Introduction

Since the opening of the Shinkansen in 1964, which marked the birth of high-speed rail (HSR), HSR has rapidly developed worldwide, particularly in countries such as Japan, France, Italy, Germany, Spain, and China (UIC, 2023). The global length of the HSR network (HSRN) has expanded from 515 km in 1964 to 58,839 km in 2021 (UIC, 2023). Although China entered the HSR sector relatively late, launching its first line in 2008, it has since achieved remarkable growth, with its HSR network reaching 45,000 km by 2023(CR, 2024a). In the recent decade, the development of HSR has made significant progress in many countries, becoming a popular choice for tourists. For example, the number of passengers of HSR reached about 2.36 billion in China in 2019 (NRA, 2019), while the number of passengers fell to 1.56 billion in 2020 due to the impact of the COVID-19 epidemic (UIC, 2023). After the epidemic, the number of passengers quickly exceeded 3.68 billion in 2023 (CR, 2024b). Therefore, in-depth exploration of the impact of HSR on tourism development is of great significance for the further development of the Chinese tourism industry and the full utilization of HSR resources.

Tourism refers to people’s travel and temporary stays outside their usual place of residence for purposes other than migration or employment. As transportation is a crucial factor impacting tourism development, HSR, with its advantages of speed, comfort, and punctuality, has become an important mode of travel for tourists. Although the closer relationship between HSR and tourism is well-established (Yu et al., 2021; Yang & Li, 2020; Albalate et al., 2023), critical gaps remain in understanding its nationwide spatial consequences, particularly how these impacts manifest at a national or country-wide scale rather than just local or regional levels. Existing research demonstrates that HSR impacts tourism development through these aspects. For example, HSR significantly enhances accessibility and shortens inter-city temporal distances (Li et al., 2020). Furthermore, HSR impacts on inter-city tourism spatial interaction (TSI) and the overall tourism spatial structure (TSS) (Wang et al., 2016). Yet these findings are largely based on the local area along an HSR line or multiple HSR lines, leaving broader areas and dynamic questions unaddressed. However, some fundamental research questions remain unanswered. First, most studies focus on a local region along an HSR line or multiple HSR lines, ignoring the impact of HSR on the whole regions of a country. Does HSR generate consistent impacts across different economic regions? Second, while the static impacts of HSR on the tourism development of cities are documented, the dynamic evolution of cities’ roles in tourism under the HSR context remains unexamined. How do cities’ positions advantages in tourism networks evolve with HSR expansion?

Recent scholarly attention has increasingly focused on the impact of HSR on TSI and TSS (Kang et al., 2018; Park et al., 2020; Wang et al., 2019; Zhang et al., 2020). HSR is an important factor in promoting the TSI between cities, especially adjoining cities (Yin et al., 2019), and cities within a certain economic circle along HSR lines (Ren et al., 2023). Specifically, the impact of HSR on TSI is reflected in the interaction between cities (Li & Chen, 2020), which is often measured by gravity models, such as constructing tourism eco-efficiency relations and tourism economic relations (Huang et al., 2019; Liu et al., 2022). Furthermore, as cities serve as key hubs for tourism activities, analyzing inter-city connections provides crucial insights into TSS. HSR can exert a significant impact on the linkages between cities, such as affecting the distribution pattern of tourism accessibility between cities, expanding the ‘1–2’ hour travel circle to enhance cities’ attraction and radiation power to tourists (Wang et al., 2023). In addition, HSR can affect the spatial structure of tourist flow between cities and may strengthen the agglomeration effect of tourist flow (Wang et al., 2016). The change in the concentration location of tourist flow may also be affected by HSR (Yin et al., 2019; Liu & Shi, 2019). Besides, the common method used to investigate the changes in TSS is to apply the gravity model to construct relations between cities, and then utilize the social network model to examine the variations of TSS (Zhang et al., 2020). Particularly, existing studies employing social network analysis to examine HSR’s impact on TSS predominantly rely on a single metric-travel time reduction-as the sole representative indicator (Wang et al., 2018).

It is obvious that the previous studies are limited to revealing the impact of HSR on a local region along an HSR line or multiple HSR lines (Zhang et al., 2020). They ignore the impact of HSR on the whole region of a country. It should be noted that the direct use of a gravity model to construct social networks somewhat undermines the significance of HSR in the tourism structure. Additionally, cities’ roles in tourism structure are considered to be static and unchanging in previous literature. In reality, these roles are dynamic and subject to frequent shifts in response to the spatiotemporal evolution of tourism systems. To fill these gaps, this study proposes a novel method to investigate the impact of HSR on TSI and TSS. Compared with previous research, this study makes several contributions as follows. First, we not only expand the research scale from a local area in previous works to a whole area within a country, but also consider the differences of all areas along different HSR lines in the country. Second, this study reveals the evolution of inter-city TSI by using the high-speed rail network (HSRN) and proposes a novel model for exploring the changes in TSS, which includes the degree centrality and the betweenness centrality of cities in the tourism network structure. This study provides empirical support for the changes of tourism structure centers under the impact of HSR. Third, this study investigates the role and the evolution of cities in tourism structure. Finally, the obtained findings of this paper can provide some implications for promoting urban tourism development and formulating effective tourism plans in cities.

This study is organized as follows. Literature review provides a comprehensive review and summary of relevant literature. Methods introduces the study area and the data sources used, and presents a description of the methodology. Results shows the findings. Finally, Conclusion concludes the results obtained and provides some suggestions.

Literature review

Impact of HSR on TSI

Spatial interaction is an important concept in geography, referring to the flow of people, goods, information, technology, and trade between geographical areas (Liu et al., 2023). TSI refers to the dynamic relations between tourist-generating areas and destinations, as well as among destinations themselves. These dynamic relations are manifested in several aspects. For example, TSI encompasses the mutual complementation of tourism resources, the flow of tourists, and the interplay of competition and cooperation among destinations. Specifically, TSI reflects the interactive relation between tourism demand and supply, while also capturing the competitive and cooperative dynamics that shape the spatial organization of tourism activities.

Efficient and convenient transportation systems serve as the basic condition for tourism (Wandelt et al., 2023). HSR can influence TSI primarily through its spatio-temporal compression effect, which alters the cost and efficiency of inter-regional interactions (Li et al., 2019, 2023; Wang et al., 2023). On the one hand, HSR can reduce travel costs, enhancing accessibility between cities by 25% to 45% (Liu & Zhang, 2018) and promoting the flow of tourism elements, e.g., tourists, capital (Cao et al., 2013; Masson & Petiot, 2009). This strengthens interactions between neighboring cities, boosting day trips and weekend travel markets (Wang et al., 2023).

On the other hand, competition and cooperation between tourist destinations are also an important aspect of tourism interaction between regions. A lot of studies demonstrate that HSR can drive the tourism competition between cities, and some cities with limited tourism resources are being further constrained by the intense competition in the tourism market (Wang et al., 2019). Differently, HSR also can facilitate cooperation through integrated tourism routes and resource complementation (Yin et al., 2019). In other words, HSR can provide various choices of tourism routes and facilitate forming a ‘chain’ cooperation structure distributed along HSR lines (Li et al., 2020). In addition, the mutual complementation of scenic spots and the coordinated development of tourism carrying capacity among key tourism cities are promoted by HSR (Luo et al., 2023). Specifically, existing studies focus on qualitative or directional impacts of HSR on TSI but rarely quantify interaction intensity or explore city-level differences.

Impact of HSR on TSS

TSS refers to the interactive relations among tourism objects, as well as the spatial distribution and agglomeration of tourism-related elements, such as tourist attractions and cities serving as large-scale tourism complexes. HSR can reshape TSS by altering the accessibility and spatial configuration of these elements, with effects influenced by factors like city location and HSR station placement (Ortega et al., 2012). As a result, the benefits from HSR exhibit regional variations, leading to a change in TSS.

Key structural changes driven by HSR include three aspects. First, HSR can drive the transfer of the core of regional tourism and its multi-core development. In other words, HSR can elevate previously remote areas to tourist cores, shifting traditional centers or fostering multi-core patterns (He et al., 2023; Li & Chen, 2022; Wang et al., 2022). Specifically, Wang et al. (2018) agree that the alteration of cities’ hinterland in HSRN results in a tourism network structure exhibiting multiple cores. Yin et al. (2019) find that the mean center of tourist distribution is initially concentrated in Beijing while gradually transforming to Tianjin with the opening of HSR services. Second, HSR can intensify the imbalance of regional development. HSR can exacerbate the Matthew effect and corridor effect, concentrating talent, technology, and tourism activities in well-connected areas (Jiao et al., 2014; Wang et al., 2016; Yoo et al., 2023). Finally, HSR has a positive spatial spillover effect of HSR on tourism. HSR can lead to an increased number of tourists within the range from 400 to 1200 km from the destination (Li et al., 2020) and generate spillover effects on surrounding regions (Zheng et al., 2019). For example, Tian et al. (2022) find that HSR has a broader range for the geographic scope of spillover. Ren et al. (2023) provide empirical support for a positive spatial spillover effect of HSR on tourist flows by using a spatial autoregressive model.

Many other methods have been employed to measure the impact of HSR on TSS, such as ArcGIS spatial analysis (Wang et al., 2018), mean center (Yin et al., 2019), and Moran’s index (Huang et al., 2019). Moreover, the social network analysis method is usually applied to investigate the alterations in TSS (Gan et al., 2021). However, when using the method, the impact of HSR is only reflected in the travel time, disregarding the operational dynamics of HSR infrastructure, which may diminish the impact of HSR on TSS to some extent.

Regional impact of HSR on tourism

HSR’s effects on tourism vary across regions, reflecting geographic and economic disparities. In other words, the establishment and development of HRSN has diversified impacts on tourism development in different regions and cities (Zhou & Zhang, 2021). Specifically, HSR can exert positive regional effects by stimulating tourism in underdeveloped or peripheral areas through improving public services and extends tourists’ stays, thereby boosting local tourism vitality (Sun & Lin, 2018; Tian et al., 2022; Yang & Li, 2020; Wang et al., 2023). For example, HSR can improve the tourism revenue of peripheral cities in Jiangsu Province (Wang et al., 2016).

Conversely, HSR can intensify inter-regional tourism competition, posing challenges to tourism development in vulnerable regions (Wang et al., 2019, 2023). It squeezes the market share of areas far from HSR lines or lacking sufficient tourism resources (Liu et al., 2020). Furthermore, HSR can facilitate the concentration of economic activities in developed cities (Li et al., 2022), thereby further marginalizing less competitive regions. For instance, the construction of HSR lines in Southern Europe leads to a spatial concentration of tourism activities primarily centered around Barcelona (Masson & Petiot 2009). However, these studies mainly focus on a small or local area along an HSR line or multiple HSR lines and lack the global perspective of analyzing a large region.

Methods

Data collection

In this paper, the time span of the data collection is from the years 2007 to 2019. China’s HSR has been open for operation since 2007. The data about the tourism industry after 2019 has been greatly affected by the COVID-19 epidemic. Therefore, in order to avoid the impact of the epidemic and other factors on the results, the study period is ultimately set from 2007 to 2019. It is noted that this period marks a crucial stage of rapid HSR development in China. The data from 2007 to 2019 can reflect the impact of HSR on tourism development under normal circumstances in China. Due to the paper’s length limitations, an overview of the study area is presented in the Appendix 1. The Appendix 1 provides a detailed introduction to the study area of this paper, which covers the seven major regions in China and 288 prefecture-level cities, as shown in Fig. A1. In addition, it outlines the selection criteria for the seven case cities.

The data used in this study mainly includes two parts. The first part covers tourism reception numbers, per capita GDP, population size, and tourism revenue. These data can be obtained from the China Statistical Yearbook, statistical yearbooks of various prefecture-level cities, and the China City Statistical Yearbook. The second part consists of the operation data of HSR. In other words, the current status and operational mileage of HSR are obtained from the official website of China’s transportation authorities.

TSI model

TSI model is one form of gravity model used to measure the spatial interaction between regions (Khadaroo & Seetanah, 2008; Yin et al., 2019; Yin et al., 2019; Liu et al., 2023). In this paper, TSI not only reflects the tourism development level of regions but also represents the strength of the tourism connection between them.

Although the gravity model has been applied to predict tourism demand since the 1960s, it has not received much attention from tourism-related research in the past few decades. Nevertheless, when the studies on tourism use the gravity model, the model parameters are often oversimplified, lacking reasonable estimation and comprehensive analysis of some key parameters, such as the spatial damping coefficient (Huang et al., 2019; Liu et al., 2023; Muñoz et al., 2023). In this paper, a revised gravity model is chosen to measure the impact of HSR on TSI based on Alan’s Wilson model, which has comprehensively considered the spatial damping coefficient under different study areas. The specific model is as follows:

$${T}_{{jk}}={A}_{k}{P}_{j}{C}_{j}^{\alpha }\exp (-\beta {r}_{{jk}})$$
(1)
$${r}_{{jk}}=({t}_{{am}}-{t}_{{sm}})+({t}_{{ah}}-{t}_{{sh}})\times 60+({t}_{{ad}}-{t}_{{sd}})\times 24\times 60$$
(2)

where \({T}_{{jk}}\) represents the TSI between city j and city k, \({A}_{k}\) represents the tourism attractiveness of destination k, tourism attractiveness refers to the factors that draw tourists to a destination, which cover a wide range of aspects, such as the economy, culture, tourism infrastructure, and natural landscapes. In this paper, tourism attractiveness measured by the number of tourist arrivals in destination k (not include inbound tourists). \({P}_{j}{C}_{j}^{\alpha }\) represents the travel ability of tourist destination cities, \({C}_{j}^{\alpha }\) represents the population size of city j, and \({P}_{j}\) represents the per capita income level of city j. Additionally, \(\alpha\) represents the income level parameter, \(\beta\) represents the spatial damping coefficient. \({r}_{{jk}}\) represents the distance between the tourist source j and the destination k, its unit is minutes. Based on the research by Li et al. (2012), \(\alpha\) is set to 0.64, and the value of \(\beta\) is the national average level of 0.00322 in this study.

In Eq. (2), \({t}_{{sm}}{,t}_{{sh}}\), \({t}_{{sd}}\) represent the minutes, hours and days in the departure time of the train from the station of city j, respectively. \({t}_{{am}}\), \({t}_{{ah}}\), \({t}_{{ad}}\) represent the minutes, hours and days in the arrival time of the train at the station of city k, respectively. If the destination city does not have operated HSR stations, travel time from other cities to it is calculated by using the travel time by conventional rail or road transportation. Otherwise, if the destination city has opened HSR stations in the first half of the year, the travel time is calculated by HSR in the same year of its opening. In this paper, the seven cities from the seven major regions are selected as the destination cities, where the HSR stations opening are listed in Table 1.

Table 1 HSR stations opening time of typical destinations.
Full size table

TSS model

Social network analysis (SNA) is a common method when investigating the change of TSS, which is originally used to describe the connection patterns and characteristics in social or interpersonal networks (Zhang et al., 2020). For example, in the field of tourism, researchers have tried to apply SNA methods to construct tourism attraction networks (Kang et al., 2018), tourism efficiency networks (Wang et al., 2014), and tourism ecological efficiency networks (Liu et al., 2022).

SNA primarily involves three elements: nodes, relations, and connections (Wang et al., 2014). According to the constituent elements of SNA, this study describes a network in which cities with HSR stations are described as nodes, and the connections indicate the presence of HSR lines between cities. The network is used to analyze the TSS in the cities’ tourism network. In order to explore the change of TSS, some adjustments have been made to the original formulas (Liu et al., 2022; Wang et al., 2022) of the degree centrality and betweenness centrality in SNA, listed as follows.

Degree centrality is used to measure the importance or centrality of nodes in a network. It quantifies the centrality of a node by the number of direct connections with other nodes. The original degree centrality (Liu et al., 2022), DC, is calculated as follows:

$${{DC}}_{i}=\frac{{K}_{i}}{N-1}$$
(3)

However, the above equation ignores the network flow of a node. Because people’s demand for tourism is the primary driving force behind tourism development and change (Wang et al., 2023). The number of tourists can reflect the development level of city tourism. Incorporating the number of tourists into the original formulas is crucial, as it allows for a more accurate reflection of the impact of HSR on TSS (Boto-García & Pérez, 2023). The adapted degree centrality model is represented by the following formula:

$${{TDC}}_{i}=\frac{{K}_{i}}{N-1}{f}_{i}$$
(4)

where \({{TDC}}_{i}\) represents the degree centrality of city i in the tourism network structure, \({K}_{i}\) represents the number of direct connections between city i and other cities, N is the total number of nodes in the network, \({f}_{i}\) represents the number of tourists received in city i.

The betweenness centrality usually reflects the importance of a node that plays the role of intermediaries or bridges in a network. It quantifies a node’s ability to act as a mediator in a network, such as serving as a critical bridge for the transfer of information or resources. Nodes with high betweenness centrality typically have a great impact on others since they can control the path of information or resource propagation. The original calculation formula for betweenness centrality, BC, is as follows (Wang et al., 2022):

$${{BC}}_{i}=\mathop{\sum}\limits_{s\ne i\ne t}\frac{{n}_{{st}}^{i}}{{g}_{{st}}}$$
(5)

In fact, the service function of a node within a network is ignored in the above equation. For example, improved traffic conditions can facilitate cities’ economic activities and form new centers in the tourism network structure (Wang et al., 2019). Furthermore, many works utilize the service frequency of HSR to measure its operating quality (Shao et al., 2017). Thus, the operating quality of HSR is considered to better investigate the impact of HSR on TSS in this study. The modified betweenness centrality model is defined by the following formula:

$${{TBC}}_{i}=\mathop{\sum}\limits_{s\ne i\ne t}\frac{{n}_{{st}}^{i}}{{g}_{{st}}}\frac{{m}_{i}}{{f}_{i}}$$
(6)

where \({{TBC}}_{i}\) represents the betweenness centrality of city i in the tourism network structure, \({g}_{{st}}\) is the total number of shortest paths between city s and city t. \({n}_{{st}}^{i}\) represents the number of shortest paths between city s and city t, passing through city i. \({m}_{i}\) stands for the frequency of the services of HSR that pass through city i, \({f}_{i}\) represents the number of tourists received in city i. When a city has a small number of tourists and a high frequency of the services of HSR, it indicates a strong betweenness centrality of the city i.

Measurement of city tourism function

Self-Organizing Map (SOM) as one of the most effective unsupervised classification algorithms, has been widely applied in various fields such as data mining, pattern recognition, and visualization (Zhou et al., 2022). Furthermore, the SOM clustering algorithm can effectively to preserve the topological structure among nodes within the HSR network. Therefore, this study used the SOM clustering algorithm to cluster cities according to two centrality indicators: degree centrality and betweenness centrality. The purpose of clustering is to identify whether a city plays a tourism transit or a tourism destination role. In this study, the SOM clustering algorithm procedure is illustrated as follows.

Input layer: give an input sample set \(x=({x}_{1},{x}_{2},\ldots ,{x}_{m})\), where m represents the number of cities, \(x\) represents the value of each city’s degree centrality (or betweenness centrality) in tourism network structure. In fact, the input layer includes a feature vector.

Competitive layer: use a 5×5 weight matrix as the output layer, in which each neuron includes a weight vector. To facilitate subsequent training, the initial values of the weight vectors for each neuron are randomly generated. Each neuron in the output layer is characterized by one vector \({w}_{{ij}}\) to characterize.

Step 1: For a given input sample, calculate its Euclidean distance with each neuron in the output layer. The specific neuron with the minimum Euclidean distance is the unique competition winner, which satisfies neuron \(y(x)\) as the best matching neuron according to Eq. (7).

$$y(x)={argmin}\Vert x-{w}_{ij}\Vert$$
(7)

Step 2: Update the weight vectors of the neighboring neurons to the winning neuron according to Eq. (8).

$${w}_{v}(c+1)={w}_{v}(c)+\varphi (u,v,c)\times \delta (c)\times \left[D\left(t\right)-{w}_{v}(c)\right]$$
(8)

where \({w}_{v}(c)\) is the current weight of the node v, c stands for the current iteration, \(\varphi\) represents the neighborhood function, which is a function of the distance between other neurons and the winning neuron, \(\delta\) represents the learning rate, D represents the input vector.

$${h}_{v}(c)=\delta (c)\times{\rm{exp}}\left(-\frac{\Vert {w}_{v}(c+1)-{w}_{v}{(c)}^{2}\Vert }{2\varphi (u,\,v,\,c)}\right)$$
(9)

where \({h}_{v}(c)\) represents the weight adjustment function, which plays a role in updating the connection weights between neurons.

The algorithm will terminate if the learning rate \(\delta\) ≤ \({\delta }_{\min }\) or if a predetermined number of iterations is reached.

Results

Evolution characteristics of TSI

Temporal evolution of TSI

In order to analyze the regional variations in the impact of the HSR on TSI and better display the obtained results, seven cities from each region are selected as the main representatives. The average TSI values from the various cities located in seven major regions to the seven representative cities are calculated over the past 13 years, illustrated in Fig. 1.

Fig. 1
Fig. 1
Full size image

Trends in average TSI between the representative cities and seven regions from 2007 to 2019.

From Fig. 1, it can be observed that the intensity of TSI is impacted by the distance, where the strength of TSI between regions that are farther apart tends to weaken. When Beijing is chosen as the destination, the average TSI between North China and Beijing is comparatively stronger compared to the other six regions. Conversely, the average TSI values corresponding to Beijing are few when the regions are the southwest and northwest areas far away from Beijing. The southwest and northwest regions correspond to low average TSI values due to the long distance from them to Shanghai. These findings provide evidence of the distance decay effect in the TSI model. Therefore, distance decay effects confirm transportation costs as a key factor for the development of a city’s tourism.

Additionally, the persistent TSI intensity differentials reveal structural imbalances in China’s tourism. Specifically, the regions of North China, East China, and Central China consistently correspond to higher TSI values than other regions, which shows a close tourism connection between these regions and the representative cities. In particular, the TSI between the eastern region and Shanghai is the strongest, with a TSI value reaching approximately 250. In contrast, the northwest and northeast regions display the weakest TSI corresponding to Nanning, with the highest value being less than 10.

The high-value “Eastern Triangle” reflects agglomeration benefits from developed transport networks and complementary tourism resources.

In addition to the distance factor affecting the TSI intensity, some specific events in the city at different times also have an impact on the TSI value. In Fig. 1, there is an obvious decreasing trend during many periods. For example, the decline in Shanghai’s tourism attractiveness to East China in 2011 is mainly because the Shanghai World Expo in 2010 attracts a large number of tourists, leading to a natural drop in the number of tourists in 2011. In addition, the rapid development of the tourism industry in other cities in East China, such as Hangzhou, Nanjing, and Suzhou, also has a diversion effect on Shanghai. For example, in 2011, Suzhou successfully established itself as a national demonstration city for tourism standardization, ranked first in the tourist satisfaction survey among 50 large and medium-sized tourist cities in China.

In 2015, the tourism industry in Shenyang faces a series of negative events. The results of the special rectification campaign on the tourism market order released by the National Tourism Administration show that some scenic spots in Shenyang, such as the Shenyang Botanical Garden, are seriously warned for failing to meet the relevant standards.

Similarly, in 2016, the tourism attractiveness of Chongqing also declined. This is also mainly because of a series of negative events. For example, the Dragon Gorge scenic area is stripped of its 5A-level scenic area title due to serious tourism safety hazards, poor environmental sanitation, and chaotic tourism order. This negative news directly affects its tourism attractiveness. Generally, these cases illustrate that special events and negative incidents can significantly impact a city’s tourism attractiveness.

In addition, Fig. 1 shows that although the opening times of the HSR in different cities are different, there is an overall upward trend in the TSI values. This can be attributed to the increase in population, per capita GDP, and the popularity of tourist attractions in these cities.

However, the most critical driver of this trend is the reduction in intercity travel time, a finding supported by our empirical analysis. Specifically, the fixed-effects model in the Appendix 2 confirms that travel time between cities has the most significant impact on TSI, as it directly mediates the flow of tourists and resources, as shown in Tables A1–A7.

As illustrated in Figs. 2 and 3, the development and optimization of HSR have effectively shortened intercity travel time. From Fig. 2, it can be seen that the number of cities connected by HSR increases from 37 in 2007 to 226 in 2019, demonstrates China’s successful infrastructure investment. In Fig. 3, the travel time from each region to the representative cities shows a downward trend overall. The decrease in travel time facilitates an increase in people’s willingness to travel and strengthens the level of TSI between cities.

Fig. 2
Fig. 2
Full size image

Number of cities with HSR from 2007 to 2019.

Fig. 3
Fig. 3
Full size image

Variations of average travel time from each region to the representative cities by HSR from 2007 to 2019.

Lastly, Fig. 1 shows that during the study period from 2007 to 2019, the growth of average TSI can be roughly divided into two stages. The first stage (from 2007 to 2014) is characterized by a gentle increase of the TSI, where the values from the southwest, northwest, and northeast regions to the representative cities are relatively low. The second stage (from 2014 to 2019) represents a phase of dramatic growth, where the value corresponding to the southwest region experiences more significant growth compared to the northwest region. The reason for the above difference in the two stages is mainly due to the different changes in average travel time from each region to the seven typical cities during the period. From Fig. 3, a noticeable decrease in the travel time can be found during the second stage. Especially during the period from 2014 to 2015, the decrease in travel time is most significant due to the notable increase in the number of cities connected by HSR.

In addition, the value corresponding to the southwest region is more significant than the northwest region. Because the development of HSR in the southwest region is more advanced than that in the northwest region. For example, it is evident that after 2009, the southwest region consistently shows a higher frequency of HSR services compared to the northwest, illustrated in Fig. 4. It reveals the differences in the average number of HSR services between the southwest and northwest regions.

Fig. 4
Fig. 4
Full size image

Variation of HSR services between the northwest and southwest regions.

Spatial differences in TSI

To indeed reveal the spatial differences, the change value and rate of the TSI between the seven representative cities and the cities located in different regions are calculated over the past 13 years as shown in Fig. 5.

Fig. 5
Fig. 5
Full size image

Spatial change in the TSI to each representative city.

From Fig. 5, it can be seen that the change value of the TSI is weak when Nanning and Shenyang are selected as the representative cities, its maximum value only reached about 290. The rate of change is also weak when considering Beijing as the representative city, with its maximum value only reaching 1740. When Chongqing is selected as the representative city, the change value is most significant, and the maximum change value can reach 2023.59. Nevertheless, most of the large changes correspond to the adjacent areas of the representative cities. For instance, the TSI between Beijing and Tianjin is higher than that between Beijing and other cities. Additionally, Fig. 5 shows that the northern, central, and eastern regions consistently exhibit deeper colors, which means that their value of TSI changes greatly owing to the HSR’s impact.

Furthermore, Fig. 6 illustrates the sum of the change values and rates of the TSI corresponding to the seven representative cities over the past 13 years. The areas with great change value are predominantly concentrated in the eastern region and the Beijing-Tianjin-Hebei region in the northern region, as well as the southeastern coastal region, as shown in the left part of Fig. 6. For example, Beijing, Tianjin, and Shanghai all have a relatively great change value of the TSI, corresponding to 4394.98, 2813.61, and 2560.84, respectively. This reveals that the HSR can exert a significant impact on the TSI change in these regions. Differently, the areas experiencing high growth rates are primarily concentrated in Southwest, South, and Central China. Specifically, cities like Kunming, Nanning, and Shenzhen all have significantly changed the rate of TSI, corresponding to 233.03, 254.52, and 298.1, respectively. In fact, the pronounced rate of the TSI change corresponding to these areas aligns closely with the Chengdu-Mianyang-Leshan HSR line and the Liupanshui-Zhanyi HSR line. This reveals that the HSR can cause a large change in these cities and regions along HSR lines.

Fig. 6
Fig. 6
Full size image

Totally spatial distribution of the TSI change value and rate.

Evolution characteristics of TSS

For investigating the TSS, the degree centrality and betweenness centrality of the seven representative cities within the tourism network structure during the study period are calculated, as shown in Fig. 7. From Fig. 7, it can be found that the centrality of the representative cities in the study period increases as a whole, except for the betweenness centrality of Beijing and Shanghai. This indicates that the tourism network structure tends to be stable and no longer depends on a single city. Initially, Beijing and Shanghai played hub roles in the tourism network. However, their roles have changed with the expansion and development of the HSRN. In other words, new hub cities are emerging in the tourism network structure, such as Wuhan and Chongqing. For example, the degree centrality of Chongqing and the betweenness centrality of Wuhan are larger than that of Beijing or Shanghai after 2017.

Fig. 7
Fig. 7
Full size image

Evolution of the TSS corresponding to the representative cities in degree centrality and betweenness centrality.

In addition, from Fig. 7, we can catch a glimpse of the development status of tourism in China. Specifically, the tourism center of China’s tourism development is no longer concentrated in North and East China. New tourism centers have successively emerged in Central, Southwest, and Northwest China, indicating that China’s tourism development is in a state of all-around prosperity.

Furthermore, the overall growth rates of the centrality of different cities within the tourism network structure are different during the period from 2007 to 2019, as presented in Table 2.

Table 2 Growth rate of degree centrality and betweenness centrality of the representative city.
Full size table

As shown in Table 2, the degree centrality of Chongqing displays the highest growth rate, followed by Xi’an, while that of Beijing has the lowest growth rate. Because Beijing, the capital of China, has a high priority to develop HSR and has also opened and operated HSR early, the impact of the HSR on the TSS of Beijing is limited, compared to other cities. However, Chongqing is late in developing HSR and does not open and operate HSR until 2015. Despite the relatively late development in HSR, the continuous improvement of the HSRN has significantly increased the connectivity between Chongqing and other regions. This leads to Chongqing becoming an important hub in the tourism structure. Differently, the betweenness centrality of Beijing and Shanghai shows negative growth while that of Shenyang has the biggest growth. This confirms that with the opening of HSR stations in other cities, the importance of Beijing and Shanghai has decreased and has even been overtaken by other cities. Moreover, these findings reveal significant variations in the impact of HSR on the TSS across different regions.

The overall evolution characteristics of the TSS may be attributed to the changing of accessibility between cities within the tourism network. Specifically, the average number of reachable cities increased from 6.11 in 2007 to 57.48 in 2019, and the maximum number of reachable cities has also shown an upward trend, rising from 21 in 2007 to 149 in 2019, as shown in Table 3. Therefore, the TSS gradually presents a multi-core development pattern. Interestingly, the minimum number of reachable cities consistently remains at 1 from 2007 to 2009. This phenomenon may be attributed to geographical constraints and the regional HSR construction plan. As a result, certain cities, like Haikou and Sanya, have only a single HSR line passing through them, and consequently, they have only one directly reachable city.

Table 3 Average accessibility between cities by HSR.
Full size table

Tourism role identification of different cities

Commonly, the role of a city can be divided into tourism destination function, tourism source function, and tourism transit function from the tourism activity perspective. Given the characteristics of the constructed indicators, they are only suitable for identifying and analyzing tourism destinations and transit points. Therefore, this study focuses on these two tourism-related functions and does not involve the identification of tourism source areas. Combined with the above result analysis of TSS, this study identifies the tourism role of different cities by using the SOM algorithm with degree centrality and betweenness centrality. The degree centrality of cities can be used to identify their role importance in the tourism destination function, while their betweenness centrality can be used to identify their role importance in the tourism transit function. It is noted that the paper classifies these cities into five levels. In this study, we refer to the number of classification levels from other research by combining with the economic classification results of Yicai to ensure that our classification results are scientifically grounded and can reflect the actual conditions of cities’ tourism development (Zhang et al., 2019; Liu et al., 2015). Additionally, 2007, 2011, 2015, and 2019 are selected as the time period of the clustering for clear visualization and interpretation, as shown in Fig. 8.

Fig. 8: Evolution of cities’ role in tourism destination function.
Fig. 8: Evolution of cities’ role in tourism destination function.
Full size image

The distinct colors stand for different cluster categories, and the size of the bubbles in the chart corresponds to the tier of cities, with larger bubbles indicating a higher tier of cities.

In Fig. 8, the cities can be clustered into five tiers based on their importance of role of tourism destination function. From Fig. 8, it can be seen that the number of cities in all five tiers exhibits an upward trend. The number of medium-tier cities increases from 9 in 2007 to 33 in 2019, and the number of high-tier cities grows from 1 in 2007 to 11 in 2019. Specifically, the number of medium-high tier cities is relatively small, while the number of low and lower tier cities far exceeds that of other tier cities. For instance, in the year 2015, there are a total of 14 cities classified as higher and high tier, while the number of low and lower tier cities amounted to 121.

Although the number of medium-high tier cities is relatively small, these cities have gradually strengthened their roles as tourism destinations with the development of HSRN. For example, the tourism destination capacity of cities such as Hangzhou and Zhengzhou became high-tier in 2011. Additionally, the tourism destination function of these cities undergoes continuous evolution over time. Specifically, Wuhan and Chongqing, initially positioned as medium-low tier cities, have gradually transitioned into higher tier cities and play significant roles in the tourism destination landscape. Interestingly, certain cities such as Beijing and Shanghai have consistently maintained their status as prominent tourist destinations. This can be attributed to the fact that these two cities have convenient transportation conditions and are also modern urban centers, offering tourists a diverse range of travel experiences.

Furthermore, it can be found that the presence of the HSR facilitates the development of the destination cities in China. In Table 4, the majority of core destination cities are located in North China, Central China, East China, and Southwest China.

Table 4 Spatial distribution of core destination cities.
Full size table

This distribution pattern is closely related to the method of selecting core destination cities, which we will elaborate on below. It is noted how to select the core destination cities. At first, we select cities from the top two levels of the SOM clustering results that have high degree centrality. Then, we further identify these selected cities according to provincial-level or sub-provincial cities.

The significant change in the spatial distribution of core destination cities is notable, which has expanded from solely encompassing North China to now including five distinct regions. In addition, the substantial impact of HSR is observed in East China, Southwest China, and Central China. Specifically, the number of core destination cities increases from 0 in 2007 to 5 in 2019 in East China, and from 0 in 2007 to 4 in 2019 in Southwest China.

In Fig. 9, the cities can also be classified into five tiers based on their importance of role of tourism transit function, in which the higher-tier cities are identified as the core transit cities due to their superior capacity in facilitating such transit compared to other tiers. The evolution of tourism transit cities can be observed in two distinct stages, shown in Fig. 9. In the initial stage, there is a noticeable increase in the number of cities across all tiers, from 2007 to 2015. In the subsequent stage, after 2015, a visible downward trend is observed, specifically among the medium-high tier cities. In other words, with the continuous development of the HSRN, a gap emerges between individual cities and other cities, resulting in the concentration of tourism transit capacity in specific cities. For example, Guangzhou, Changsha, and Zhengzhou are classified as core cities of tourism transit in 2019.

Fig. 9: Evolution of cities’ role in tourism transit function.
Fig. 9: Evolution of cities’ role in tourism transit function.
Full size image

The distinct colors stand for different cluster categories, and the size of the bubbles in the chart corresponds to the tier of cities, with larger bubbles indicating a higher tier of cities.

Similar to the evolution of tourism destinations, there are also only a few medium-high tier in tourism transit cities, whereas the low and lower tier cities make up a larger proportion of the total number of cities. For instance, in 2007, there are a mere 11 medium-high tier cities, while the number of low and lower tier cities reaches 26. Likewise, the number of medium-high tier cities is 23 in 2019, whereas an overwhelming 194 low and lower tier cities are identified. In addition, cities like Zhengzhou and Shijiazhuang have ascended to the first and second tiers from 2015 to 2019, indicating a consistent strengthening of their transit functions. This indicates that the tourism transit function of cities, such as Jinan and Nanjing, changes over time. Furthermore, some cities, such as Guangzhou, Changsha, Nanjing, and Jinan, are identified as consistently occupying the core positions among tourism transit cities.

Finally, it can also be observed that there are regional differences in the impact of the HSR on cities’ tourism function. In Table 5, the core and sub-core transit cities are mostly located in North China during the early period. This distribution pattern is closely related to the method of selecting hub cities. Therefore, it’s important to note how to identify hub cities. Initially, we choose from high-betweenness centrality cities in clustering results. Subsequently, we further select these chosen cities according to provincial-level and sub-provincial cities and the list of the hub cities released by the National Development and Reform Commission in China. For example, in 2007, 3 of the core and sub-core component cities are located in North China. However, there has been a shift in the distribution of these transit cities toward East and Central China over time. In 2011, 5 of the core and sub-core component cities are located in East China, while 6 and 4 of the core and sub-core component cities are located in Central China, in 2015 and 2019, respectively. This suggests that the impact of HSR development is significant in Central China and East China.

Table 5 Spatial distribution of core transit cities.
Full size table

Conclusion

This study investigates the spatio-temporal evolution of tourism interaction and structure under the impact of HSR in Mainland China during the period from 2007 to 2019, rather than that in a limited small region in a given year. To do that, the TSI and TSS models based on SNA and SOM clustering algorithms are proposed, and 288 cities are selected as a case study. The obtained results can be summarized as follows.

Firstly, the TSI between cities has shown an upward trend over the past 13 years, with two distinct stages including gradual growth and rapid expansion. Furthermore, the spatial evolution of TSI reveals that the impact of the HSRN improvement varies across different regions in China. Specifically, the impact is great significant on Central China, East China, and North China. This further reveals regional heterogeneities in HSR network impacts, supplementing prior studies that focused on a single area or HSR line (Wang et al., 2016).

Secondly, the evolution of the TSS in the representative cities, except for the betweenness centrality of Beijing and Shanghai, shows an upward trend from 2007 to 2019. For the degree centrality of cities in the tourism network structure, hub cities shift from Beijing and Shanghai to Wuhan and Chongqing, reflecting a decentralized trend in China’s TSS, an observation not fully captured in earlier research on multi-core patterns (He et al., 2023). In terms of betweenness centrality, Beijing and Shanghai show a downward trend.

Lastly, for the evolution of cities’ role in tourism destination function, the number of all five tiers of cities shows an upward trend from 2007 to 2009. Differently, for the evolution of cities’ role in the tourism transit function, the number of all five tiers increases on the whole before 2015, while the number of medium-high tier cities decreases after 2015. This phenomenon reflects a shift in the roles of HSR-driven cities, extending prior findings that HSR contributes to the emergence of new hot hubs by highlighting functional dynamics of cities (Li et al., 2023). Additionally, the role of cities in tourism functions undergoes constant change.

Management implications

Some management implications can be found from the above findings. These implications are tailored to different types of cities, considering the varying roles they play in the tourism network and the diverse nature of their tourism resources, to ensure more differentiated and diversified recommendations.

For core cities with well-developed HSR networks and abundant tourism resources. On the one hand, leverage their rising degree centrality to integrate regional tourism networks. For example, develop intercity HSR tourism alliances to coordinate route planning and share marketing resources, mitigating the “siphon effect” on peripheral areas. On the other hand, continuously improve the transportation infrastructure around HSR stations. This includes building more direct transportation links between HSR stations and major tourist attractions to reduce the transfer time for tourists. For instance, dedicated shuttle buses or subway lines.

For sub-core or peripheral cities, active cooperation with surrounding core cities should be pursued, such as the implementation of joint ticket sales or the provision of free admission to tourists arriving at the HSR stations of core cities, to enhance the appeal of sub-core or peripheral cities. Moreover, peripheral cities should leverage their distinctive resources, such as rural landscapes or ethnic cultural heritage, to enhance tourism competitiveness. Strategic collaboration with HSR-adjacent cities can facilitate “day-trip extension” itineraries, integrating these destinations into regional tourism networks.

Limitations

Nevertheless, this study has certain limitations. When exploring the evolution of urban tourism functions under the impact of HSR, we use the measurement of betweenness centrality and degree centrality by incorporating tourist arrivals and HSR service frequency. However, urban functions are inherently complex, our approach may not fully capture the multifaceted nature of urban tourism dynamics. Tourism-related infrastructure quality and service efficiency are crucial to tourism development, which directly influence tourism experiences. Future research could consider integrating additional dimensions, for example, incorporating online review sentiment analysis (e.g., from platforms like Ctrip) to quantify service quality. In addition, for research purposes, this paper selects a relatively long research period, specifically from 2007 to 2019. This leads to difficulties in obtaining data at the prefecture-level city and problems of missing data. Therefore, this study only uses the single indicator of the number of tourist arrivals to represent the tourism attractiveness of cities in the TSI model. However, relying solely on this single indicator may not comprehensively reflect a city’s tourism potential. In future, we can consider adjusting the research period according to research questions and the availability of data, selecting a time frame if related data can be more easily obtained.