Introduction

As the global urbanization trend continues to progress, the discourse surrounding transportation has gained increasing significance. The proliferation of motor vehicles has given rise to various challenges, including traffic congestion and atmospheric contamination, thereby imposing a multitude of predicaments upon urban locales1. Public transportation, as an alternative to private transport, has proven to be an effective solution to address transportation2. Nevertheless, within renowned tourist destinations where the efficacy of public transportation is crucial, issues such as comfort, punctuality, and accessibility have been criticized. In response to these challenges, shared bicycles have witnessed rapid development over the past decade, being introduced in major cities and hailed as one of China’s “new four inventions”3. With technological advancements, electric-powered shared bicycles, as an upgraded version of shared bicycles, have greatly facilitated urban commuting. In recent years, shared electric bicycles have become the mainstream choice for the “first mile” and “last mile” of urban tourist travel4. Shared electric bicycles are not only a clean and reliable mode of transportation but are also characterized by requiring less road space compared to other transportation methods5,6. Furthermore, they enhance the passenger flow of public transportation systems, promoting the integration of shared electric bicycles with public transit7,8. In China, shared electric bicycle projects and urban rail transit systems are undergoing unprecedented expansion, becoming mainstream not only in major cities but also gradually extending to suburban and small-to-medium-sized cities9, with particularly rapid expansion in tourist cities. In areas lacking bus services, where the distance between the nearest bus stop and the destination exceeds a reasonable walking distance, shared electric bicycles can serve as an effective means of reaching certain locations. Their emergence and development have facilitated the integration of shared electric bicycles with public transportation in urban areas, making them a significant research focus in the field of shared mobility10,11.

Kunming, situated in the southwest of China, stands as one of the nation’s vital tourist destinations12. The inception of shared bicycles in Kunming dates back to 2016, and since then, bicycle usage has witnessed a steady ascent. Presently, approximately 6 out of every 100 residents in the Kunming urban area opt for bicycles as their primary mode of transportation. According to the “2022 Annual Report on Shared Bicycle/Electric Bicycle Riding in Major Cities in China,” the utilization of shared bicycles among commuters in Kunming surpasses 80%, significantly outstripping the national average of 50% in cities with shared bicycle programs, ranking Kunming third nationwide. Furthermore, the reduction in carbon emissions per vehicle in Kunming has risen by 17.6 kg per year, underscoring the notable popularity and effectiveness of the shared bicycle system in the city. As of June 7, 2023, the total count of shared bicycles and electric bicycles in Kunming has reached 130,000. As the capital of China, Beijing has also actively explored the field of shared electric bicycle. However, as a super-large city, Beijing faces more complex traffic management and urban planning issues. As a highly urbanized country, Singapore’s shared electric bicycle system also presents unique development characteristics. However, due to the limited urban area, the parking and charging facilities of shared electric bicycle have become an important factor restricting its development. As a famous tourist city, Kunming attracts a large number of tourists with its spring-like climate, beautiful scenery, and good air, which provides a unique advantage for the development of shared electric bicycle. In terms of urban structure, Kunming’s urban structure is relatively flat, the population density is large, and the road infrastructure is relatively complete, making cycling a convenient and comfortable way of travel. In terms of transportation mode, Kunming’s shared electric bicycle have become an important choice for citizens and tourists to travel with their high flexibility. This flexibility is not only reflected in the seamless connection with public transportation, but also in the ability to quickly shuttle between various scenic spots. Hence, investigating the usage patterns of shared electric bicycles in Kunming bears immense significance for the city’s urban management and the advancement of its status as a tourist destination.

Shared mobility trends

With the continuous development of intelligent technology and big data, the transportation field is undergoing profound changes. Intelligent transportation systems (ITS) can realize real-time collection, processing and analysis of traffic data through technologies such as the Internet of Things (IoT), big data and artificial intelligence (AI), thus playing an important role in traffic scheduling, route optimization and travel mode prediction13. As a new mode of transportation, the operational efficiency and user experience of shared electric bicycles are increasingly dependent on these advanced technologies14. Not only is the technology innovating, but the research perspective of shared electric bicycles is also constantly changing. Urban tourism has emerged as a pivotal focal point in the development of cities as society enters the urbanization era, attracting substantial investments in land, capital, and expert labor15,16,17,18. Nonetheless, research concerning the transportation facets of tourist cities, both domestically and internationally, remains relatively limited. Macao, being a prominent international tourist city, has seen studies conducted by Zhao et al.19and Qiao20 delving into public bus service standards and passenger satisfaction data, employing commonly used satisfaction models (CCSI, ECSI, SCSB, and ACSI). Cui et al.21, utilizing the ACSI model, conducted a survey encompassing 267 passengers and constructed a passenger satisfaction model for online taxi services. The existing body of research on urban development in tourist cities primarily hinges on survey data to explore satisfaction. In this study, we adopt a geospatial and temporal approach to scrutinize travel and planning development in tourist cities, yielding more visually compelling and pragmatic research findings to effectively bolster the progress of tourist cities.

Transportation interconnections

The interconnection between shared bicycle systems and other modes of transportation has long been a prominent subject of study in the field of shared bicycles. Most of the literature on bicycles as transportation interconnections has explored the factors influencing the choice of bicycles for entering and exiting subways, buses, or trains, especially in developing countries. For instance, Ji et al.22 examined the disparities in the utilization of shared bicycles and subways among individuals in various demographic groups. They reported that individuals who had experienced bicycle theft and young males were more inclined to use shared bicycles to reach subway stations. Tang et al.23 investigated the capacity of public transportation stations for accommodating shared bicycles, focusing on the first and last kilometer issues of travel. Their findings suggested that the shared bicycle system would encourage more people to utilize public transportation. Ma et al.24 analyzed usage patterns in shared bicycle-subway connections, considering different dimensions of transfers. They observed that young users constituted the largest proportion of this travel mode, with significant variations based on user age. Regarding data sources, Bachand-Marleau et al.25 conducted a study using survey data collected from shared bicycle systems and found that most subway commuters also used shared bicycles. In related studies, Ji et al.26 and Zhao et al.27utilized smart card data to more precisely determine commuters’ utilization of shared bicycles and subway transfers. Additionally, Radzimski and Dzięcielski28 reported that an increase in public transportation frequency led to more shared bicycle travel, while Yang et al.29noted that 62.8% of commuters in the suburbs of Nanjing planned to shift their mode of travel to using shared bicycles for subway connections. Mobility as a Service (MaaS), as an integrated travel platform, is becoming an important part of future urban transportation. MaaS connects various modes of transportation, such as shared electric bicycles, public transportation, taxis and car rentals, through a unified digital platform to provide users with one-stop travel services30. Users can plan and pay for the entire travel process through the MaaS platform without worrying about the conversion and payment issues between different modes of transportation31. For shared electric bicycles, the integrated characteristics of MaaS enable electric bicycles to be seamlessly connected with other modes of transportation, improving the convenience of users’ travel. For example, when users use public transportation, electric bicycles can be used as a last-mile travel mode to solve the problem of short-distance travel between bus stops or subway stations and destinations32. The popularization of MaaS platforms will further promote the integration of shared electric bicycles into urban transportation systems and improve the efficiency and sustainability of urban transportation33. However, most of these studies concentrated on the interconnection between shared bicycles and subways in specific cities, with limited exploration of the interconnection with bus stations. Moreover, many studies relied on survey data for interconnections, which offer a macroscopic and spatially limited perspective; furthermore, detailed exploration of spatiotemporal heterogeneity was lacking to meet the current requirements of precise urban management.

Built environment impacts

The built environment encompasses human-made spaces and infrastructure that support daily activities, including living, working, and recreation34. Similarly, in recent studies, the built environment has emerged as a crucial influencing factor on people’s travel patterns35,36, particularly concerning shared bicycles. Scholars have jointly researched this topic, investigating the impact of various built environment factors on shared travel. Weliwitiya et al.37found that the proportion of local low-speed roads significantly influenced the number of people cycling to stations. Barajas38, in a study on road networks, discovered that higher intersection density on American roads led to more people cycling to stations. Wu et al.39 observed that population and residential land density positively affected bicycle traffic volume. However, it was noted by Hu et al.40that near the city center, population density exhibited a negative impact, gradually transitioning to a positive influence within a specific distance from the city center. In the case of commercial land density, it was considered positively correlated by Cheng and Lin41 and Zhou et al.42, whereas Chan and Farber43 regarded it as negatively correlated. In other investigations, a negative impact was discerned in the case of employment density, as found by Ma et al.44, yet Wang et al.’s45 research indicated a positive correlation between employment density and shared bicycle travel volume. These varying results highlight the contextual differences in different cities. However, most of these studies primarily focused on the impact of built environment factors on the travel volume of shared bicycles, with limited exploration of the interconnection between the built environment and shared bicycles and public transportation. To address this gap, some scholars have employed modeling methods. Zhang et al.46 employed multiple linear regression models to investigate the influence of building environmental factors on the interconnection of shared bicycles with public transportation stations. Generalized additive mixed modeling was utilized by Cheng et al.47to examine the impact of the built environment surrounding stations on the joint utilization of shared bicycles and subways. Negative binomial regression was applied by Guo and He48 to analyze the effect of the built environment on the combined usage of shared bicycles and subways. Their findings indicated that the density of the work area significantly influenced commuting demand, resulting in a substantial increase in the combined usage of shared bicycles and subways. However, it is worth noting that most of these modeling methods originate from mathematical modeling. In addressing the spatial aspects of the interconnection between shared bicycles and public transportation systems, a geospatial and temporal perspective would be more pertinent. Therefore, this paper integrates the above research ideas, starting from a spatiotemporal perspective, and uses various modeling methods for comparison to study the degree to which the built environment impacts the interconnection of shared bicycles with subways and buses in Kunming, a key tourist city. This study provides corresponding guidance and recommendations for urban planning.

Therefore, this paper integrates the above research ideas, starting from a spatiotemporal perspective, and uses various modeling methods for comparison to study the degree to which the built environment impacts the interconnection of shared bicycles with subways and buses in Kunming, a key tourist city. This study provides corresponding guidance and recommendations for urban planning.

Study area overview and data sources

Study area

As shown in Figure 1, the research area is situated in Kunming, Yunnan Province, China, a notable city renowned for its prominence in tourism, commerce, and trade. Administratively, the city is divided into 7 districts, 3 counties, overseeing 1 county-level city, and 3 autonomous counties. Covering a total area of 21,012.54 square kilometers, Kunming is home to a permanent population of 8.6 million people, with an urbanization rate of 81.1%. In November 2019, the Kunming Municipal Urban Management Committee formally introduced shared electric bicycles, with an initial deployment of 9,000 bicycles. Shared electric bicycles, as an advanced iteration of shared bicycles, rely on electric propulsion, providing users with a more convenient mode of transportation. At present, shared electric bicycles have become widely popular and have become a convenient public transportation option for short-distance travel in cities. Presently, shared electric bicycles have garnered widespread popularity, serving as a convenient choice for short-distance urban travel in the realm of public transportation. For this study, the primary urban region of Kunming, including Chenggong District, Guandu District, Panlong District (urban area), Wuhua District (urban area), and Xishan District, was designated as the research area. These districts represent areas characterized by a concentrated population and relatively high levels of socioeconomic development within Kunming. They include various land-use functions, including dining establishments, residential zones, workplaces, educational institutions, extensive commercial districts, and transportation hubs. The selected regions are indicative of the overall landscape of short-distance travel within Kunming, reflecting high-demand scenarios for shared electric bicycle utilization. Notably, Kunming has experienced substantial demand for shared electric bicycle rides, rendering it a noteworthy representation of short-distance travel patterns within the chosen research area. As an essential transportation tool for bridging the “first mile” and “last mile,” shared electric bicycles continue to witness increasing demand. The role of shared electric bicycles in this densely populated, contemporary tourist-centric city is of paramount importance, intimately intertwined with the daily travel patterns of both residents and tourists alike. Consequently, the selection of shared electric bicycle data from Kunming holds considerable significance in the context of municipal traffic management.

Fig. 1
figure 1

Study area. The map was created using ArcGIS Pro (version 3.0.1, https://www.esri.com/en-us/arcgis/products/arcgis-pro).

Full size image

Data sources

Shared bicycle spatiotemporal data

Spatiotemporal data pertaining to shared bicycles during the period of April 23 to 29, 2022, were acquired from the Urban Management Bureau of Kunming, Yunnan Province, China. The dataset includes in excess of one million trip records, including particulars such as unique identifiers for shared electric bicycles, commencement times, conclusion times, and the geographical coordinates (latitude and longitude) of the inception and termination points of each bicycle’s journey. To narrow our scope to the designated research area, trip records situated beyond the stipulated longitude and latitude boundaries were excluded. Furthermore, any defective data, including duplicate entries, critical omissions, and location discrepancies, were eradicated. To ensure data reliability and consistency, several preprocessing steps were undertaken to enhanced the dataset’s accuracy and usability for spatial and temporal analysis. Data cleaning was performed by removing null values, corrupted entries, duplicate records, and trip records situated outside the stipulated longitude and latitude boundaries of the designated research area. Data filtering was further applied to eliminate anomalies caused by GPS errors or human factors. Subsequently, the dataset was integrated into vectorized shapefile (shp) format for subsequent geospatial analysis.

Points of interest (POI) Data

POI data were extracted from the application programming interface (API) of Amap, a widely utilized mapping service in China (https://lbs.amap.com/). The latitude and longitude for each POI are included within the dataset. The study included diverse POIs, including educational and cultural institutions, commercial enterprises, dining establishments, retail and entertainment venues, financial institutions, governmental entities, and lodging establishments(Table 1). The public transportation stations incorporated in this research comprised subway stations and bus stations. Importantly, it should be noted that the subway station map does not include Line 5, as this line commenced operation in Kunming on June 29, 2022, subsequent to the acquisition of shared electric bicycle spatiotemporal data for April 2022.

Table 1 Legend (350 words max). Example legend text.
Full size table

Road network data

The road network data used in this study were sourced from OpenStreetMap (OSM) (http://www.openstreetmap.org/). OSM has evolved into one of the largest open-source volunteered geographic information (VGI) projects. It provided freely accessible spatial data and was utilized to study road networks in this research.

Population data

Population data were procured from the WorldPop dataset (https://www.worldpop.org/). In contrast to alternative population grid datasets, WorldPop’s data demonstrates superior precision and granularity within the Chinese context. Given that this dataset furnishes worldwide gridded population information, with a spatial resolution of 100 meters for each year from 2000 to 2020, and considering the impact of the COVID-19 pandemic on population dynamics during the period from 2020 to 2022, the population data for the year 2020 were used as the reference point due to their relevance to this investigation.

Research methods

Technical approach

This paper conducts spatiotemporal data analysis on shared electric bicycles in Kunming city. When studying the impact of the built environment on the connectivity of shared electric bicycles, the connectivity situations are categorized into two types: “access” and “egress”. The modes of connectivity are divided into two types: “shared electric bicycle-to-bus connectivity” and “shared electric bicycle-to-subway connectivity.” The study period is finely divided into four segments: weekday morning peak, weekday evening peak, weekend morning peak, and weekend evening peak. For feature selection, the XGBoost machine learning algorithm was utilized to compare the importance indices of relevant factors and multicollinearity analysis was conducted to screen feature factors and avoid multicollinearity issues. Finally, four models, namely, OLS, GWR, GTWR, and MGWR, undergo comparison and analysis. The specific research approach is depicted in Figure 2.

Fig. 2
figure 2

Technical approach.

Full size image

Connectivity usage and variable processing

In the study of connectivity usage (Using shared electric bicycles and subways/buses before and after), Wu et al.49 initiated a pioneering exploration into the feasibility of using dockless shared bicycle location data to infer the integration of shared bicycles with the subway system. Based on this groundwork, Wang et al.50 conducted an analysis of individual subway station entrances and exits. They identified dockless shared bicycle trips within 100 meters of each station entrance or exit as connectivity trips aimed at subway usage. This study considers each subway station in its entirety. Through spatial analysis, the minimum distance between any two subway stations in Kunming was calculated to be 295 meters. The comprehensive use of shared bicycles and the subway in this context is determined by calculating the total effective use of shared electric bicycles within a 300-meter buffer zone around each subway station, a computation of significant research importance.

Furthermore, when compared to subway coverage, bus stations in Kunming have a more extensive network. Spatial analysis revealed that the average distance between bus stations in Kunming and their nearest neighboring bus station is 321 meters. To understand the actual impact coverage area of bus stations on travel patterns, the comprehensive use of shared bicycles and the bus is assessed by calculating the total effective use of shared electric bicycles within a 160-meter buffer zone around each bus station.

In addition, other digitally related variables were computed within a 1.6-kilometer buffer zone surrounding each transportation station. This buffer zone size was determined by examining the percentile distribution of riding distances for shared electric bicycles in Kunming. Approximately 80% of shared electric bicycle rides in Kunming are less than 20 minutes, with 80% of linear riding distances generally within 3200 meters. The median riding distance is 1603 meters. Considering the starting and ending points of a ride, the final range distance of 1600 meters was established as significant for studying the influence of the built environment on shared bicycle-bus and shared bicycle-subway scenarios. The details of the interconnection diagram are shown in Figure 3.

Fig. 3
figure 3

Feature Importance of Factors.

Full size image

The importance of the XGBoost machine learning features is shown in Figure 4. In the study of factors related to subway connectivity, features such as subway stations, hotel accommodations, and sports and fitness venues exhibited significantly lower importance indices; thus, these factors were not considered influencing factors in subsequent research. Similarly, in the examination of factors related to bus connectivity, subway stations were deliberately omitted as influential factors in subsequent investigations. Comprehensive usage can be described separately as shared bicycle-bus journeys or shared bicycle-subway journeys, each signifying the frequency of shared electric bicycles serving as feeder modes for round trips between subway or bus stations. The dependent variable under scrutiny is the comprehensive connectivity utilization rate (The number of trips using shared electric bicycles and subways/buses before and after) of shared bicycles in conjunction with buses or subways. More specifically, this comprehensive utilization is stratified into weekdays and weekends. Additionally, in the context of this connectivity analysis, variance inflation factor (VIF) values are computed for each feature factor. Feature factors exhibiting relatively elevated VIF values are gradually eliminated to ensure that the final VIF indices for all feature factors remain below 10; thus, averting issues associated with multicollinearity among independent variables51. Subsequent to the data analysis and processing, within the realm of shared electric bicycle-subway connectivity, five feature factors-namely, financial institutions, tourism and entertainment venues, shopping and dining establishments, automobile-related establishments, and life service establishments-have been excluded. In parallel, within the framework of shared electric bicycle-bus connectivity, an equivalent set of five factors-comprising financial institutions, hotel accommodations, shopping and dining establishments, automobile-related establishments, and life service establishments-have also been excluded from consideration.

Fig. 4
figure 4

Feature Importance of Factors.

Full size image

Model selection

Ordinary Least Squares (OLS)

In this study, the OLS model assumes the role of the fundamental model and can be articulated as Equation 1.

$$\begin{aligned} y_i=\beta _0+\sum _{j=1}^m\beta _jx_{ij}+\varepsilon _i \end{aligned}$$
(1)

\(\textrm{y}_{i}\) represents the dependent variable, \(x_{ij}\) is the \(\textrm{j}\)-th independent variable, \(\textrm{m}\) is the total number of independent variables, \(\beta _{0}\) denotes the estimated intercept, \(\beta _{j}\) represents the regression coefficient of the \(\textrm{j}\)-th independent variable, and \(\varepsilon _i\) denotes the random error term.

Geographically Weighted Regression (GWR)

GWR, a local coefficient regression model, addresses spatial heterogeneity. GWR produces locally varying regression coefficients, deriving these coefficients from the spatial location information of sample data. The fundamental formula is presented as Equation 2.

$$\begin{aligned} Y_i=\beta _0(u_i,v_i)+\sum _k\beta _k(u_i,v_i)X_{ik}+\varepsilon _i \end{aligned}$$
(2)

\({Y}_{i}\) refers to the dependent variable value for the \(\textrm{i}\)-th observation, \(X_{ik}\) is the \(\textrm{k}\)-th explanatory variable for the \(\textrm{i}\)-th observation, \((u_i,v_i)\) indicates the spatial location of the \(\textrm{i}\)-th observation, \(\beta _0(u_i,v_i)\) represents the constant term for the iii-th observation, \(\beta _k(u_i,v_i)\) denotes the regression coefficient for the \(\textrm{k}\)-th explanatory variable, and \(\varepsilon _i\) represents the random error term.

Spatiotemporal geographically weighted regression (GTWR)

While the regression parameters of the independent variables in the GWR model vary with spatial geographic location, in the GTWR model, these parameters vary with both spatial and temporal location. The basic expression of the GTWR model is as Equation 3.

$$\begin{aligned} y_i=\beta _0(u_i,v_i,t_i)+\sum _{k=1}^k\beta _k((u_i,v_i,t_i))x_{ik}+\varepsilon _i \end{aligned}$$
(3)

\((u_i,v_i)\) denotes the geographic coordinates (longitude and latitude) of the \(\textrm{i}\)-th sample point, \({t}_{i}\) is the observation time, \({y}_{i}\) is the dependent variable value for the iii-th sample point, and \(X_{ik}\) refers to the \(\textrm{k}\)-th explanatory variable for the iii-th sample point. \(\beta _0(u_i,v_i,t_i)\) represents the regression constant for the iii-th sample point, \(\beta _k(u_i,v_i,t_i)\) is the regression coefficient for the \(\textrm{k}\)-th explanatory variable, and \(\varepsilon _i\) denotes the model’s error term.

Multi-Scale Geographically Weighted Regression (MGWR)

Multi-scale geographically weighted regression (MGWR) developed by the Oshan team at Arizona State University52. In the MGWR model, a separate bandwidth is determined for each independent variable, thereby enabling spatial variation in the relationship scale between the independent variables and the dependent variable. This multi-bandwidth approach yields spatial process models that are more reflective of reality53, as elucidated in Equation 4.

$$\begin{aligned} y_i=\beta _{bw0}(u_i,v_i)+\sum _{i=1}^m\beta _{bwj}\big ((u_i,v_i)\big )\chi _{ij}+\varepsilon _i \end{aligned}$$
(4)

Here, \(\beta _{bw0}(u_i,v_i)\) is the local intercept for the i-th observation, \(\beta _{bwj}((u_i,v_i))\) is the parameter for the j-th independent variable \(x_{ij}\)’s \(\varepsilon _{i}\) random error term, and \((u_i,v_i)\) is the spatial location of the j-th observation. In this study, urban built environment factors are the independent variables, and the comprehensive connectivity utilization rate is the dependent variable.

The evaluation of the models

The four evaluation indicators of RSS, AICc, \(\hbox {R}^{2}\) and adjusted \(\hbox {R}^{2}\) are of great significance in the study of shared electric bicycles. They can evaluate the predictive ability, fitting effect and explanatory ability of the model from different angles, thus helping us to more accurately understand the development status and management effect of shared electric bicycles. RSS can reflect the difference between the model prediction value and the actual observation value. AICc is a standard for measuring the goodness of fit of a statistical model. Through AICc, the optimal model can be screened out from multiple models to more accurately describe the relevant data of shared electric bicycles. Through \(\hbox {R}^{2}\), the explanatory power of the model can be evaluated, so as to select the model with the strongest explanatory power. Through adjusted \(\hbox {R}^{2}\), the optimal model can be screened out from multiple models, while avoiding the problem of overfitting53,54. Smaller RSS and AICc values indicate better model performance, while larger \(\hbox {R}^{2}\) and Adjusted \(\hbox {R}^{2}\) values indicate better model fit.

Results and discussion

Model comparison

In this study, the OLS model, GWR model, and MGWR model were compared using four indicators: RSS, AICc, \(\hbox {R}^{2}\), and adjusted \(\hbox {R}^{2}\). Smaller RSS and AICc values indicate better model performance, while larger \(\hbox {R}^{2}\) and Adjusted \(\hbox {R}^{2}\)values reflect a stronger goodness-of-fit55. According to the Table 2 and Table 3, MGWR shows significantly better performance in model fitting, outperforming the OLS, GWR, and GTWR models, with the GTWR model notably performing worse (Unlike other models, the GTWR model incorporates temporal variables, making it unsuitable for direct comparison based solely on peak periods). Taking the case of shared electric bicycles to bus access during weekday morning peak hours as an example, the \(\hbox {R}^{2}\) values are 0.580 in the OLS model, 0.661 in the GWR model, and 0.731 in the MGWR model, signifying a noteworthy enhancement in the numerical values. The RSS values are 452.336 in the OLS model, 364.673 in the GWR model, and 289.039 in the MGWR model, indicating a substantial reduction in the numerical values. Through comprehensive comparative analysis, it becomes evident that the MGWR model exhibited a more robust explanatory power in this study. Subsequent analyses in this paper will primarily focus on interpreting the results of the MGWR model. The superior fitting capability of the MGWR model enables a more accurate analysis of the spatial variation effects in this research. Additionally, MGWR effectively leverages the spatial relationships among independent variables, addressing potential disparities in the built environment across different regions of Kunming.

Table 2 Comparison of the egress models.
Full size table
Table 3 POI Categories.
Full size table

MGWR analysis

The MGWR analysis is separately conducted for four scenarios: access to shared e-bike and bus, egress from shared e-bike and bus, access to shared e-bike and subway, and egress from shared e-bike and subway. The purpose of this analysis is to examine the influence of built environment features on the overall usage of shared e-bikes in conjunction with buses or subways.

Table 4 presents a summary of the statistical data for local parameter estimates derived from the MGWR model for the shared e-bike and bus access and egress scenarios. Table 5 provides the statistical data for local parameter estimates obtained from the MGWR model for the shared e-bike and subway access and egress scenarios. These tables indicate the percentage of coefficients that are statistically significant (P \(\le\) 0.1). with P \(\le\)0.1 indicating a meaningful relationship between the explanatory variables and the dependent variable at a 10% significance level56.

Table 4 Significant coefficient percentages of environmental factors for shared electric bicycle-bus output and input(Each factor corresponds to the first row of data for output and the second row of data for input).
Full size table
Table 5 Percentage of significant coefficients for environmental factors in shared e-bikes and subway egress and access(The first row corresponds to departures, and the second row corresponds to arrivals).
Full size table

As illustrated in Figure 5,6,7.8 positive parameter estimates indicate a positive impact of explanatory variables on overall usage, while negative parameter estimates suggest a negative impact. Additionally, regions shaded in gray denote variables that do not have a significant effect on the overall usage of shared e-bikes and public transportation. Similar to the research on subways, Shared e-bike-to-bus egress refers to riding a shared electric bicycle and then transferring to the bus. Shared e-bike-to-bus access refers to riding a shared electric bicycle after taking the bus. The categorization of coefficient values is based on the natural break classification method as offered by ArcGIS Pro 3.0.1 software, dividing the coefficient range into five segments for analysis.

Fig. 5
figure 5

Spatial distribution of local coefficients for shared e-bikes to bus egress. The map was created using ArcGIS Pro (version 3.0.1, https://www.esri.com/en-us/arcgis/products/arcgis-pro).

Full size image
Fig. 6
figure 6

Spatial distribution of local coefficients for shared e-bike-to-bus access. The map was created using ArcGIS Pro (version 3.0.1, https://www.esri.com/en-us/arcgis/products/arcgis-pro).

Full size image
Fig. 7
figure 7

Spatial distribution of local coefficients for shared e-bikes and subway egress. The map was created using ArcGIS Pro (version 3.0.1, https://www.esri.com/en-us/arcgis/products/arcgis-pro).

Full size image
Fig. 8
figure 8

Spatial distribution of local coefficients for shared e-bikes and subway access. The map was created using ArcGIS Pro (version 3.0.1, https://www.esri.com/en-us/arcgis/products/arcgis-pro).

Full size image

Shared electric bike-bus egress and access

For the analysis of the impact on companies and business establishments, Zhou et al.42identified a positive effect of the quantity of workplace facilities on the combined usage of shared e-bikes and bus connections. Guo and He48 uncovered a negative correlation between employment density and the joint utilization of bicycles and public transportation in Shenzhen. In this investigation, apart from the weekend morning peak period, corporate and business establishments across various areas in Kunming exhibit a significant positive correlation with shared e-bike and bus access and egress during each peak period. The local coefficient is notably higher in the Chenggong area, primarily because Kunming’s workplace facilities are not saturated, and shared e-bikes and buses play a pivotal role in the commuting process.

Concerning the influence of government institutions, there is no significant impact during the morning peak period for shared e-bike and bus egress. However, during other time periods for shared e-bike and bus egress, as well as all time periods for access, positive correlations are evident. The presence of various government institutions, as integral components of the city, contributes to an increase in the number of facilities around bus stations, thereby promoting combined travel involving shared e-bikes and buses.

In the assessment of the impact of tourist and entertainment destinations, similar to Zhou et al.’s research42in Beijing, significantly negative local coefficients are observed for leisure services in the urban area, indicating a relatively strong association between leisure services and the comprehensive use of public transportation. These leisure service areas correspond to the tourist and entertainment venues in this study. In alignment with Chan and Farber’s study43, which identified a negative correlation between the proportion of commercial land and bicycle arrivals at bus stops, this study in Kunming reveals that tourist and entertainment places only exhibit significance during the weekend morning peak period for shared e-bike and bus egress, consistently displaying negative local coefficients. As a prominent tourist city, Kunming has seen a saturation in the number of tourist and entertainment venues, emphasizing the need to focus on the quality of these establishments.

In a study on the impact of educational and cultural places, Wang et al.57 identified a positive correlation between shared bike travel and bike stations near universities. In this study, scientific, educational, and cultural factors exhibit a positive correlation with shared e-bikes and bus access and egress during the weekend evening peak period, while displaying a negative correlation during other significant time periods. For areas with a concentration of scientific, educational, and cultural resources, the weekend evening peak period is the peak period of student movement throughout the week, during which the combined travel of shared e-bikes and buses is effectively promoted.

In the analysis of the impact of medical and insurance places, there is a positive correlation during significant time periods for shared e-bike and bus combined travel, with the most significant impact during the weekday morning peak period. This is closely related to the service hours of healthcare places during working days.

Regarding the impact of sports and fitness places, the significantly affected areas exhibit relative scattering, but all demonstrate a positive correlation with shared e-bikes and bus travel access and egress. Notably, the local coefficients display significant increases in the northern region of Panlong District, the central area, and the northern sector of Chenggong District.

According to the analysis of bus station density, for shared e-bikes and bus travel, there is a positive correlation only during the morning peak period during egress. The uncertainty of bus arrival times and the desire for faster travel during the morning peak period result in a higher density of bus stops around a destination, increasing the potential for connecting with shared e-bikes. In the case of shared e-bikes and bus access, the denser the bus stop is around a bus stop, the stronger the competitive relationship is between the surrounding stops, particularly during the evening peak period when more destinations become random, and demand is not high.

In the investigation of road network density, as observed by Barajas38, the density of road intersections is associated with an increased usage of bicycles for reaching bus stops. In this research, road network density exhibits significance during the daily morning peak period for access and the weekend morning peak period for egress. During the daily morning peak period for egress, there is a negative correlation, whereas during the weekend morning peak period for access, a positive correlation is evident.

Regarding the analysis of population density, as highlighted in Chan and Farber’s study43, population density plays a constructive role in facilitating combined travel involving multiple modes of transportation. This observation is also validated in the study area, where population density significantly influences shared e-bike and bus travel during the weekday morning peak period for egress, and both exhibit a positive correlation.

Shared electric bike-Subway egress and access

According to Table 5, government institution locations and road density factors are not significant for shared e-bikes or subway connections.

A study on the impact of companies and business places revealed that during the morning peak hours, there is a significant positive correlation in the southern region between shared e-bikes and subway connections, while some areas near Kunming Changshui Airport show no significant correlation. The number of workplaces was found to positively impact the use of shared e-bikes and subways, especially in the Chenggong district. In contrast, Ma et al.44 found in their Nanjing study that higher workplace density led to a decrease in the use of bikes and subways, possibly because Nanjing’s workplace density was already high, unlike Kunming, where it is not as saturated. In a study on the impact of educational and cultural places, Hu et al.40 discovered that shared bikes are more popular in subway station areas with universities due to increased trip frequency, higher BnRs, and enhanced usage. In this study, this factor is significant solely during the evening peak period in the departure scenario and remains significant outside the Chenggong district. All factors significantly influencing subway stations exhibit positive local coefficients. For weekend evening peak access, Chenggong District displays higher local coefficients, while in other shared e-bikes and subway access and departure periods, the main urban area exhibits higher local coefficients. This indicates that travel originating from scientific, educational, and cultural places primarily occurs in the main urban area, thereby significantly promoting the use of shared e-bikes and subway connections.

A study on the impact of medical and insurance places reveals their significance at all time periods for both access and departure, with higher coefficients observed in the Wuhua and Panlong districts during the weekend morning peak access.

Regarding the impact of bus station density, Wu et al.39 reported a negative correlation between bus station density and shared bike and subway combinations in Shenzhen. Similarly, Zhou et al.42 reported that an increase in the number of bus stations has a negative impact on combined shared bike and subway trips in Beijing. This study in Kunming yields analogous findings, pointing to a negative correlation between bus station density and shared e-bikes and subway access and departure. It is noteworthy that significantly lower local coefficients are observed in the Wuhua, Panlong, and Xishan districts.

In a study on the impact of population density, Hu et al.40 reported that the impact of shared bikes and subways combined with travel is negative near city centers and positive further away. Guo et al.58 concluded that population density has a positive impact on multimodal combined travel. However, in this study, the population density factor does not exhibit significance during the morning peak period for shared e-bikes or subway departures and displays a negative correlation in various areas. During the evening peak period, the Chenggong and Guandu districts exhibited a stronger negative correlation, while during the morning peak period, the Wuhua, Guandu, and Panlong districts showed a stronger negative correlation. As a nationally important tourist city, Kunming’s relatively high population density limits people’s inclination to engage in excessive travel.

Conclusion

Studying the transportation patterns in a tourist city is pivotal for urban planning and tourism development. Environmental characteristics play a crucial role in the connectivity of shared e-bikes with other transportation modes. This study conducts a comparison of four models-OLS, GWR, GTWR, and MGWR-and concludes that the MGWR model significantly outperforms in the analysis of shared e-bikes in Kunming. The analysis differentiates between access and egress and includes two modes of connectivity: shared e-bike to bus and shared e-bike to subway. Furthermore, the study categorizes time into four periods: weekday morning peak, weekday evening peak, weekend morning peak, and weekend evening peak.

The research findings indicate that a positive correlation exists among companies in various regions of Kunming during peak periods concerning shared e-bikes and bus connectivity. Government institutions exert a positive influence on shared e-bikes and bus egress during nonpeak times and on access throughout all periods. The location of tourism and entertainment venues significantly impacts shared e-bikes and bus egress only during the morning peak, consistently yielding negative local coefficients. Educational and cultural institutions demonstrate positive correlations during the weekend evening peak, affecting both access and egress, but display negative correlations during other significant periods. Medical facilities significantly influence shared e-bike and bus joint travel, particularly during weekday morning peaks. Sports and fitness facilities exhibit positive correlations with shared e-bikes and bus access and egress, particularly in specific regions. Bus station density has a negative impact on shared e-bikes and bus access solely during the morning peak. Road network density plays a significant role during weekday morning access and weekend evening egress, yielding negative and positive correlations, respectively. Population density shows significant correlations with shared e-bikes and peak egress during bus mornings, displaying positive correlations.

Regarding shared e-bikes and subway connectivity, companies in the southern region significantly affect egress during the morning peak, whereas Kunming Changshui Airport access does not demonstrate significance. Workplace density positively influences shared e-bikes and subway connectivity. Educational and cultural institutions, particularly outside the Chenggong district, significantly impact egress during the evening peak. All significant subway stations exhibit positive local coefficients, with higher coefficients observed in Chenggong during weekend evening peaks. Medical facilities significantly influence shared e-bike and subway joint travel throughout all periods, with higher coefficients in Chenggong during weekend morning peaks. Bus station density negatively impacts shared e-bike and subway connectivity, especially in the Wuhua, Panlong, and Xishan districts. Population density does not significantly impact shared e-bikes or subway morning peak egress but exhibits negative correlations in various regions during evening peaks.

The study offers specific policy recommendations based on the research findings. For shared e-bikes and bus connections, it is advisable to increase the number of shared e-bikes in proximity to company areas, particularly in Chenggong, and near bus stops. Concerning shared e-bikes and subway connectivity, the focus should be on workplace areas, along with an increase in the number of shared e-bikes around subway stations, particularly in Chenggong. Additionally, adjustments to the number of shared e-bikes near high-density bus stops, especially in Wuhua and Panlong, are suggested. Due to the saturation of tourist and entertainment venues, an emphasis on enhancing the quality of these locations is recommended.

This study reveals the significant impact of the built environment on the use of shared electric bicycles for connection with public transportation, and provides some specific suggestions for implementing these findings in urban policies and pilot projects. In terms of the integration of shared electric bicycles with public transportation systems, the layout of shared electric bicycle connection points should be optimized, and shared electric bicycle parking points should be reasonably arranged near high-frequency transportation transfer points such as subway stations, bus stations and commercial areas. In the operation of shared electric bicycles, ticketing can be integrated with the urban public transportation system. In order to encourage citizens to choose shared electric bicycles in their daily travel, the government can introduce subsidy policies, such as providing discounts for users of electric bicycles during specific periods (such as peak hours); for urban planning similar to tourist cities, shared electric bicycle rental points should be set up around major attractions and commercial areas so that tourists can easily switch travel modes between scenic spots, and shared electric bicycles should be included in the planning of tourist routes to form an integrated travel plan. In addition, big data and intelligent analysis technologies should be used to monitor tourists’ travel patterns and behaviors in real time, analyze tourists’ travel needs during peak travel periods, and accurately allocate shared electric bicycle resources59. It is also possible to consider combining the charging facilities of shared electric bicycles with renewable energy sources such as solar energy and wind energy to promote green and low-carbon travel60.

This study acknowledges the necessity for further refinement in future research, particularly in terms of data timespan, spatial-temporal trajectory analysis, and consideration of socioeconomic and weather factors. For example, areas with poor air quality may cause residents to have health concerns, thus affecting their willingness to ride bicycles61. Rising temperatures may lead to more frequent high temperatures in summer, which in turn affects the use of bicycles. In particular, during the hot summer period, users may be more inclined to choose vehicles with air conditioning rather than shared electric bicycles62.

Building on this foundation, future research could incorporate supplementary urban factors such as temperature, air quality, green infrastructure, and socio-economic indices to enrich the analysis and provide deeper insights into the multifaceted interactions between the built environment and shared mobility. Expanding the dataset temporally to examine seasonal or annual trends would offer a more comprehensive understanding of temporal dynamics. Additionally, cross-city comparisons could help generalize the findings, validate the methodology, and identify context-specific patterns or strategies applicable to various urban settings.