis paper chooses the bandwidth selection estimated regression coefcients are the same and constant method for the golden section search. Te corrected Akaike in the whole study area. However, since spatial data are information criterion (AIC ) is used to evaluate the ftness usually heterogeneous and highly dependent on local re- to obtain the best bandwidth. Te coefcients in the OLS model are constant, and the gional characteristics, the region cannot be completely homogeneous. As a basic extension of the OLS model, the diference between the OLS model and the GWR model is that its coefcients vary with geographical location. geographical location factor is added to the regression pa- rameters to quantify the spatial efect, and the neighborhood Terefore, we use four regression models to discuss the relationship is simulated by calibrating the model with local relationship between subway access/egress distances and coefcients. Te formula is explanatory variables. Tese four models are represented as OLS_Access, OLS_Egress, GWR_Access, and GWR_Egress. In order to further evaluate the spatial nonstationarity of the y � β u , v + β u , v x + ε , (4) i i0 i i ik i i ik i coefcients, we use AIC and adjusted R to measure the k�1 model performance of OLS and GWR. Lower AIC and where for the station i, (u , v ) represents the geographical higher adjusted R values show better model ftting. i i coordinates of the subway station, β (u , v ) is the intercept i0 i i term, β (u , v ) is the regression coefcient associated with 5.Results and Discussions ik i i the k th environmental variable, x is the k th explanatory ik 5.1. Descriptive Statistics and Analysis variable. According to the frst law of geography [32], the in- 5.1.1. Origin and Destination of Trips. Figure 3 shows the teraction between adjacent stations is more signifcant kernel density distribution of the origin and destination of than that between distant stations. Tis location uses the subway trips, respectively. Te display rule of the distri- latitude and longitude location of each subway station. bution is quantile. Te deeper the red color, the higher the Terefore, constructing a spatial weight matrix is neces- number of trip generations or destinations. Tis fgure in- sary to estimate the value of β (u , v ), which can be ik i i dicates that the trip generation or attraction is mainly calculated as follows: distributed within the loop line (Line 7), and the intensity −1 T T outside the loop line is relatively low. Te origin and des- (5) β u , v � X W X X W Y, i i i i i tination of trips mainly fall near the subway station. Tat is, the farther away from the subway station, the less the trip. In where for the station i, β (u , v ) � (β , β , · · · , β , · · · β ) i i i i0 i1 ik im addition, the fgure also shows that there are diferent in- is the estimated coefcient of the independent variable k, X, tensities around the same subway station, which may be and Y are the vector-matrix of the independent and de- related to land use. pendent variables, respectively, and W is the spatial weighting matrix, which can be expressed as follows: Journal of Advanced Transportation 9 Origin density Destination density 0 - 2,221,430 0 - 2,043,254 2,221,431 - 6,664,290 2,043,255 - 6,129,762 6,664,291 - 13,328,580 6,129,763 - 12,259,525 13,328,581 - 24,435,730 12,259,526 - 22,475,795 24,435,731 - 566,464,640 22,475,796 - 521,029,792 (a) (b) Figure 3: Spatial distribution of the origin and destination. (a) Kernal density of access trip and (b) kernal density of egress trip. 5.1.2. Analysis of the Access and Egress Distances. average distances of diferent types of users are various. Te Table 4 shows the descriptive statistics of the access and average distance between males (1.126 km) is higher than egress distances. Te distances are calculated as the Eu- that of females (0.983 km), consistent with most previous clidean distances between the origin or destination with the studies. Whether walking or bicycling, the distance between subway stations. Te table indicates the average access males is higher than that of females [26, 33]. In terms of age, distance is 1.059 km, and the 85th percentile distance is the distance of people aged 16–25 is the shortest. With the 2.031 km. However, the egress distance is lower than the increase in age, the distance is longer. Te higher the ed- ucational background, the farther the distance. It is an in- access distance, with an average distance of 0.998 km and the 85th percentile distance of 1.930 km. In order to obtain the teresting discovery, which may be related to their travel purpose. Compared with other income groups, middle- distance of the total trips, the access and egress trips of each station are added. Results show the average distance of the income people have the longest distance. Because the sub- total trips is 1.030 km, higher than the walking distance way provides an afordable travel choice for people, middle- (0.8 km) often used in practice [5, 6], and lower than the income people are more likely to choose the subway. Te bicycle distance (2 km) calculated based on GPS trajectory in egress distances are smaller, but diferent users show Shanghai [19]. Te 85th percentile distance is often used as a similar pattern compared with the access distances. the threshold for people willing to walk or bicycle to reach Te Kruskal–Wallis test is used to compare whether the subway service [7, 10, 19]. Te 85th percentile distance of there are statistical diferences in the travel distance of the total trips is 1.983 km, which indicates that the subway in diferent users. Te Kruskal–Wallis test is suitable for comparing grouping variables with two or more levels (if Chengdu provides services for most people within this distance without considering the feeder mode. there are two levels, equivalent to the Mann–Whitney U test). In addition, it is possible to judge whether the mean Figure 4 is a histogram of access and egress distances. Te left side of the vertical axis is the frequency of trips, and values of several populations are equal or not without the right side is the cumulative proportion of trips corre- making any assumptions. Te results show that the null sponding to the solid line. Te fgure shows that both the hypothesis that there is no diference between samples can access and the egress distances are in the form of decay, and be rejected (P < 0.001), which means that the distances of the farther the distances are, the fewer the trips are. It can diferent user types are signifcantly various. also be found that the commonly used 0.8 km only accounts for 59% of trips, as shown by the blue dotted line, which 5.1.4. Spatial Distribution of Access and Egress Distance. means that a large proportion of trips beyond 0.8 km are still Diferent subway stations may have diferent access and not covered, which may lead to the underestimation of the egress distance thresholds, considering the spatial hetero- service coverage of subway stations. geneity. We use the standardized circle size of the subway station to represent the threshold of the 85th percentile 5.1.3. Analysis of Diferent Users’ Access and Egress Distances. distance shown in Figure 5. It can be found that the access Table 5 shows the descriptive statistics of trips for diferent and egress distances between stations vary obviously. users, including age, gender, educational background, and Generally speaking, the stations with smaller distances are income. Both access and egress distances show that the mainly distributed in the central city, while the stations far 10 Journal of Advanced Transportation Table 4: Descriptive statistics of distance (unit: km). Type Mean SD 25th 50th 75th 85th Max Access distance 1.059 1.101 0.341 0.645 1.305 2.031 5.282 Egress distance 0.998 1.027 0.318 0.616 1.256 1.930 4.858 Distance of the total trips 1.030 1.067 0.330 0.631 1.282 1.983 5.282 found that hotel, residence, life, fnance, road density, and mixed land are negatively correlated with distance, while 85 other variables are positively correlated with distance. 5.2.2. Analysis Results of Global Moran’s I. In order to test whether the GWR model is suitable for exploring the re- lationship between subway access/egress distance and ex- planatory variables, this study frst makes a global Moran’s I through ArcGIS to check whether the selected variables have spatial autocorrelation. Table 7 shows the global Moran’s I 0 0 result. Tis test measures the spatial autocorrelation of 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 a specifc element according to its position and numerical Distance (km) value. Te null hypothesis is that there is no spatial corre- lation. According to the result, this hypothesis is rejected. P Access values of all variables are signifcant, showing a strong spatial Egress correlation. Te Z-value is greater than 0, indicating that Figure 4: Histogram of access and egress distances. each variable presents a spatial aggregation pattern. Te above evidence shows that the global OLS model cannot efectively analyze the relationship between subway travel away from the central city have a longer distance. Tis result distance and interpretation. Terefore, it is advisable to use is logical because the density of subway stations is high in the the GWR model to explore the spatial heterogeneity of data. central city, and people can reach the nearest subway station by traveling a short distance. However, in the outlying area of the central city, the density of the subway and other 5.2.3. Results Analysis of the GWR Model. In order to transportation facilities is low, and even if it is far from the compare the results of the global regression model, we use all subway station, it has to travel long distances to reach the variables in the OLS model to discuss the infuence of spatial subway station. variation of explanatory variables. GWR 4.0 software is used to model GWR. Tables 8 and 9 show the regression results of access and egress distance in GWR, respectively. Tese two 5.2. Model Results tables show descriptive statistics of regression coefcients: 5.2.1. Results of Ordinary Least Squares Regression Model. mean, standard deviation, minimum, lower quartile, upper 2 2 In this study, the OLS model is used to model the 85th quartile, maximum, and range. Te R , adjusted R , and percentile access distance to determine which factors may AICc in statistics are widely used indicators to evaluate the afect the catchment area of the subway. Te Pearson cor- applicability and performance of the model. Terefore, these relation coefcient and variance infation factor (VIF) are indexes of the GWR model in Tables 8 and 9 are compared used to eliminate the collinearity between variables. with those of the OLS model in Table 6. As the R value is First, if the correlation coefcient between the variables larger and the AICc value in GWR is smaller, the model is is greater than 0.7, it is considered that there is a high more suitable for observation data, which indicates that the correlation between the two variables, and it is deleted. Ten, GWR model is superior to the traditional OLS model in this we calculate the VIF of other variables, which shows that case study. In addition, descriptive statistical indicators there is no signifcant collinearity among variables (VIF < 5). provide an overall understanding of the distribution char- In order to select variables better, the backward stepwise acteristics of regression coefcients. For example, the resi- regression method is used, which allows a series of re- dence has a negative impact on the distance of subway travel gression models to be established by deleting and adding (average � −5.460). However, its standard deviation independent variables and evaluating which variables should (SD � 2.499) shows that the regression coefcient distribu- be kept. Te results of OLS model are shown in Table 6. Te tion of residence is more dispersed than other variables, and adjusted R of the model is 0.510 and 0.505, respectively, more than 75% of them have a negative impact on distance. indicating that the independent variables in this study ex- Tese results will help decision-makers understand the range plain at least 50.5% of the distance variation. If the P value of of local coefcients between explanatory variables and a variable is less than 0.05, the null hypothesis that there is no distances and then help to implement targeted planning relationship between variables can be rejected. It can be measures at diferent stations. Trips Cumulative trips (%) Journal of Advanced Transportation 11 Table 5: Descriptive statistics of diferent users (unit: km). Access distance Egress distance Type Percentage (%) Mean SD P value Mean SD P value Gender <0.001 <0.001 Female 39.9 0.983 1.039 0.925 0.967 Male 60.1 1.126 1.148 1.062 1.071 Age <0.001 <0.001 16–25 42.7 0.967 1.018 0.912 0.948 26–35 47.3 1.153 1.171 1.083 1.093 36–45 7.0 1.174 1.186 1.115 1.106 Over 46 3.0 1.231 1.222 1.165 1.152 Education <0.001 <0.001 Low 48.6 1.048 1.092 0.993 1.022 Median 37.2 1.060 1.102 0.997 1.027 High 14.2 1.104 1.129 1.023 1.045 Income <0.001 <0.001 Low 54.9 1.036 1.080 0.974 1.005 Middle 33.2 1.107 1.137 1.045 1.064 High 11.8 1.044 1.096 0.985 1.019 a: low education: high school and below, medium education: college and undergraduate, high education: graduate and above. b: low income: monthly income below 7,000 yuan, middle income between 7,000 and 15,000 yuan, and high income above 15,000 yuan. Access distance (km) Egress distance (km) 0.420 - 1.092 0.158 - 1.184 1.093 - 1.518 1.185 - 1.543 1.519 - 1.928 1.544 - 1.916 1.929 - 2.682 1.917 - 2.593 2.683 - 4.851 2.594 - 4.591 (a) (b) Figure 5: Spatial distribution of the 85th quantile distances. (a) Spatial distribution of access distance and (b) spatial distribution of egress distance. As the estimation coefcient of each independent var- distance is negative, but it varies from space to space. From iable varies from station to station, Figures 6–10 show that the spatial point of view, the hotel has a smaller negative subway stations are marked with diferent colors in the impact on the distance between the city center and the north. On the contrary, in areas far away from the city center, the fgure based on the value of their estimation coefcient to understand better the infuence of the spatial change of increase in hotels has a greater negative impact on the service independent variables. Because of the layout limitation, this scope of subway stations. Tese results indicate that the paper only shows and discusses that the two GWR models increase in the proportion of hotels beside suburban subway have common variables. On the whole, the infuence of most stations shortens the distance between subway stations. variables on the access and egress distances, respectively, Figure 7 shows the spatially varying efects of residence showed similar spatial variation, with only Figure 10 on subway stations’ access/egress distance. Te fgure shows showing some more signifcant diferences. that the proportion of residence is negatively correlated with Figure 6 shows the spatially varying efects of hotel on the access/egress distance. Tis may be because people subway stations’ access/egress distance. Te fgure shows usually like to live around subway stations, so the increase in that the relationship between hotel and access/egress residential proportion reduces the distance. From the spatial 12 Journal of Advanced Transportation Table 6: Results of the ordinary least squares regression model. OLS_Access OLS_Egress Variables Coefcient T-value P value VIF Coefcient T-value P value VIF (Intercept) 1.463 1.280 0.201 — 0.170 0.252 0.800 — ∗∗ ∗∗∗ Hotel −2.441 −2.678 0.008 1.190 −3.050 −3.725 <0.001 1.202 ∗∗ ∗ Residence −4.240 −3.021 0.002 1.057 −2.921 −2.269 0.024 1.113 Life — — — — −2.432 −2.191 0.029 1.306 Education 2.444 1.856 0.064 1.153 — — — — Finance — — — — −5.634 −2.152 0.032 1.292 ∗∗∗ ∗∗∗ Road density −0.020 −4.753 <0.001 1.497 −0.017 −4.113 <0.001 1.867 ∗∗ Mixed land −0.660 −2.858 0.004 1.352 — — — — ∗∗ Age 36–45 — — — — 8.553 2.704 0.007 1.572 ∗∗∗ ∗∗∗ Male 5.648 6.087 <0.001 1.585 3.528 3.999 <0.001 1.867 ∗ ∗∗∗ High education 3.719 2.197 0.029 1.144 6.154 4.001 <0.001 1.058 R 0.529 0.527 Adjusted R 0.510 0.505 AICC 439 395 ∗∗∗ ∗∗ ∗ P< 0.001, P< 0.01, P< 0.05. Table 7: Global Moran’s I result of variables. Variables Moran’s I Expected I Variance Z-value P value Access distance 0.474 −0.004 0.001 12.247 <0.001 Egress distance 0.409 −0.004 0.001 10.574 <0.001 Hotel 0.445 −0.004 0.001 13.309 <0.001 Residence 0.364 −0.004 0.001 9.735 <0.001 Life 0.381 −0.004 0.001 9.850 <0.001 Education 0.223 −0.004 0.001 5.951 <0.001 Finance 0.323 −0.004 0.001 8.406 <0.001 Road density 0.684 −0.004 0.001 17.601 <0.001 Mixed land 0.430 −0.004 0.001 11.318 <0.001 egress Age 36–45 0.277 −0.004 0.001 7.223 <0.001 access Male 0.399 −0.004 0.001 10.361 <0.001 egress Male 0.431 −0.004 0.001 11.175 <0.001 access High education 0.077 −0.004 0.001 2.161 0.030 egress High education 0.119 −0.004 0.001 3.286 0.001 Access represents the access distance. Egress represents the egress distance. Table 8: Regression results of GWR_Access. Variables Mean SD Minimum Lower quartile Upper quartile Maximum (Intercept) 2.315 2.390 −2.131 0.539 3.870 6.537 Hotel −2.379 0.845 −5.148 −2.805 −1.705 −1.442 Residence −5.460 2.499 −10.709 −7.243 −4.180 1.991 Education 3.378 1.678 0.250 2.195 4.642 6.283 Road density −0.019 0.003 −0.030 −0.020 −0.016 −0.014 Mixed land −0.695 0.414 −1.424 −0.964 −0.392 0.040 Male 4.357 1.804 1.308 2.819 5.762 7.685 High education 2.430 1.357 −0.252 1.391 3.208 6.987 R 0.609 Adjusted R 0.548 AIC 434 point of view, the impact of residential on distance is usually with the distance. Tis may be because in areas with high greater in the northeast of the city. Tese results indicate that road density, the trafc accessibility around the subway is people who live in the city’s northeast are usually more greater, thus shortening the access/egress distance of the sensitive to the access and egress distance. subway station. Te access/egress distances have shown Figure 8 shows the spatially varying efects of road similar results, and the infuence of road density on the density on subway stations’ access/egress distance. Te distance is usually greater in urban suburbs. Tis indicates fgure shows that the road density is negatively correlated that the increase in road density around suburban subway Journal of Advanced Transportation 13 Table 9: Regression results of GWR_Egress. Variables Mean SD Minimum Lower quartile Upper quartile Maximum (Intercept) 0.198 0.918 −1.370 −0.525 0.696 2.208 Hotel −3.058 1.368 −8.748 −3.528 −2.181 −1.513 Residence −4.113 1.971 −7.286 −5.436 −2.871 2.923 Life −1.143 1.477 −4.074 −2.373 0.127 1.422 Finance −7.370 1.492 −11.680 −8.286 −6.290 −2.808 Road −0.012 0.005 −0.025 −0.015 −0.008 −0.006 Age 36–45 9.091 2.835 3.486 7.485 10.742 17.309 Male 3.010 1.543 −0.483 1.998 4.184 5.934 High education 5.892 1.119 2.694 5.178 6.651 8.906 R 0.625 Adjusted R 0.547 AIC 393 Hotel Hotel -5.149 - -3.089 -8.748 - -3.664 -3.088 - -2.395 -3.663 - -3.032 -2.394 - -1.899 -3.031 - -2.485 -1.898 - -1.634 -2.484 - -2.092 -1.633 - -1.442 -2.091 - -1.514 (a) (b) Figure 6: Spatially varying efects of a hotel on the access/egress distance. (a) Efects of the hotel on the access distance and (b) efects of the hotel on the egress distance. Residence Residence -10.710 - -7.477 -7.287 - -5.658 -7.476 - -5.975 -5.657 - -4.979 -5.974 - -4.475 -4.978 - -3.986 -4.474 - -3.847 -3.985 - -2.610 -3.846 - 1.992 -2.609 - 2.923 (a) (b) Figure 7: Spatially varying efects of residence on the access/egress distance. (a) Efects of residence on the access distance and (b) efects of residence on the egress distance. 14 Journal of Advanced Transportation Road density Road density -0.030 - -0.021 -0.025 - -0.018 -0.020 - -0.018 -0.017 - -0.011 -0.017 - -0.018 -0.010 - -0.009 -0.008 -0.017 - -0.016 -0.015 -0.007 (a) (b) Figure 8: Spatially varying efects of road density on the access/egress distance. (a) Efects of road density on the access distance and (b) efects of road density on the egress distance. Male Male 1.308 - 2.274 -0.484 - 1.762 2.275 - 3.916 1.763 - 2.627 3.917 - 5.162 2.628 - 3.480 5.163 - 6.008 3.481 - 4.361 6.009 - 7.685 4.362 - 5.935 (a) (b) Figure 9: Spatially varying efects of male on the access/egress distance. (a) Efects of male on the access distance and (b) efects of male on the egress distance. stations can afect the service coverage of subway stations Figure 10 shows the spatially varying efects of high more than that in downtown. education on subway stations’ access/egress distance. Te Figure 9 shows the spatially varying efects of male on fgure shows that high education is positively correlated with subway stations’ access/egress distance. Te fgure shows distance. In the previous descriptive statistics (Section 5.1.3), that there is a positive correlation between males and dis- there are diferences in the distance between education. Te higher the education, the farther the distance is. Te distance tance. In the previous descriptive statistics (Section 5.1.3), there are diferences in the distance between the gender. at the station level also shows the same result. In the GWR However, their results do not refect whether there is model, the access distance model has a larger coefcient in a statistical diference in the distance at the station level. In the north, which means that people with higher education the GWR model, the coefcient in the north is larger, which tend to travel longer distances in these stations in the north. means that in these stations in the north, males are more In addition, it can also be seen that the coefcient of egress inclined to travel a longer distance to the subway station, distance is greater in the southern part compared to the thus expanding the coverage of the subway station. access distance, perhaps because the transportation is not so Journal of Advanced Transportation 15 High education High education -0.253 - 1.233 2.694 - 5.030 1.234 - 1.958 5.031 - 5.586 1.959 - 2.676 5.587 - 6.025 2.677 - 3.337 6.026 - 6.748 3.338 - 6.987 6.749 - 8.906 (a) (b) Figure 10: Spatially varying efects of high education on the access/egress distance. (a) Efects of high education on the access distance and (b) efects of high education on the egress distance. convenient in the southern part of Chengdu, and the highly speaking, the distance from the city center is shorter, and the educated people are more likely to transfer to other trans- distance in the suburbs is usually longer. portation modes to their destinations after leaving the Secondly, this study takes the 85th percentile distance as the key indicator of the service coverage of the station. Te subway station, which is more likely to increase the egress distance. OLS models are established to fnd out the key factors af- fecting the distance: hotel, residence, life, fnance, road- density, and mixed land are negatively correlated with the 6.Conclusion access/egress distance of the subway station. In contrast, As an essential means of transportation in big cities in China, education, 36–45 years old, male, and high education are the subway signifcantly infuences people’s travel. If the positively correlated. transportation facilities and living environment in the Finally, the GWR model is used to analyze the spatial proper area of the station are improved, more people may be variation relationship between the distance and various factors. Te goodness of ft shows that the GWR model has attracted to the subway. Terefore, this study extracts the subway access and egress trips from Chengdu’s mobile better performance than the OLS model for the same var- iable, AICc is signifcantly smaller, and the adjusted R phone positioning data to obtain the station’s service dis- is tance. Ten, the access/egress distances and the environ- higher. Te infuence of explanatory variables on distance mental variables related to the trip are calculated. In order to also varies from space to space. explore the infuencing factors of access/egress distance, this Te above results in this study are obtained based on the study applied the GWR model to test the relationship be- existing stations and the historical behavior of users in tween access/egress distances and building environment Chengdu and can provide some useful information for variables. Finally, the spatially varying efects of these ex- future public transport planning or other urban construc- planatory factors are analyzed. Te main results are as tions. In the construction of subway networks, we should follows: carefully study the characteristics of the built environment of First of all, the access and egress distance of the subway candidate stations in diferent spatial locations, and fully station varies with the diference in user characteristics and consider the behavioral desires of users to rationalize the the spatial location of the station. Comparing gender, age, planning and optimize the layout of the surroundings of education, and income shows that the average access/egress diferent subway stations. distance of males is longer than that of women. Te higher Tere are also some shortcomings in this study. First, the age and educational background, the longer the distance. although this study opens the relationships between the Compared with other income groups, middle-income access/egress distances of the subway station and the built groups have a longer distance. Tese results strengthened environment and user characteristics, travel to the subway some previous studies, emphasizing that diferent social and station is a complex behavior, depending on various factors demographic factors show signifcant diferences in the besides these factors. For example, travel purpose greatly subway access/egress distance. Te major discovery in space infuences access/egress distances, but these data cannot be is that the access/egress distance varies from station to obtained. In addition, the availability of other types of public station, showing signifcant spatial diferences. Generally transport (such as bike-sharing and ride-sharing) is also an 16 Journal of Advanced Transportation transit,” Journal of Transport Geography, vol. 82, Article ID essential factor afecting the access/egress distance, as these 102560, 2020. data are not available and we are not able to quantify what [14] J. He, R. Zhang, X. Huang, and G. Xi, “Walking access dis- efect they have on the distances, which provides a direction tance of metro passengers and relationship with demographic for future research. characteristics: a case study of nanjing metro,” Chinese Geographical Science, vol. 28, no. 4, pp. 612–623, 2018. Data Availability [15] R. Rastogi and K. V. Krishna Rao, “Travel characteristics of commuters accessing transit: case study,” Journal of Trans- Te data used to support the results of this study are available portation Engineering, vol. 129, no. 6, pp. 684–694, 2003. from the corresponding author upon request. [16] H. Pan, Q. Shen, and S. Xue, “Intermodal transfer between bicycles and rail transit in Shanghai, China,” Transportation Research Record: Journal of the Transportation Research Conflicts of Interest Board, vol. 2144, no. 1, pp. 181–188, 2010. [17] H. H. Hochmair, “Assessment of bicycle service areas around Te authors declare that they have no conficts of interest. transit stations,” International Journal of Sustainable Trans- portation, vol. 9, no. 1, pp. 15–29, 2014. References [18] A. Weinstein Agrawal, M. Schlossberg, and K. Irvin, “How far, by which route and why? A spatial analysis of pedestrian [1] Z. Gan, M. Yang, T. Feng, and H. J. P. Timmermans, “Ex- preference,” Journal of Urban Design, vol. 13, no. 1, pp. 81–98, amining the relationship between built environment and metro ridership at station-to-station level,” Transportation [19] W. Li, S. Chen, J. Dong, and J. Wu, “Exploring the spatial Research Part D: Transport and Environment, vol. 82, Article variations of transfer distances between dockless bike-sharing ID 102332, 2020. systems and metros,” Journal of Transport Geography, vol. 92, [2] S. Li, L. Dijiang, L. Xiaoping et al., “Te varying patterns of rail Article ID 103032, 2021. transit ridership and their relationships with fne-scale built [20] D. Lin, Y. Zhang, R. Zhu, and L. Meng, “Te analysis of environment factors: big data analytics from Guangzhou,” catchment areas of metro stations using trajectory data Cities, vol. 99, Article ID 102580, 2020. generated by dockless shared bikes,” Sustainable Cities and [3] J. Zhao, W. Deng, Y. Song, and Y. Zhu, “What infuences Society, vol. 49, Article ID 101598, 2019. Metro station ridership in China? Insights from Nanjing,” [21] K. Gao, Y. Yang, A. Li, and X. Qu, “Spatial heterogeneity in Cities, vol. 35, pp. 114–124, 2013. distance decay of using bike sharing: an empirical large-scale [4] X. Xu, Y. Lu, Y. Wang, J. Li, and H. Zhang, “Improving service analysis in Shanghai,” Transportation Research Part D: quality of metro systems—a case study in the beijing metro,” Transport and Environment, vol. 94, Article ID 102814, 2021. IEEE Access, vol. 8, Article ID 12573, 2020. [22] H. Yang, Y. Zhang, L. Zhong, X. Zhang, and Z. Ling, “Ex- [5] R. Daniels and C. Mulley, “Explaining walking distance to ploring spatial variation of bike sharing trip production and public transport: the dominance of public transport supply,” attraction: a study based on Chicago’s Divvy system,” Applied Journal of Transport and Land Use, vol. 6, no. 2, p. 5, 2013. Geography, vol. 115, Article ID 102130, 2020. [6] B. Canepa, “Bursting the bubble: determining the transit- [23] L. Yang, K. W. Chau, W. Y. Szeto, X. Cui, and X. Wang, oriented development’s walkable limits,” Transportation Re- “Accessibility to transit, by transit, and property prices: search Record, vol. 1992, no. 1, pp. 28–34, 1992. spatially varying relationships,” Transportation Research Part [7] A. El-Geneidy, M. Grimsrud, R. Wasf, P. Te´treault, and D: Transport and Environment, vol. 85, Article ID 102387, J. Surprenant-Legault, “New evidence on walking distances to transit stops: identifying redundancies and gaps using variable [24] J. Zhao and W. Deng, “Relationship of walk access distance to service areas,” Transportation, vol. 41, no. 1, pp. 193–210, 2013. rapid rail transit stations with personal characteristics and [8] Z.-j. Wang, F. Chen, and T.-k. Xu, “Interchange between station context,” Journal of Urban Planning and Development, metro and other modes: access distance and catchment area,” vol. 139, no. 4, pp. 311–321, 2013. Journal of Urban Planning and Development, vol. 142, no. 4, [25] B. W. Alshalalfah and A. S. Shalaby, “Case study: relationship of walk access distance to transit with service, travel, and [9] Y. L. Xi, S. Saxe, and E. Miller, “Accessing the subway in personal characteristics,” Journal of Urban Planning and Toronto, Canada: access mode and catchment areas,” Development, vol. 133, no. 2, pp. 114–118, 2007. Transportation Research Record Journal of the Transportation [26] X. Ma, Y. Jin, and M. He, “Measuring bikeshare access/egress Research Board, vol. 2543, no. 1, pp. 52–61, 2016. transferring distance and catchment area around metro sta- [10] T. Zuo, H. Wei, and A. Rohne, “Determining transit service tions from smartcard data,” Information, vol. 9, pp. 289–311, coverage by non-motorized accessibility to transit: case study of applying GPS data in Cincinnati metropolitan area,” [27] L. Yang, X. Chu, Z. Gou, H. Yang, Y. Lu, and W. Huang, Journal of Transport Geography, vol. 67, pp. 1–11, 2018. “Accessibility and proximity efects of bus rapid transit on [11] J. Lee, K. Choi, and Y. Leem, “Bicycle-basedtransit-oriented development as an alternative to overcome the criticisms of housing prices: heterogeneity across price quantiles and space,” Journal of Transport Geography, vol. 88, Article ID the conventional transit-oriented development,” In- ternational Journal of Sustainable Transportation, vol. 10, 102850, 2020. [28] R. Pan, H. Yang, K. Xie, and Y. Wen, “Exploring the equity of no. 10, pp. 975–984, 2015. [12] J. Wang and X. Cao, “Exploring built environment correlates traditional and ride-hailing taxi services during peak hours,” Transportation Research Record, vol. 2674, no. 9, pp. 266–278, of walking distance of transit egress in the Twin Cities,” Journal of Transport Geography, vol. 64, pp. 132–138, 2017. 2020. [29] C. Wang, X. Wang, R. Pan, and Y. Yan, “Infuence of built [13] T. Tao, J. Wang, and X. Cao, “Exploring the non-linear as- sociations between spatial attributes and walking distance to environment on subway trip origin and destination: insights Journal of Advanced Transportation 17 based on mobile positioning data,” Transportation Research Record, vol. 2676, no. 9, pp. 693–710, 2022. [30] L. Ni, X. C. Wang, and X. M. Chen, “A spatial econometric model for travel fow analysis and real-world applications with massive mobile phone data,” Transportation Research Part C: Emerging Technologies, vol. 86, pp. 510–526, 2018. [31] X. Yang, Z. Fang, Y. Xu, L. Yin, J. Li, and S. Lu, “Spatial heterogeneity in spatial interaction of human move- ments—insights from large-scale mobile positioning data,” Journal of Transport Geography, vol. 78, pp. 29–40, 2019. [32] W. R. Tobler, “A computer movie simulating urban growth in the detroit region,” Economic Geography, vol. 46, pp. 234–240, 1970/06/01 1970. [33] J. C. Garc´ıa-Palomares, J. Gutierrez, ´ and O. D. Cardozo, “Walking accessibility to public transport: an analysis based on microdata and GIS,” Environment and Planning B: Planning and Design, vol. 40, no. 6, pp. 1087–1102, 2013.

Journal of Advanced Transportation – Hindawi Publishing Corporation

**Published: ** Feb 28, 2023

Loading...

You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!

Read and print from thousands of top scholarly journals.

System error. Please try again!

Already have an account? Log in

Bookmark this article. You can see your Bookmarks on your DeepDyve Library.

To save an article, **log in** first, or **sign up** for a DeepDyve account if you don’t already have one.

Copy and paste the desired citation format or use the link below to download a file formatted for EndNote

Access the full text.

Sign up today, get DeepDyve free for 14 days.

All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.