Spatio-temporal Variation of Urban Bus Ridership Using Smart Card Data in a Compact City

Abstract

Buses are one of the most essential parts of the urban transportation system since they can cater to residents’ daily transit demands. This study investigates the spatio-temporal patterns of urban bus traveler activities using a one-month bus trajectory dataset in Kumamoto City based on smart card data. Based on the idea of a compact city, the ordinary least squares (OLS) regression model is implemented to explore the factors that affect the bus ridership at the bus stop level in the 17 essential core districts in Kumamoto City. Then using the geographically weighted regression (GWR) model, we can get the spatial heterogeneity of the bus ridership and visualize the spatial distributions of parameter estimations. By comparing the results of the two models, we find that the two models performed similarly both in global fit and explanatory accuracy. These results can provide valuable suggestions for estimating bus demand, which may exert important implications for bus route optimization. They can also provide a basis for policy formulation by city and transportation planning and management authorities. The results of the study demonstrate the effectiveness of compact urban development in Kumamoto City from the perspective of bus ridership in each core district.

introduction

Public transportation is an important means of transportation in the daily lives of citizens and is indispensable for the formation of affluent local communities, such as the development of local economies. In recent years, several countries have begun to focus on upgrading their public transport infrastructure to make it more attractive to people. It is critical to understand what factors may affect transit ridership and to develop reasonable ways to improve the attractiveness of public transportation and quantify the impact of these factors. Internal factors are those that are under the control of the transit operator, such as quality of public service, operating management, and transit fares (Taylor, Miller, et al., 2009). External factors are those beyond the control of the transit operator, including urban development policies, socioeconomic conditions, individual passenger attributes, and built environment factors (Li, Wang, et al., 2015). With the progress and spread of ICT (Information and Communication Technology), a wide variety of big data has come to be generated and accumulated. In recent years, scholars began to have interests in how to extract relevant and useful information from automatically collected data (ETC data, GPS data, smart card data, et al) (Tang, Liu, et al., 2020). Smart card data has been widely used to evaluate urban bus ridership (Bagchi and White, 2005). Hosoe, Kuwano, et al. (2018) examined a methodology for extracting user movement patterns from a large amount of data accumulated daily for transportation IC cards. There are great expectations for new findings obtained by quantitatively analyzing big data, and they are being used as a means for evaluating and examining transportation infrastructure plans and maintenance measures (X. Zhang, Q. Zhang, et al., 2018). In this context, it’s necessary to find out the main factors that affect bus travel behaviors and ridership to improve the urban and transport planning.

Smart card data is helpful for smart cities (Lim, Kim, et al., 2018; Y. Zhao, Zhang, et al., 2018) and can provide opportunities for innovation in transportation research, such as exploring the impact of the built environment on travel behaviors (P. Zhao and Hu, 2019). Although there are many existing studies that use smart cards to analyze the behavioral characteristics of people using public transportation, there are few studies that analyze public transit use within small living communities based on a compact city perspective, and this study can effectively remedy this gap. In Kumamoto City, in the city master plan, even though the population is expected to decline, and the population is aging, to maintain the vitality of the city over the long term, the aim is to create a compact and sustainable city where everyone can move and live comfortably. "Multi-core cooperation city" is listed as the future image of the urban structure. The basic policy of urban development in Kumamoto City is this multi-core cooperation type of urban development. When public transportation stations are arranged around a community, high density and mixed land uses can enhance the strength of the community. In this study, we will grasp the current state of behavioral characteristics of bus users in Kumamoto City and explore the effects of different built environment factors on bus rides. After that, we analyze how built environment factors affect the spatial differentiation of bus usage and aim to explore the effectiveness of Kumamoto City's compact city policy based on multi-core urban development.

Literature Review

Built environment measurements

Much research examines transportation ridership in an attempt to connect ridership with the built environment across different contexts. The importance of the built environment is widely recognized to support the transportation. The built environment can play an important part as a barrier to transit use by identifying the departures and destinations around each station, defining the physical context of access to the transit system, and linking stations to the urban fabric (Vergel-Tovar and Rodriguez, 2018). The built environment is known to affect travel demand in three main aspects —density, diversity, and design (Cervero and Kockelman, 1997). There is a need to treat density more cautiously and it should be discussed when other main factors have been controlled under the background of high-density cities (Pan, Zheng, et al., 2020), as density is considered to be able to impact the spatial arrangement in urban areas (Soltani, Gu, et al., 2020). In later research, Cervero (2007) discovered through deeper research that accessibility must also be a factor in building the environment, and as bus accessibility and destination accessibility increase, bus utilization rates. Density can increase the accessibility to the destinations and the amount of public transit ridership. Earlier research has focused on discussing the different factors that affect transit ridership at the macro level (region or nation) (Chakour and Eluru, 2016). There are also many studies showing that transit ridership is related to the built environment factors at the station level (Jun, Choi, et al., 2015; Zhao, Deng, et al., 2013). People's choice of public transportation is a personal decision-making behavior based on their economic status, lifestyle, preferences, and values. Population density and urban vitality are closely related (Blainey, 2010). Distance to destination is directly related to the cost of residents' transportation use and is one of the crucial variables in constructing the built environment variables, and can influence residents' transportation choices in many ways, including the length of time spent using transportation and the distance over which they use transportation. According to a previous study by scholars Van Acker and Witlox (2010), as the distance from the CBD increases, the percentage of households owning passenger cars also increases. Furthermore, the percentage of passenger car use is increasing, while the percentage of bus use is decreasing. Regarding land use variables, land use types are usually categorized into residential, commercial, and office land uses to calculate land use/building area and land use mix (Sohn and Shim, 2010; Sung, Choi, et al., 2014; J. Zhao, Deng, et al., 2013). In addition, scholars have started to use POI data from open-source maps to calculate the number of different public facilities and relate them to transit use (Shi, Zhang, et al., 2018; J. Zhao, Deng, et al., 2014). The road characteristics are also a very important part of the built environment, and the road length in a region is usually taken into account in transportation-related studies (Liu, Homma, et al., 2020).

Concept of compact city

From the 1960s, the concept of the compact city began to appear in urban policy (Jacobs, 1965). Nowadays, the notion of a compact city is still an essential response to multiple societal challenges in cities. The compact city is put forward for the urban spread of disorderly development to achieve sustainable development. The process of urban development attracts a large influx of people and the construction of large-scale buildings, and the mixed-use and intensive development of land is an important expression of the compact city. It is kind of a new concept based on the efficient use of land use recourses and city development. The concept of the compact city is primarily realized through spatial design, which emphasizes urban form as a decisive factor in shaping a sustainable society and insists on specific growth boundaries to curb sprawl (Westerink, Haase, et al., 2013). However, population decline has become a social problem in many developed countries, and policymakers need to propose strategies to improve urban development based on this issue (Oswalt and Rieniets, 2006), the focus on the discussion on how to plan a more sustainable urban form has shifted from ideal to reality. In many European cities, the compact city form has trickled down into the urban policy. For example, the City of Barcelona pursued its Mediterranean version of the compact city. And the City of Rotterdam tried to densify the housing stock in the urban area to provide a more attractive living environment to maintain a balance between life and jobs (Kain, Adelfio, et al., 2022). The UK government promoted land use and transport integration planning to reduce transport energy consumption (Breheny, 1995). Even in the United States, scholars argued against urban sprawl and advocated growth-management policies (Chinitz, 1990). In addition, in Japan, the government responded to urban population decline by the compact city policy, which is to induce people to relocate where there is high accessibility to the downtown areas and public transportation networks (Sakamoto, Iida, et al., 2018; Tsuboi, Ikaruga, et al., 2015). Achieving compactness of urban form at the community level is an important part of a sustainable development strategy in cities with high-density development. Mixed land use plays an important role in encouraging public transportation because it reduces travel distances and car use.

Materials and Methods

Study area and data

The study area (Figure 1.) covers Kumamoto City. Kumamoto City is in the northern part of Kumamoto Prefecture. It is a prefectural government-designated city in Kumamoto Prefecture with five districts including Nishi ward, Kita ward, Chuo ward, Higashi ward, and Minami ward. In this study, we examined the 238 bus stops in 17 core districts in Kumamoto City. The range of the core districts is basically within the urbanized area, and the radius is approximately 800 meters from the center point of the regional base considering the topography, features, and urban development, excluding the industrial area (generally within a 10-minute walk area where walking and biking are the main means of transportation). These 17 core districts are representative and concentrated population areas.

Figure 1. The study area

The data on bus trips were collected by Kumamoto City Transportation Bureau (KCTB). Daily bus trips varied between about 7,900 and 21,000 based on the observation from June 1, 2019, to June 30, 2019. Since there are some problems among the collected IC card data such as defects and errors, it is necessary to pre-process the collected initial data before performing data analysis to improve the quality of the data and ensure the accuracy of the research. In addition, since we focused only on bus trips of getting on and off within the Kumamoto City area, we analyzed the latitude and longitude after the link and remove the bus trip data outside Kumamoto City. Finally, in this study, we organized a total of 483419 valid IC card data for June 2019.

Methodology

The maximum service radius of a bus stop is 500 meters, while the urban land use and the distribution of facilities within 500 meters around the stop greatly affect the choice of travel mode of residents. Therefore, a distance threshold of 500 m was considered for the bus stops in this study area. Explanatory variables that represent the built environment can be broadly divided into three categories: area conditions, city design, and land use. Concerning previous studies (Morrall and Bolge, 1996; Sung and Oh, 2011), this study used open-source data as much as possible and incorporated built environment indicators that have a significant impact on bus rides (Table 1). In order to obtain more specific facility types, we used Python to obtain some facility POIs in Kumamoto City and classify these POIs.

Table 1. Candidate list of explanatory variables

Category	Name	Description
Area condition	Population density	Total population within the 500m catchment area.
	Distance from CBD	The length of the distance from CBD to each bus stop. (km)
City design	Road density	Road density within the 500m catchment area.
Land use	Building density	The number of buildings within the 500m catchment area.
	Park density	The number of parks within the 500m catchment area.
	Hospital density	The number of hospitals within the 500m catchment area.
	School density	The number of schools within the 500m catchment area.
	Shopping mall density	The number of shopping malls within the 500m catchment area.
	Restaurant density	The number of restaurants within the 500m catchment area.
	Public facility density	The number of public facilities within the 500m catchment area.
	Recreational facility density	The number of recreational facilities within the 500m catchment area.

The spatial patterns of both the dependent and explanatory variables are spatially non-stationary due to the inherent function of urban form (Qian and Ukkusuri, 2015). Moran's I is the global spatial autocorrelation coefficient, and the range of values obtained is between -1 and 1. At some prominent level, if Moran's I is significantly positive, there is a significant positive correlation between the observations, if there is a spatially high or low aggregation pattern, and if Moran's I is significantly negative, there is a significant positive correlation between the observations. It shows a remarkable negative correlation, high and low differences, and spatial dispersion. the "Spatial Autocorrelation (Global Moran's I)" is used to examine whether each built environment factor as a self-variable is an agglomeration distribution or a dispersion distribution and examine the degree of aggregation or dispersion.

First, the spatial distribution patterns of the 11 variables in Table 1 were analyzed using the Global Moran's I index to determine if any spatial autocorrelation occurred. In addition, a bivariate correlation matrix was generated between the candidate impact factors to prevent multicollinearity in the model. Multicollinearity can cause regression models to perform badly.

Table 2. Moran’s I test for spatial autocorrelation of explanatory variables

Variable	Moran's I	Expected index	Pattern	Z Score	p
Population density	0.463514	-0.004219	Clustered	45.503277	0.0000*
Distance from CBD	0.781170	-0.004219	Clustered	91.569784	0.0000*
Road density	0.218073	-0.004219	Clustered	25.944091	0.0000*
Building density	0.189864	-0.004219	Clustered	22.673096	0.0000*
Park density	0.202731	-0.004219	Clustered	24.157566	0.0000*
Hospital density	0.220879	-0.004219	Clustered	26.421178	0.0000*
School density	0.150580	-0.004219	Clustered	18.252285	0.0000*
Shopping mall density	0.079123	-0.004219	Clustered	9.928177	0.0000*
Restaurant density	0.041712	-0.004219	Clustered	5.456347	0.0000*
Public facility density	0.315094	-0.004219	Clustered	37.448913	0.0000*
Recreational facility density	0.144295	-0.004219	Clustered	17.555214	0.0000*

Note:*Indicates that it correlates significantly with the level of 0.1 (p <0.1).

Regression Analysis is a simple and intuitive method of statistically analyzing data to find out if there is a correlation between two or more variables, the direction, and intensity of association, and to observe a particular variable (Deng, Fannon, et al., 2018). It has been widely used in practical applications (Yan and Su, 2009). In this study, the OLS (ordinary least squares) model is used to analyze the effects of built environment factors on the bus ridership in Kumamoto City. The dependent variable is the daily bus rides in each core district at the bus stop level and the explanatory variables include 3 domains to represent the built environment. The primary purpose of modeling in this way is to explore the influence of built environment factors on travelers' decision-making. The structure of this model can be shown as follows.

$y_i={\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+\cdots {\beta }_p{x}_p+{\epsilon }_i$ (1)

In Formula (1), y_iis the dependent variable, x_p is the explanatory variable, p is the number of variables, and ε_i is the random error item. The OLS model has been widely applied in real life, intuitively reflecting the changes in variables that accompany several arguments, and quantitatively describing the relevance of individual arguments and factor variables.

In the traditional OLS model, the relationships between all dependent and explanatory variables are assumed global. And for spatial data, spatial heterogeneity will exist due to different geographical locations. Geographically weighted regression (GWR) is an extension of the usual linear regression model, incorporating the spatial position of the data into the model parameters. It is a new method for modeling spatial heterogeneity. Spatial heterogeneity shows that in each point/site (geographical coordinates) there is a different correlation between the dependent and explanatory variables due to the dependence of the parameters or coefficients of the model on the point/site (Fotheringham, Charlton, et al., 2001; Szymanowski and Kryza, 2012). The GWR model has been applied to several fields like urban science and transportation mobility for the past few years. Ji, Ma, et al. (2018) established the GWR model to study the spatial change in the relationship between the number of subway-public bicycle refutations and urban construction environment variables. Currently, there are few studies on the relationship between bus ridership and built environment factors using geospatial weighted regression analysis methods, and this measure makes new attempts to fill the gap in this area. The structure of this model can be expressed as follows.

$y_i={\beta }_0(u_i,v_i)+{\beta }_1(u_i,v_i)x_{1,i}+{\beta }_2(u_i,v_i)x_{2,i}+\cdots +{\epsilon }_i={\sum }_k^{}{\beta }_k(u_i,v_i)x_{k,i}+{\epsilon }_i$ (2)

Where (u_i, v_i ) are the coordinates of sampling point i ; β_k (u_i, v_i ) is the kth regression parameter on sampling point i, and it is a function of geographic location.

The bus ridership model for each bus stop was performed using the OLS and GWR models. We obtained the results of the model through the ArcGIS spatial model results. When the OLS and GWR models were obtained, they would be compared for their global fit and the distributions of the residuals.

Results and Discussion

Exploratory analysis

According to the time getting on/off the bus of passengers’ smart card, the data can be sorted by day of the week for a valid bus smart card with a bus code, and the weekly status of the bus rides within one month of Kumamoto bus passengers can be obtained.

Figure 2. Number of bus rides per day in a week

On weekdays from Monday to Friday, the number of Kumamoto citizens getting on/off the bus is relatively stable. It can be seen that an average of about 20,000 people get on the bus every day on weekdays. The number of people taking a ride on the bus on weekdays does not change too much, but the number of bus rides on weekends decreases, and the average number of bus rides on two days off is about half that on weekdays. Sunday is not the busiest day of the week. The number of bus rides in Kumamoto City is high on weekdays and has decreased significantly on weekends (Figure 2).

Figure 3. Bus rides on weekdays and weekends

Figure 3 shows the distribution of time of bus rides on weekdays and weekends for Kumamoto citizens. Bus rides in Kumamoto City on weekdays are concentrated from 7:00 to 9:00 in the morning and from 16:00 to 19:00, and it can be said that this time zone is rush hour, which the current distribution of commuting time in Japan is the same. The number of people taking a ride on the bus from 10:00 to 15 hours on weekdays is stable at around 650 people. The number of people getting on/off the bus before 6 am and after 23:00 is small, less than 100 people.

The number of bus rides in Kumamoto City on weekends has decreased significantly, and the fluctuation range is small. The number of people getting on the bus on weekends is less than 4000 at any time of the day.

In the core districts, the largest number of people get on and off at bus stops is in the central city area, with about 752 people every day. There are many bus stops in the central city area, the population density and the bus utilization rate are high. The central city area is more closely connected to the bus routes in other areas, and the bus stops in this area are connected to the bus stops in the other 15 areas. In addition, the largest number of passengers go from the city center to the Suizenji-Kuhonji Area by bus, with about 540 people every day. The area where the number of bus rides is low is Hakenomiya-Shimizukamei Area, and the number of passengers going from this area to other areas is less than 10 people every day. In addition, the bus routes in the Shiroyama Area are only in the center of the city, and the number of daily bus passengers is less than 30.

Figure 4 shows a statistical analysis of the average daily bus rides in each region (including the number of bus riders getting on and off at bus stops in the entire city range other than the other base areas and the central city area). It can be found that the number of bus riders getting on and off the bus each day in each base area and the central city area is about the same. Among them, the number of people taking the bus is the largest in the central city area, about 6000. Hakenomiya-Shimizukamei Area and Tomiai A rea bus passengers are less than 10 people in total.

Figure 4. Number of bus rides by area

Results of models

Before building the models, Global Moran’s I analysis was conducted to determine where the explanatory variables were spatially autocorrelated. All explanatory variables’ estimated Moran’s I values were higher than the expected E(I) (Table 2), proving that the positive spatial autocorrelation existed.

Through correlation analysis, variables that have a significant correlation with bus rides in each core district are selected as explanatory variables (Table 3), and the number of daily bus riders in each region at the bus stop level is subjected to regression analysis using the dependent variable. In urbanized areas, where land use is diversified, people may be encouraged to choose diverse modes of travel. The correlation coefficients of the variables represent the degree of influence of the variables on bus ridership.

The variance inflation factor (VIF) is a measure of the severity of covariance in a multiple linear regression model. It can express the ratio of the variance of the estimated regression coefficients compared to the variance when no linear correlation between explanatory variables is assumed. Usually, when VIF is less than 10, there is no multicollinearity among the variables (Mason, Gunst, et al., 1989). After excluding multicollinearity, we selected variables with correlation with transit ridership for regression analysis. The estimated coefficients of variables and other detailed information about the OLS model are shown in Table 4. The results of the model with OLS showed that only 1 independent predictor variable had a significant effect on bus ridership. Commercial land use is associated with bus travelers (Chan and Miranda-Moreno, 2013). The density of restaurant distribution in each core area of the city has a positive impact on bus ridership. Specifically, every time a restaurant is added to the 500-meter radius buffer area around each bus stop in core districts, the bus ridership would increase by 8.52 passengers. The R square of the OLS model is 0.2430 and the adjusted R square is 0.2233, which means this model could explain over 24% of the variability in the bus ridership data.

Table 3. Result of the correlation analysis

Explanatory variables	Cor. Coe.	Sig.
Population density	.064	.326
Distance from CBD	-.183**	.002
Road density	.161**	.007
Building density	.150*	.010
Park density	.042	.518
Hospital density	.389**	.000
School density	.009	.886
Shopping mall density	.403**	.000
Restaurant density	.481**	.000
Public facility density	.417**	.000
Recreational facility density	.445**	.000

Table 4. Result of the OLS model

Variable	Coefficient	Std-error	t-statistic	P (Sig.)	VIF
Intercept	1372.254514	2166.900231	0.633280	0.527178	—
Distance from CBD	906.544129	13571.257947	0.066799	0.946788	1.579256
Road density	-0.126622	0.185778	-0.681580	0.496184	2.002916
Building density	-0.006193	0.010132	-0.611214	0.541661	2.066078
Shopping mall density	-5.427964	3.784038	-1.434437	0.152811	5.991053
Restaurant density	8.519413	2.308203	3.690929	0.000289*	4.782493
Public facility density	4.124235	2.534322	1.627352	0.105039	9.403672
Number of observations	238
Number of variables	6
Adjusted R2	0.2233
R2	0.2430
AICc	4720.118172

Unlike the OLS model, the GWR model provided varying coefficients for each parameter. As a local model, the regression coefficients of each sample spatial element of the GWR model changed by region. It can show the spatial unsteadiness of the model interpreter variables in a better way. A GWR model with the same explanatory variables as the OLS model was built assuming the influence that each variable has on ridership of each bus stop varies spatially. The statistics result of coefficients estimation in the GWR model is shown in Table 5. Due to the uneven development of urban space, the built environmental factors in the 500-meter catchment area of each bus stop have certain differences in the process of development, and due to such spatial inhomogeneity and difference, these variables are different in each region. The magnitude of the effect is different, and the regression coefficient is also different. Regression coefficient parameters allow a deep analysis of the effects of each variable on the causes of spatial differences between the variables. In this study, changes in the regression coefficient spatial distribution of each variable are spatially visualized in ArcGIS, and the effects of built environmental factors in each 500-meter catchment area of the bus stop on the spatial differentiation of buses and their possible causes are discussed and analyzed.

Table 5. Result of the GWR model

Variables	Min.	Max.	Mean	Std Desv.
Intercept	1371.772902	1372.802772	1372.332212	0.260752
Distance from CBD	906.645883	912.34881	909.293266	1.4484
Road density	-0.126682	-0.126587	-0.126635	0.000022
Building density	-0.006196	-0.006193	-0.006195	0.000001
Shopping mall density	-5.429587	-5.427345	-5.428384	0.000569
Restaurant density	8.518796	8.519632	8.519188	0.000213
Public facility density	4.124639	4.125281	4.124919	0.000121
Number of observations	238
Number of variables	6
Adjusted R2	0.2234
R2	0.2429
AICc	4720.128912
Bandwidth	2.032363

In Figure 5, we can see in the western and northern core districts, the distance from CBD has the strongest effect on the bus ridership. From western districts to the eastern side, this effect becomes weak gradually. The road density variables have negative coefficients in all core districts. Chakour and Eluru (2016) suggested that the length of highways exerts a negative impact on bus ridership. It can be observed at in areas with high road density, residents in this area are likely to travel by car and fewer people will board the bus. The regression coefficients of building density are very small, and it has little impact on bus ridership. According to Table 5, the distribution of shopping malls has a negative impact on bus ridership. It can be found that the negative impact is greater in the eastern core districts. This is probably because each core district has a concentration of urban functions, the distribution of shopping malls in the core districts can meet the living needs of the surrounding residents within walking distance, and the residents will not need to travel by bus. In the northern core districts, the city center area, and the southwestern core districts, the distribution density of restaurants has a greater positive impact on transit ridership. In addition, public facility density has a stronger effect on the ridership of each bus stop in northern core districts. Commercial facilities like restaurants and public service facilities will attract residents from other districts to take public transportation.

Figure 5. Coefficients of variables in the GWR model

Comparison can be made between the two regression models. We can analyze the degree of influence of various explanatory variables on bus rides from the perspective of the global and local and explain the distribution of built environment factors more comprehensively. To quantify the simulation effect of the comparative OLS model and GWR model, Adjusted R2 and AICc are examined as evaluation indexes of model reliability. As can be seen from Tables 4 and Table 5, the values of adjusted R square and AICc are similar for both models, indicating that both models performed similarly for the simulation of these variables. Due to the unique pattern of urban development in Kumamoto City, urban activity is low in many areas and daily ridership at some bus stops is in the single digits. In fact, within each core district, the central bus stop has a high number of passengers, and the other neighboring stations have few passengers. The big size of sampling bus stop points leads to the lower global fit of the OLS model and the GWR model.

Conclusion

Buses are an important means of transportation in the daily lives of citizens. Few studies have investigated the specific geographical effects of facility-scale built environment indicators on the daily bus ridership. In this study, we examined the current state of bus rides, built environmental factors in urban construction, and the relationship between the built environment and bus usage. Direct stop-level bus ridership model estimation models have advantages over the traditional models. The innovation of this study is to use the region as the base point and select the bus stop as the object of the evaluation model, allowing an in-depth analysis of the environment around bus stops, which is used to assess the impact of changes in the built environment on bus ridership.

Kumamoto City has selected the more active traffic points as the center and delineated an 800-meter area as an urban function core district. Kumamoto City is a city with severe aging as well as a large population loss. The establishment of a regional core living circle is intended to ensure the convenience of daily life for residents and the livability of the area in the face of a declining urban population. Taking Kumamoto City as an example, the behavioral characteristics of residents were clarified in detail through the collection and analysis of Kumamoto City bus smart card data. By analyzing the data on the smart card, it was clarified that the bus ridership in Kumamoto City changes over time. In addition, we focused on the quantitative relationship between the characteristics of bus rides at 17 regional core districts and the built environmental factors, then clarified the factors that affect bus rides in each region. We analyzed the degree of influence of the built environmental factors on bus rides from the viewpoints of the whole and parts of 17 regional core districts.

The result of the OLS analysis model revealed that the density of restaurants around bus stops has a significant effect on bus rides per day. The GWR regression model showed the result from the spatial perspective, revealing that distance from CBD, road density, building density, shopping mall density, restaurant density, and public facility density variables exert significant spatial differences in bus rides by stop. Then, by comparing the two models, it can be found that the two models perform similarly. Through global and local regression models, we explored the impact of built environment factors on bus ridership within each core district, using bus stops as the study target. The model results verify that the facilities within the current core districts have largely met the needs of residents. Spatially, the distance from the CBD would be the strongest factor influencing people to take public transportation. This study proved the feasibility of Kumamoto City's compact city strategy of developing a multi-core district as a base. This study started from the viewpoint of compact city planning, sought the relationship between public transportation and the built environment in each core district, and enriched the quantitative and scientific research methods of city planning. And the methodology used in this study can be extended to some other related fields to explore spatial relationships between variables.

It’s certain that this study is not without limitations. We recognize that capturing the impact of urban design is a subtle process that cannot be separated from the critical activities of people. However, in this study, we only discussed the temporal and spatial effects of the built environment on transit use but did not combine it with cabs or cycling. In the future study, we will consider discussing bus ridership in combination with cabs or cycling.

Author Contributions

Conceptualization, Q. F.; methodology, Q. F., Q. L., and Q. Z.; software, Q. F.; investigation, Q. F., and Q. L.; resources, R. H.; data curation, Q. F.; writing—original draft preparation, Q. F.; writing—review and editing, Q. F., R. H., T. I., Q. L, and Q. Z.; supervision, R. H. All authors have read and agreed to the published version of the manuscript.

Ethics Declaration

The authors declare that they have no conflicts of interest regarding the publication of the paper.

Acknowledgments

The authors thank Kumamoto City Transportation Bureau for providing the data and information used in this paper. This work was supported by JST SPRING, Grant Number JPMJSP2136.

References

•  Bagchi,  M., and  White , P. R. (2005). "The potential of public transport smart card data". Transport Policy, 12(5), 464–474. https://doi.org/https://doi.org/10.1016/j.tranpol.2005.06.008
•  Blainey,  S. (2010). "Trip end models of local rail demand in England and Wales". Journal of Transport Geography, 18(1), 153–165. https://doi.org/https://doi.org/10.1016/j.jtrangeo.2008.11.002
•  Breheny,  M. (1995). "The Compact City and Transport Energy Consumption". Transactions of the Institute of British Geographers, 20(1), 81–101. https://doi.org/https://doi.org/10.2307/622726
•  Cervero,  R. (2007). "Transit-Oriented Development’s Ridership Bonus: A Product of Self-Selection and Public Policies". Environment and Planning A: Economy and Space, 39(9), 2068–2085. https://doi.org/https://doi.org/10.1068/a38377
•  Cervero,  R., and  Kockelman,  K. (1997). Travel demand and the 3Ds: Density, diversity, and design. Transportation Research Part D: Transport and Environment, 2(3), 199–219. https://doi.org/https://doi.org/10.1016/S1361-9209(97)00009-6
•  Chakour,  V., and  Eluru,  N. (2016). "Examining the influence of stop level infrastructure and built environment on bus ridership in Montreal". Journal of Transport Geography, 51, 205–217. https://doi.org/https://doi.org/10.1016/j.jtrangeo.2016.01.007
•  Chan,  S., and  Miranda-Moreno,  L. (2013). "A station-level ridership model for the metro network in Montreal, Quebec". Canadian Journal of Civil Engineering, 40(3), 254–262. https://doi.org/https://doi.org/10.1139/cjce-2011-0432
•  Chinitz,  B. (1990). Growth Management: Good for the Town, Bad for the Nation? Journal of the American Planning Association, 56(1), 3–8. https://doi.org/https://doi.org/10.1080/01944369008975739
•  Deng,  H.,  Fannon,  D., et al. (2018). Predictive modeling for US commercial building energy use: A comparison of existing statistical and machine learning algorithms using CBECS microdata. Energy and Buildings, 163, 34–43. https://doi.org/https://doi.org/10.1016/j.enbuild.2017.12.031
•  Fotheringham , A. S.,  Charlton , M. E., et al. (2001). Spatial Variations in School Performance: A Local Analysis Using Geographically Weighted Regression. Geographical and Environmental Modelling, 5(1), 43–66. https://doi.org/https://doi.org/10.1080/13615930120032617
•  Hosoe,  M.,  Kuwano,  M., et al.. (2018). "Movement Pattern Extraction from Smart Card Data using Non-negative Tensor Factorization". Journal of the City Planning Institute of Japan, 53(3), 1320–1326.
•  Jacobs,  J. (1965). The Death and Life of Great American Cities (Second Edition). Penguin Books.
•  Ji, Y.,  Ma,  X., et al. (2018). "Exploring Spatially Varying Influences on Metro-Bikeshare Transfer: A Geographically Weighted Poisson Regression Approach". Sustainability, 10(5), 1526. https://doi.org/https://doi.org/10.3390/su10051526
•  Jun,  M.-J.,  Choi,  K., et al. (2015). Land use characteristics of subway catchment areas and their influence on subway ridership in Seoul. Journal of Transport Geography, 48, 30–40. https://doi.org/https://doi.org/10.1016/j.jtrangeo.2015.08.002
•  Kain,  J.-H.,  Adelfio,  M., et al. (2022). Towards a systemic understanding of compact city qualities. Journal of Urban Design, 27(1), 130–147. https://doi.org/https://doi.org/10.1080/13574809.2021.1941825
•  Li, L.,  Wang,  J., et al. (2015). Analysing the impact of weather on bus ridership using smart card data. IET Intelligent Transport Systems, 9(2), 221–229. https://doi.org/https://doi.org/10.1049/iet-its.2014.0062
•  Lim,  C.,  Kim,  K.-J., et al. (2018). Smart cities with big data: Reference models, challenges, and considerations. Cities, 82, 86–99. https://doi.org/https://doi.org/10.1016/j.cities.2018.04.011
•  Liu,  Q.,  Homma,  R., et al. (2020). "Quantitative Evaluation on Public Bicycle Trips and its Impact Variables among Different Land Uses". International Review for Spatial Planning and Sustainable Development, 8(2), 118–130. https://doi.org/https://doi.org/10.14246/irspsda.8.2_118
• Mason, R. L., Gunst, R. F., et al. (1989). Statistical design and analysis of experiments: Applications to engineering and science. Wiley.
•  Morrall,  J., and  Bolge,  D. (1996). "The relationship between downtown parking supply and transit use". ITE Journal, 66(2), 32–36.
• Oswalt, P., and Rieniets, T. (2006). Atlas of Shrinking Cities. Hatje Cantz Publishers.
•  Pan,  H.,  Zheng,  Y., et al. (2020). "Metro Proximity and Built Environment on Commuting CO2 Emissions in Shanghai". International Review for Spatial Planning and Sustainable Development, 8(2), 18–36. https://doi.org/https://doi.org/10.14246/irspsd.8.2_18
•  Qian,  X., and  Ukkusuri , S. V. (2015). "Spatial variation of the urban taxi ridership using GPS data". Applied Geography, 59, 31–42. https://doi.org/https://doi.org/10.1016/j.apgeog.2015.02.011
•  Sakamoto,  K.,  Iida,  A., et al. (2018). "Spatial patterns of population turnover in a Japanese Regional City for urban regeneration against population decline: Is Compact City policy effective?" Cities, 81, 230–241. https://doi.org/https://doi.org/10.1016/j.cities.2018.04.012
•  Shi,  Z.,  Zhang,  N., et al. (2018). "Exploring Spatiotemporal Variation in Hourly Metro Ridership at Station Level: The Influence of Built Environment and Topological Structure". Sustainability, 10(12), 4564. https://doi.org/https://doi.org/10.3390/su10124564
•  Sohn,  K., and  Shim,  H. (2010). "Factors generating boardings at Metro stations in the Seoul metropolitan area". Cities, 27(5), 358–368. https://doi.org/https://doi.org/10.1016/j.cities.2010.05.001
•  Soltani,  S.,  Gu,  N., et al. (2020). "Spatial Configuration and Density: How Building Density Affects Spatial Arrangement of a Neighbourhood". International Review for Spatial Planning and Sustainable Development, 8(3), 87–100. https://doi.org/https://doi.org/10.14246/irspsda.8.3_87
•  Sung,  H.,  Choi,  K., et al. (2014). "Exploring the impacts of land use by service coverage and station-level accessibility on rail transit ridership". Journal of Transport Geography, 36, 134–140. https://doi.org/https://doi.org/10.1016/j.jtrangeo.2014.03.013
•  Sung,  H., and  Oh,  J.-T. (2011). "Transit-oriented development in a high-density city: Identifying its association with transit ridership in Seoul, Korea". Cities, 28(1), 70–82. https://doi.org/https://doi.org/10.1016/j.cities.2010.09.004
• Szymanowski, M., and Kryza, M. (2012). "Local regression models for spatial interpolation of urban heat island—an example from Wrocław, SW Poland". Theoretical and Applied Climatology, 108(1), 53–71. https://doi.org/https://doi.org/10.1007/s00704-011-0517-6
•  Tang,  T.,  Liu,  R., et al. (2020). "Incorporating weather conditions and travel history in estimating the alighting bus stops from smart card data". Sustainable Cities and Society, 53, 101927. https://doi.org/https://doi.org/10.1016/j.scs.2019.101927
•  Taylor , B.  D.,  Miller, D., et al. (2009). "Nature and/or nurture? Analyzing the determinants of transit ridership across US urbanized areas". Transportation Research Part A: Policy and Practice, 43(1), 60–77. https://doi.org/https://doi.org/10.1016/j.tra.2008.06.007
•  Tsuboi,  S.,  Ikaruga,  S., et al. (2015). "Study on a Method of Making a Concentrated Urban Structure Model Based on an Urban Master Plan". International Review for Spatial Planning and Sustainable Development, 3(4), 33–45. https://doi.org/https://doi.org/10.14246/irspsd.3.4_33
•  Van Acker,  V., and  Witlox,  F. (2010). "Car ownership as a mediating variable in car travel behaviour research using a structural equation modelling approach to identify its dual relationship". Journal of Transport Geography, 18(1), 65–74. https://doi.org/https://doi.org/10.1016/j.jtrangeo.2009.05.006
•  Vergel-Tovar , C. E., and  Rodriguez , D. A. (2018). "The ridership performance of the built environment for BRT systems: Evidence from Latin America". Journal of Transport Geography, 73, 172–184. https://doi.org/https://doi.org/10.1016/j.jtrangeo.2018.06.018
•  Westerink,  J.,  Haase,  D., et al. (2013). Dealing with Sustainability Trade-Offs of the Compact City in Peri-Urban Planning Across European City Regions. European Planning Studies, 21(4), 473–497. https://doi.org/https://doi.org/10.1080/09654313.2012.722927
• Yan, X., and Su, X. G. (2009). Linear Regression Analysis: Theory and Computing. World Scientific Publishing Co. Pte. Ltd.
•  Zhang, X.,,  Zhang, Q., et al. (2018). "Evaluation of urban public transport priority performance based on the improved TOPSIS method: A case study of Wuhan". Sustainable Cities and Society, 43, 357–365. https://doi.org/https://doi.org/10.1016/j.scs.2018.08.013
•  Zhao,  J.,  Deng,  W., et al. (2013). "What influences Metro station ridership in China? Insights from Nanjing". Cities, 35, 114–124. https://doi.org/https://doi.org/10.1016/j.cities.2013.07.002
•  Zhao,  J.,  Deng,  W., et al. (2014). "Analysis of Metro ridership at station level and station-to-station level in Nanjing: an approach based on direct demand models". Transportation, 41(1), 133–155. https://doi.org/https://doi.org/10.1007/s11116-013-9492-3
•  Zhao,  P., and  Hu,  H. (2019). "Geographical patterns of traffic congestion in growing megacities: Big data analytics from Beijing". Cities, 92, 164–174. https://doi.org/https://doi.org/10.1016/j.cities.2019.03.022
•  Zhao,  Y.,  Zhang,  H., et al. (2018). "Improving the approaches of traffic demand forecasting in the big data era". Cities, 82, 19–26. https://doi.org/https://doi.org/10.1016/j.cities.2018.04.015