Evaluation of Tourism Ecological Security and Its Driving Mechanism in the Yellow River Basin, China: Based on Open Systems Theory and DPSIR Model

Abstract

Tourism ecological security (TES) has gradually become a frontier topic because it is related to the virtuous circle of ecosystems and sustainable development, especially in river basins with fragile ecosystems. Based on the Driver–Pressure–State–Impact–Response (DPSIR) model and open systems theory, we constructed a TES evaluation system in the Yellow River Basin (YRB), China. Then, the TES index was measured from 2004 to 2019 and its spatio-temporal characteristics and driving mechanism were analyzed. The results show that: (1) In terms of temporal evolution, the comprehensive TES index shows a steady upward trend, but the difference between cities increases over time. Moreover, the proportion of cities with low status levels of TES declined rapidly, while the proportion of cities with high status levels of TES has grown slowly. (2) Spatially, low-TES value cities have always been in the majority, and the high-value cities show a scattered spatial distribution, most of which are along the river. Moreover, TES is randomly distributed in space before 2013, but it shows a significant positive spatial clustering feature thereafter. Specifically, the range of hot spots extends from the intersection of the middle and upper reaches to downstream, while the cold spots are always scattered. Furthermore, the trend surface in the east–west direction is always smooth, while it gradually manifests an inverted U-shape in the north–south direction. (3) In the dynamic transfer, TES lacks the vitality of transfer, but the probability of shifting upward becomes more significant when adjacent to higher-level cities; the opposite is true when adjacent to lower-ranked cities. (4) In terms of the driving mechanism, the factors related to tourism and the economy are the most important driving forces, and the effect of tourism-related factors on TES is becoming increasingly significant. Moreover, the driving mechanism is constructed. Finally, this study provides targeted policy implications for improving TES in the YRB, which has reference value for the development of ecological protection and high-quality tourism.

1. Introduction

Tourism has the characteristics of low resource consumption and fewer ecological hazards, and is considered a “green” and “clean industry” [1,2]. Some international organizations, such as the World Tourism Organization (UNWTO), have high expectations that tourism can contribute to achieving the sustainable development goals (SDGs). Specifically, in SDG15 (Life on Land), tourism is considered to play a positive role in protecting, restoring, and promoting the sustainable use of terrestrial ecosystems and halting biodiversity loss [3]. In addition, after studying different case sites, some scholars found that an appropriate tourism development has a positive effect on the urban ecological environment [4], green development efficiency [5], and residents’ awareness of environmental protection [6].

However, with the expansion of the global tourism market, the rapid growth of the tourism economy has also exerted tremendous pressure on the ecological safety of tourism destinations [7]. In some fragile ecosystem areas, the prevalence of over-tourism has brought a series of ecological and environmental problems [8], such as atmospheric pollution [9], soil erosion [10], water pollution [11], and marine litter pollution [12]. Venice, for example, the world’s most popular tourist destination, suffers from the serious adverse effects of over-tourism. In particular, the decline seen in the lagoon ecosystem service capacity due to excessive cruise tourism is the most alarming [13]. In recent years, the intense conflict between tourism and ecology has also led to protests by local civil society organizations, such as the “No Big Ships Committee” [14]. In the Middle East, coastal and marine tourism is widespread in the Gulf and Red Seas, which has degraded the quality of marine ecosystems. Specifically, the construction of coastal infrastructure and reclamation has led to the severe destruction of coral reefs, wetlands, and mangroves. Moreover, shipping, tourists’ uncivilized behaviors, and wastewater discharge have also led to marine ecological problems, such as marine snow, coral diseases, the overgrowth of macroalgae, and the loss of seagrass [15]. What is of more concern is that some niche destinations also experience conflicts between tourism development and the ecological environment. Benitez-Capistros et al. [16] confirmed that tourism activities on the Galapagos Islands triggered ecological degradation. Zhao and Guo [17] found that TES in China’s old revolutionary region of Dabie Mountain was generally low.

China is one of the fastest-growing countries in tourism, and its tourism revenues reached CNY 6.63 trillion, accounting for 11.05% of the GDP in 2019 [4]. However, China’s tourism industry has adopted an extensive growth mode, resulting in a severe threat to ecological security, especially in areas with fragile ecosystems, such as arid zones and large river basins. Among them, the Yellow River Basin (YRB), one of the most important ecological barriers in China, has received wide attention from scholars [18,19,20]. It is facing the crisis of ecosystem degradation due to unreasonable tourism development and over-tourism [21]. Specifically, in the upper reaches of the YRB, over-tourism has accelerated the disappearance of the Crescent Moon Spring in Dunhuang [22]. The level of tourism ecological security (TES) in the Qilian Mountains of the Zhangye section has also shown a significant declining trend in the short term [23]. In the midstream region, the quality of the fragile ecosystem of the Loess Plateau has declined due to tourism development [24]. The lower reaches of the YRB are one of the most densely populated areas in China, and the human–land conflict due to tourism development is increasing. Moreover, over-tourism can also further increase the general adverse effects of congestion diseconomies [25]. In general, irrational tourism development patterns have led to the degradation of destination ecosystems and have hindered sustainable tourism development in recent years [1]. In the face of these problems, the Chinese government has made ecotourism development a critical task in promoting ecological civilization construction since 2012, and it established ecotourism demonstration zones for the first time in 2013 to reverse the extensive tourism growth mode. Moreover, The National Development and Reform Commission of China (NDRC) released the “The National Ecotourism Development Plan (2016–2025)”, which provides directional guidance for China’s ecological tourism development [26]. Practical policy guidance is vital for improving tourism ecological security (TES). Meanwhile, high-level academic research is also essential because it can provide scientific theoretical and empirical evidence for policy implementation [17,23]. Therefore, the study of TES is urgent and meaningful.

TES has received wide attention from ecology, tourism, and geography scholars. Although existing studies have presented some scientific results, they have mainly focused on the spatial distribution characteristics of TES by using ArcGIS visualization technology. A summary of the spatial dynamic evolution characteristics based on long time series data is lacking, which may lead to a decrease in the accuracy of research conclusions and cannot be used to accurately predict future trends [2,27]. In addition, most scholars explored the influencing factors of TES from a static perspective without analyzing the changes in the influence level of each factor over time, which may lead to a time lag in the guidance provided by research findings to policy [28]. Furthermore, previous scholars mainly selected small-scale areas, such as national parks or macroscopic regions, as case sites. The former studies are of great practical significance to the case sites, but lack guidance relevant to other regions [1]. Although the latter studies help to provide a reference for constructing the national TES system, it is not easy to focus on the reality of specific regions to propose targeted practical values [23,28]. To improve the existing research, this study took the YRB as a case site, using long time series data from 2004 to 2019 to evaluate the TES from the perspective of spatial–temporal dynamic evolution and dissect the dynamic evolutionary characteristics of the influencing factors of TES.

The main contributions and innovations of this study are as follows: (1) In terms of research contents, different from the description of the spatial distribution characteristics of TES in previous studies, the spatial dynamic evolution law of TES was revealed by using dynamic spatial analysis methods, including spatial Markov chain and spatial trend surface. (2) Based on the conventional research on the influencing factors of TES, the changes in the influence level of each factor on TES were analyzed in critical years using the Geo-Detector, and were presented visually, which revealed the driving mechanism of TES more profoundly. (3) In addition, unlike previous micro- or macroscale studies, the Yellow River Basin, one of the four major civilization birthplaces, was selected as a case study site, which is also a medium-scale region where TES is most seriously threatened worldwide.

The remainder of this paper is organized as follows. Section 2 presents the literature review. Section 3 includes the materials and methods. Section 4 presents the results, including the temporal and spatial evolution characteristics of TES, the dynamic transfer characteristics of the TES, and the driving mechanism of TES. Section 5 is the discussion, and the conclusions are provided in Section 6.

2. Literature Review

2.1. Tourism Ecological Security

As ecological degradation is widespread worldwide, ecological safety has become an important research theme in ecology [29]. It is derived from the concept of environmental security, first introduced in the report “Our Common Future” adopted by the World Commission on Environment and Development (WCED) in 1987, and has received substantial attention from scholars [30]. In recent years, based on the research findings of ecological security, scholars have applied the concept to the study of specific ecological resources, such as forests [31], land [32], and water resources [33]. TES is derived from ecological safety and is essential to tourism sustainable development research [26]. The study of TES first started in the 1990s and originated from scholars’ concern about the positive impact of tourism development on ecological safety [34]. In this early stage, TES has not yet formed a mature evaluation system. Scholars have mainly focused on the interaction between tourism and the ecological, economic, and social systems of destination and sustainable tourism [35,36]. Recently, as the TES concept and evaluation system has gradually been clarified, scholars have conducted many empirical studies with different case sites [2]. After a comprehensive review of the existing studies, this research on TES includes three main aspects: (1) the connotations of TES; (2) the evaluation model and quantitative measurement of TES; and (3) the empirical study of TES.

The definition of TES has not been explicitly defined, but some of the fundamental connotations are generally accepted by scholars. On the one hand, TES means that tourism development does not cause a decline in the quality of the destination’s ecology and irreversible damage to the internal structure of the ecosystem; on the other hand, it refers to a harmonious state between tourism development and the ecological environment [21]. Among representative studies, Li et al. [37] consider TES to refer to a harmonious and orderly state of tourism development and an ecological–economic–social system based on rational tourism exploration and effective management of tourism activities. Wang et al. [23] argue that TES is a sustainable state of tourism development, natural resource use, and ecological protection. Yang et al. [38] assume that TES is a guarantee that the complex tourism ecosystem, which includes ecological, environmental, social, and tourism resources, is in a healthy, non-conflicting state. Liu and Yin [1] consider TES to be a state in which nature, the social environment, and economic development coexist harmoniously. Under this condition, the growth rate of tourism may be moderate, but the ecological environment would not be seriously threatened by tourism.

As the research on the connotation of TES tends to mature, the trend from “qualitative research” to “quantitative evaluation” is noticeable, and the evaluation models and quantitative measurement of TES have gradually become the study focus. Most TES evaluation models are derived from the evaluation framework of ecological safety, mainly involving touristic, ecological, economic, and social elements, but they do not yet have a unified standard. Among them, the Pressure–State–Response (PSR) model is widely used to evaluate the sustainability level of various ecosystems, which reflects the pressure caused by human activities on the ecosystem, the changes in the state of the ecosystem, and the human response to changes in the ecosystem [39]. Based on this model, Wang et al. [23] applied it to the evaluation of TES in the Qilian Mountains of the Zhangye section, China. Tang et al. [40] established the PSR evaluation index system of TES in Beijing. In addition, combined with the ecological–economic–social complex ecosystem theory, Li et al. [41] constructed the Pressure–State–Response–Society–Economy –Environment (PSR-SEE) model, which is an upgrade of the PSR model. The Driver–Pressure–State–Impact–Response (DPSIR) model, which is the most widely used model in the assessment of TES, was created by the European Environment Agency (EEA) in 1993 to address growing ecological problems [27]. It consists of five components: Driver (D), Pressure (P), State (S), Impact (I), and Response (R). Compared to the PSR model, it is more scientific and logical because it fully incorporates the core ideas of open systems theory [42]. Specifically, Ruan et al. [27] and Liu and Yin [1] constructed the DPSIR evaluation system of TES in the Yangtze River Delta and the provincial regions of China, respectively, and meticulously analyzed the mechanisms between the subsystems. The Driver–Pressure–State–Impact–Response-Economy–Environment–Society (DPSIR-EES) model is an upgraded version of the DPSIR model that fully combines the advantages of the logical rigor of the DPSIR model and the comprehensiveness of the Economy–Environment–Society (EES) model [43]. However, the high demand for data acquisition is a major obstacle to its widespread use. Furthermore, scholars have also used the Threat–Quality–Regulation (TQR) model [44], Carrying–Supporting–Attraction–Evolution–Developing (CSAED) model [45], and Pressure–State–Control (PSC) model [46] to construct the evaluation system of TES. The quantitative measures of TES can be divided into two types. On the one hand, from the development level perspective, scholars used the polygon composite indicator method [1], the improved Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) method [40], the entropy weight TOPSIS method [17], the ecological footprint method [47], Structural Equation Modeling (SEM) [48], the catastrophe progression method, and the Analytic Hierarchy Process (AHP) and Delphi methods [49] to measure the TES level. On the other hand, from an efficiency perspective, the Data Envelopment Analysis (DEA) method is used to measure the quality of TES [2,27]. In contrast, the former is the most popular TES quantitative measurement method, while the latter is still in the exploration stage.

In the empirical study of TES, scholars have conducted extensive research in different scale cases based on constructing the evaluation system of TES and quantitatively measuring its development level. At the microscale, scholars mostly took cities and ecologically fragile areas as objects. Tang et al. [40] studied the TES’s spatial and temporal dynamics in ecological conservation development areas in Beijing, concluding that the TES is improving but still has potential for improvement. Liu et al. [49] assessed the TES of the Kanas Natural Heritage Site in Xinjiang, China, and found that its TES status was good. However, obstacles such as forest cover and the natural population growth rate prevent it from continuing to progress. Wang et al. [20] found that the TES levels in the Zhangye Qilian Mountains National Park showed a “U-shaped” trend. At the medium and macro scales, Ruan et al. [27] identified that TES levels in the Yangtze River Delta showed an “N” pattern of “rising–declining–rising” and had significant urban heterogeneity. Liu and Yin [1] studied the TES in China and found that it showed a steady increasing trend, and the distribution of cold and hot spots shifted from an east–west to a north–south direction. In terms of research content, existing studies mainly involved the spatial–temporal characteristics [1,27], barrier factor [23], dynamic simulation [50], and grey correlation degree of TES [23]. In addition, studying the influences of TES by using Geo-Detector [32] and spatial econometric models is also an essential part of the field [2].

2.2. Current Gaps in the Research

As an essential theme in tourism ecology, TES has attracted a wealth of research findings, which have laid a good foundation for further study. However, after a systematic review of the literature, it was found that the current research gaps in TES were addressed in this study.

First, in terms of the TES evaluation models, most previous scholars directly applied existing models, excluding detailed explanations of their theoretical basis. Specifically, the DPSIR model, the most widely used in TES evaluation, is supported by open systems theory and has a complex operational logic [27]. However, the previous studies selected evaluation indicators based on five dimensions (D-P-S-I-R), without an in-depth analysis of the interaction among the subsystems. Therefore, in this study, based on the interpretation of the theoretical sources of the DPSIR model, the operation mode among the subsystems was explained, the basis for selecting each indicator was clarified, and the procedure of using the DPSIR model in TES research was standardized.

Second, in terms of the study duration, the time series used in the existing studies was relatively short, with most not exceeding ten years due to the severe lack of early data [2,27]. Studies with short time series do not fully reveal the dynamic evolution of TES and provide scientific predictions [1]. To fill this gap, time series data with a length of 16 years were used to comprehensively compare the dynamic change pattern of TES in the YRB before and after the adoption of the ecological civilization construction strategy in China.

Third, in the selection of study case sites, previous scholars preferred microscale cases, such as cities and national parks, or macroscopic regions. The former is more practical but lacks the analysis of spatial heterogeneity; the latter focuses on the spatial characteristics of TES in different regions, but it struggled to provide targeted improvement measures for each part. In contrast, medium-scale regions, such as urban agglomerations and watersheds, are better options because they combine the advantages of both types above. Therefore, the Yellow River Basin was selected as the study area, and its spatial and temporal dynamic evolution characteristics and driving mechanism of TES were analyzed in depth.

Fourth, in terms of research content, scholars have paid more attention to TES’s spatial distribution and agglomeration characteristics rather than its dynamic evolutionary characteristics and the trend of hierarchical change. Therefore, to fill the gap in existing studies, the spatial evolutionary characteristics of TES were analyzed using the spatial trend surface. In addition, based on the Markov chain, the transfer pattern of the hierarchy of TES in the YRB was also revealed, which provided a new perspective for the study of TES.

Finally, previous scholars constructed econometric models, such as the spatial Durbin model and the panel regression model, to analyze the driving mechanism of TES. Even if a few scholars used Geo-Detector to compensate for the shortcomings of economic models, they ignored the variability in the influence level of driving factors. Therefore, to advance the research on driving mechanisms, the impact of each factor was measured in critical years and their changing trends were analyzed using a Geo-Detector, which helps provide policymakers with more immediate policy insights.

3. Materials and Methods

3.1. Study Area

The Yellow River, China’s “Mother River”, is the sixth longest river in the world, with a total length of 5464 km. It is located at 32~42° N, 95~120° E, with a watershed area of 795,000 km², which originates from the Bayankara Mountain Range on the Qinghai–Tibet Plateau and injects into the Bohai Sea from Kenli County in Shandong Province. It flows through nine provinces in China, including Qinghai, Sichuan, Gansu, Inner Mongolia, Ningxia, Shanxi, Shaanxi, Henan, and Shandong. The scope of the Yellow River Basin (YRB) delineated in the paper is referred to in the document issued by the Yellow River Water Conservancy Commission of the Ministry of Water Resources (YRCC), PRC [51]. In addition, the document titled “Outline of the Yellow River Basin’s Ecological Protection and High-quality Development Plan” issued by the State Council of China in 2021 is also an important reference for this study [52]. However, due to the severe lack of data in 11 cities (Alxa League, JiCNY, Linxia Hui Autonomous Prefecture, Gannan Tibetan Autonomous Prefecture, Haidong, Haibei Tibetan Autonomous Prefecture, Huangnan Tibetan Autonomous Prefecture, Hainan Tibetan Autonomous Prefecture, Guoluo Tibetan Autonomous Prefecture, Yushu, and Haixi Mongolian and Tibetan Autonomous Prefecture), only 78 prefecture-level cities in the basin were selected as the case study sites (Figure 1). The reasons for taking the YRB as a case study site of TES are as follows:

(1)

The YRB plays a vital role in China’s economic and social development. In 2021, 421 million people lived in the basin, accounting for 29.77% of China’s population. Among them, the populations of Zhengzhou and Xi’an exceed 10 million (respectively, 10.35 million and 10.20 million), while the populations of Wuhai City and Shizuishan City are only 566,100 and 805,900, respectively. Moreover, the GDP is approximately CNY 28.68 trillion in the YRB, accounting for 25.05% of China’s total in 2021. Among them, the GDPs of Zhengzhou, Jinan, and Xi’an rank among the top three, respectively, CNY 1269.10 billion, CNY 1143.22 billion, and CNY 1068.83 billion. In contrast, the GDPs of Zhongwei, Baiyin, and Shizuishan are only CNY 50.47 billion, 57.1 billion, and 61.7 billion, respectively. The per capita GDP of the YRB is CNY 68,489, which is 84.58% of the Chinese average. Among them, the per capita GDPs of Ordos, Dongying, and Yulin rank in the top three, at CNY 172,686, 134,022, and 120,908, respectively, while the per capita GDPs of Dingxi, Longnan, and Tianshui are only CNY 19,873, 20,938, and 25,178, respectively;

(2)

The YRB is a critical ecological barrier, which has diverse ecosystems, including desert ecosystems, such as the Mawwusu Desert and Loess Plateau, grassland ecosystems, such as the Ordos Grassland and Aba Yellow River Steppe, wetland ecosystems, such as the Loop Plain Wetlands and Estuary Delta Wetlands, and forest, farmland and urban ecosystems. However, the rapidly developing economy and society have gradually created violent conflicts with the ecological environment. Specifically, the wetlands have been destroyed in upstream areas, and extensive ecological problems, such as land desertification, soil erosion, and salinization, have developed in the middle and lower reaches;

(3)

Ecological safety in the YRB is a vital concern for the Chinese government. In 2021, China’s State Council released an outline document on the ecological protection and high-quality development of the Yellow River Basin, elevating ecological conservation and the high-quality development of the YRB to a national strategy. It aims to strengthen the water conservation capacity in the upstream areas, the soil and water conservation capacity in the midstream areas, and the resilience of wetland ecosystems in the downstream areas by implementing a series of natural restoration and ecological protection projects;

(4)

The YRB is an important tourist destination in China and has rich tourism resources, including natural tourism resources, such as Hukou Waterfall and Mount Hua, and humanistic tourism resources, such as the Terracotta Warriors and the Longmen Grottoes. It has 20 World Heritage and 84 AAAAA tourist attractions, accounting for 35.71% and 26.42% of the total number in China, respectively, laying a solid foundation for developing tourism in the YRB. In 2019, the tourism revenue of the nine provinces in the YRB reached CNY 3.57 trillion, accounting for 53.85% of China’s total, which showed that it occupied an important position in China’s tourism industry. However, with the rapid development of tourism, unreasonable tourism exploration and over-tourism have led to the increasingly prominent problem of TES.

In summary, the YRB is representative of a case study site for TES, and this research can provide a reference for studying TES in other large river basins worldwide.

3.2. Data Sources

Due to the lack of data before 2004, some of which have not been updated since 2019, this study collected statistics for 78 prefecture-level and above cities in the YRB from 2004 to 2019. The data were obtained from the China Statistical Yearbook (2005–2020), China Urban Statistical Yearbook (2005–2020), China Environmental Statistical Yearbook (2005–2020), China Tourism Statistical Yearbook (2005–2020), and the statistical bulletin of each city. In addition, the data on the annual average concentration of inhalable particles were obtained through the China Environmental Energy Economics Database (https://www.epsnet.com.cn/index.html#/Index). The missing data were filled in by linear interpolation. The administrative division, river, and digital elevation (DEM) data were obtained from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences (http://www.resdc.cn/).

3.3. Methods

3.3.1. The Tourism Ecological Security System

TES is a derivative of the concept of ecological safety, and it is an important research topic in tourism ecology for which there is no uniform definition yet. Currently, most scholars emphasize that TES refers to the ecosystem of tourist destinations that do not experience a decline in quality or even irreversible damage due to tourism activities [1,2]. In addition, it is a harmonious and orderly state between tourism and social, economic, and ecological systems [53,54]. Based on the above analysis of TES and existing studies, combined with open systems theory, the DPSIR model was embedded into the study of TES and the TES conceptual model was constructed (Figure 2).

The open systems theory was proposed by Bertalanffy in 1932, who believed that any system is an organic whole with complex functions composed of several subsystems in a specific structure and form [55]. Moreover, the whole function is not reducible to the isolated state of each subsystem. Currently, it is widely used in ecological research and fully integrated into other theories. Among them, the social–economic–natural complex ecosystem theory suggests that evaluating the regional ecosystem requires a holistic view that considers the interactions among the elements of social, economic, and natural systems [56,57]. The DPSIR model was proposed by the EEA in the 1990s and is mainly used to address resource and environmental issues, combining the advantages of the PSR model and the DSR model [27]. It is widely used in the study of ecological security, such as water ecological security, land ecological security, and forest ecological security. Specifically, it consists of five components: Driver (D), Pressure (P), State (S), Impact (I), and Response (R).

In the TES system, the mechanism of each part is as follows. Tourism’s social and economic development are the drivers of the TES system. However, under the extensive mode of growth, tourism and the ecosystems of tourism destinations are under tremendous pressure, and the stable state is under attack. In this context, the sustainable development of tourism is strongly negatively affected. Finally, tourism stakeholders respond positively and take targeted measures to mitigate conflicts between tourism development and the ecological environment. In summary, D-P-S-I-R systems are interlinked, and TES can be improved only when the five systems are coordinated and orderly. According to existing studies and expert advice, this study selected 27 indicators to construct the evaluation system of TES based on the available data (

Table 1. Evaluation index system of tourism ecological security.

Standard Level	Factor Level	Index Level	No.	Weight	Index Significance	References
Driver (D)	Tourism development	Growth rate of tourism revenue (%)	D1	0.023	To reflect the potential damage caused by tourism development to the ecological environment.	[1,2,27,28]
		Growth rate of tourists (%)	D2	0.028
	Social development	Urbanization rate (%)	D3	0.024	To reflect the potential damage caused by urbanization and population growth to the ecological environment of tourist destinations.
		Natural growth rate of population (%)	D4	0.078
	Economic development	GDP per capita (CNY)	D5	0.025	To reflect the potential damage caused by economic development to the ecological environment of tourist destinations.
		GDP growth rate (%)	D6	0.019
Pressure (P)	Tourism and social pressure	Tourism spatial index (Person/km2)	P1	0.033	To reflect the pressure of tourists and residents on the tourist destinations, respectively.	[1,17,37,41]
		Population density (Person/km2)	P2	0.101
		Tourist density index (%)	P3	0.013	To reflect the degree of tourists’ interference in local residents’ life through the ratio of the tourists’ number to the total number of permanent residents.
	Ecological pressure	Industrial wastewater discharge (tons)	P4	0.022	To reflect the pressure of pollutant discharge on the ecological environment.
		SO2 emission (tons)	P5	0.033
		Solid waste output (tons)	P6	0.026
State (S)	Tourism economy	Domestic tourism income (million CNY)	S1	0.032	To reflect the changes of tourism economy state in the process of system operation.	[1,2,27,58]
		Tourism foreign exchange income (million CNY)	S2	0.031
		Per capita tourism income (CNY)	S3	0.029
	Ecological environment	Green coverage rate of built-up region (%)	S4	0.011	To reflect the changes of ecological environment state in the process of system operation.
		Annual average concentration of inhalable particles (mcg/m3)	S5	0.064
Impact (I)	Industrial economy	Proportion of total tourism revenue in GDP (%)	I1	0.060	To reflect the impact of the system operation on the industrial economy.	[1,20,28,37]
		Proportion of tertiary industry (%)	I2	0.017
		Tourism economic density (CNY10,000/km2)	I3	0.012
		Tourism industry cluster (%)	I4	0.047
	Tourism employment	Employees in accommodation and catering industry (people)	I5	0.081	To reflect the impact of the system operation on tourism employment.
Response (R)	Talent supply	Number of students in ordinary colleges and universities (people)	R1	0.103	To reflect the talent supply response required to optimize the system.	[1,2,27]
	Economic investment	Proportion of fiscal expenditure in GDP (%)	R2	0.058	To reflect the economic investment response required to optimize the system.
	Environmental governance	Sewage treatment rate (%)	R3	0.006	To reflect the environmental governance response to optimize the system.
		Domestic waste treatment rate (%)	R4	0.012
		Comprehensive utilization rate of solid waste (%)	R5	0.012

).

(1)

Driver (D). The driving factor includes three parts—tourism, and social and economic development. Specifically, the growth rate of tourism revenue (D1) and the growth rate of tourists (D2) are used to measure the tourism development level. The urbanization rate (D3) and the natural growth rate of the population (D4) are used to measure the social development level. Finally, the GDP per capita (D5) and the GDP growth rate (D6) are used to measure the economic development level;

(2)

Pressure (P). The pressure factor is explained in terms of tourism, social, environmental, and ecological pressure. The tourism spatial index (P1), population density (P2), and tourism density index (P3) are used to assess tourism and social pressure. Moreover, industrial wastewater discharge (P4), SO₂ emissions (P5), and solid waste output (P6) are used to comprehensively assess ecological pressure.

(3)

State (S). The state factor is evaluated based on the tourism economy and ecological environment. Domestic tourism income (S1), tourism foreign exchange income (S2), and per capita tourism income (S3) are used to assess the tourism economy. Moreover, the green coverage rate of the built-up region (S4) and the annual average concentration of inhalable particles (S5) are used to evaluate the ecological environment.

(4)

Impact (I). The impact factor is evaluated based on the industrial economy and tourism employment. The evaluation indicators of the industrial economy include the proportion of total tourism revenue in GDP (I1), the proportion of tertiary industry (I2), the tourism economic density (I3), and the tourism industry cluster (I4). In addition, since tourism is a comprehensive industry and tourism employment cannot be directly counted, employees in the accommodation and catering industry (I5) were selected to characterize the tourism employment level.

(5)

Response (R). The response factor is evaluated in three parts: talent supply, economic investment, and environmental governance. Specifically, the number of students in ordinary colleges and universities (R1) represents talent supply. The proportion of fiscal expenditure in GDP (R2) represents an economic investment. Moreover, the sewage treatment rate (R3), domestic waste treatment rate (R4), and comprehensive utilization rate of solid waste (R5) are used to comprehensively assess ecological pressure.

3.3.2. The Entropy-Weighted TOPSIS Method

The method combines the entropy weight and the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) [59]. First, based on data standardization, the entropy weight method is used to assign index weights based on the discrete degree of data, which can effectively reduce subjectivity and improve the credibility of indicator weights. Then, the TOPSIS method is used to calculate the Euclidean distance between the index values of each evaluation object and the optimal and inferior values to derive the comprehensive evaluation index.

The first step is to normalize the data. To eliminate the influence of different dimensions of each indicator on the evaluation results, the range method was used to standardize the original data. Formulas (1) and (2) correspond to the calculation process of the positive and negative indices, respectively.

Positive indicator:

$$X_{ij}=\frac{x_{ij}-\min x_{ij}}{\max x_{ij}-\min x_{ij}}$$

(1)

Negative indicator:

$$X_{ij}=\frac{\max x_{ij}-x_{ij}}{\max x_{ij}-\min x_{ij}}$$

(2)

where X_ij denotes the normalized value of indicator j of city i and x_ij denotes the original value of indicator j of city i.

The second step is to normalize the data.

$$P_{ij}=Y_{ij}/{\displaystyle \sum _{i=1}^m{Y}_{ij}}$$

(3)

The third step is to calculate the information entropy E_j of indicator j in year t.

$$E_j=-\frac{1}{\ln m}{\displaystyle \sum _{t=1}^m{P}_{ij}}\ln P_{ij}$$

(4)

The fourth step is to calculate the weight W_j of each indicator.

$$W_j=\frac{1-E_j}{{\displaystyle \sum _{j=1}^n(1-E_j)}}$$

(5)

The fifth step is to calculate the weighting matrix.

$$\begin{array}{l}R={(r_{ij})}_{mn},r_{ij}=w_j{x}_{ij}\\ (i=1,2,\cdots ,m;j=1,2,\cdots ,n)\end{array}$$

(6)

The sixth step is to determine the optimal and inferior solutions.

$$S_j^{+}=\max (r_{1j},r_{2j},\cdots ,r_{nj}),S_j^{-}=\min (r_{1j},r_{2j},\cdots ,r_{nj})$$

(7)

The seventh step is to calculate the Euclidean distance between the various solutions and the optimal and inferior solutions.

$$se{p}_i^{+}\sqrt{{\displaystyle \sum _{j=1}^n{(s_j^{+}-r_{ij})}^2}},se{p}_i^{-}\sqrt{{\displaystyle \sum _{j=1}^n{(s_j^{-}-r_{ij})}^2}}$$

(8)

Finally, the comprehensive evaluation index of TES is calculated.

$$C_i=\frac{se{p}_i^{-}}{se{p}_i^{+}+se{p}_i^{-}}$$

(9)

The classification of tourism ecological security has not yet formed a unified standard. Referring to the existing classification standards [17,21,28,40], and according to the measurement results of TES, the TES of the YRB were divided into I (Extreme Insecurity), II (Insecurity), III (Relative Insecurity), IV (Critical Insecurity), V (Critical Security), VI (Relative Security), VII (Security), and VIII (Extreme Security) (Figure 3).

3.3.3. Spatial Autocorrelation

Tobler’s First Law of Geography states that everything is related to everything else, but spatially close things are more related [60]. Therefore, this paper uses the global Moran’s I to measure the global spatial autocorrelation of TES.

$$I=\frac{n{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{w}_{ij}(Y_i-\overline{Y})(Y_j-\overline{Y})}}}{S^2{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{w}_{ij}}}}(i\ne j)$$

(10)

where i and j denote different cities, $S^2=\frac{1}{n}{\displaystyle \sum _{i=1}^n(Y_i}-\overline{Y}{)}^2$ and $\overline{Y}=\frac{1}{n}{\displaystyle \sum _{n=1}^n{Y}_i}$, Y_i and Y_j represent the observations on the spatial cell, w_ij is the spatial weight matrix, and I takes values between −1 and 1. The spatial correlation is negative when the value of I is less than 0, spatially uncorrelated when it is equal to 0, and spatially positive when it is greater than 0. The larger the absolute value of I is, the stronger the spatial correlation.

In addition, since global spatial autocorrelation cannot present the local spatial correlation characteristics of TES, Getis-Ord Gi* was used to analyze TES’s cold spots and hot spots.

$${G_i}^{*}(d)={\displaystyle \sum _{j=1}^n{w}_{ij}(d)}{P}_j/{\displaystyle \sum _{j=1}^n{P}_j}$$

(11)

where n denotes the number of cities, d represents the geographic distance between cities, P_j represents the observed values of TES in spatial units, w_ij is the spatial weight matrix, and Gi* is close to 0, which means that the observed values are randomly distributed in the region. The larger the absolute value of Gi*, the more likely it is to form a hot spot area or a cold spot area.

3.3.4. Spatial Trend Surface Analysis

The trend surface is used to analyze TES’s spatial distribution pattern and evolutionary trends. The trend surface model was constructed assuming that R_i is the TES value of a city in the YRB and (X_i, Y_i) is the city’s coordinates.

$$R_i(X_i,Y_i)=T_i(X_i,Y_i)+{\epsilon }_i$$

(12)

where R_i(X_i, Y_i) is the trend function, T_i(X_i, Y_i) is the trend surface fit value, and ε_i is the autocorrelation random error term. In this paper, a second-order polynomial is used to measure the trend value of TES.

3.3.5. Markov Chain

The Markov chain is a classic prediction model often used to study the random transfer probability of a social phenomenon. It was used to analyze the probability of the evolution of the TES types over time. The research steps are as follows: First, the continuous attribute values of TES at different years are discretized to determine the state space and parameter set of the Markov prediction model. Subsequently, TES is classified into k types, and the probability distribution and changes of each type are calculated to construct the state shift probability matrix M. Supposing the probability of a city that is type i in year t changing to type j in year t + 1 is P_ij, the calculation formula is:

$$P_{ij}=\frac{n_{ij}}{n_i}$$

(13)

where n_ij denotes the number of cities that are type i in year t and changed to type j in year t + 1 during the study period, and n_i denotes the number of cities of type i.

The spatial Markov chain is a combination of the Markov chain and spatial lag term, which makes up for the traditional Markov chain’s deficiency in examining spatial factors [61]. By calculating the probability of an upward or downward shift of TES, the relationship between the probability of a shift of TES types and the TES types in neighboring cities is revealed. Specifically, based on TES being discretized into k types, the traditional N × N order Markov matrix is decomposed into k (N × N) order transfer probability matrix, which can clearly express the possibility of an upward or downward shift of TES types in different neighborhood contexts. Taking the kth condition matrix as an example, P_ij(k) denotes the probability of a city moving from type i in year t to type j in year t + 1 when conditioned on the spatial lag type k. The spatial lag type (k) is divided by the spatial lag value of TES in the initial year. The formula is as follows.

$$P_{ij,t+1}(k)=P_{ij,t}(k)N{W}_k$$

(14)

where N is the N × N transfer matrix, W_k is the spatial lag term type, and P_ij, t + 1 is the probability matrix of type k in year t + 1, with k as the spatial lag condition.

3.3.6. Geo-Detector

Geo-Detector is a statistical method used to detect the influencing factors of the spatial differentiation of geographic phenomena, and it is widely used in the study of geography and ecology. This paper applies it to identify the impact of each factor on TES. The formula is as follows.

$$q=1-\frac{{\displaystyle \sum _{h=1}^L{N}_h{\sigma }_h^2}}{N{\sigma }^2}$$

(15)

where q is the explanation degree of TES offered by Factor X. The value range is [0, 1], and the closer q is to 1, the stronger the explanatory power of X on TES. L is the stratum of X or TES. N_h and σ_h² are the number of units and variances of stratum h, respectively. Moreover, N_h and σ_h² are the number of units and variances of the whole area, respectively.

3.4. Research Framework

This paper proposes a research framework to assess the spatial–temporal dynamic evolution and driving mechanism of TES in 78 cities in the YRB. First, we collect relevant data and build an evaluation system for TES based on the DPSIR model. Second, the entropy-weighted TOPSIS method is used to measure TES level. The spatiotemporal dynamic evolution characteristics of TES are explored by combining spatial autocorrelation, spatial trend surface, and Markov chain. Moreover, its driving mechanism is analyzed by Geo-Detector. Finally, the study proposes relevant policy implications based on the results (Figure 4).

4. Results

4.1. Temporal Evolution Characteristics of Tourism Ecological Security

According to Formulas (1)–(9), the entropy-weighted TOPSIS method was used to measure the TES value of 78 cities in the Yellow River basin from 2004 to 2019 (Appendix A), and its temporal evolution characteristics were analyzed based on the results. Figure 5a shows that the average value of TES in the YRB has a steady upward trend, from 0.084 in 2004 to 0.192 in 2019, with an average annual growth rate of 8.5%. However, it is still deficient, indicating that the YRB faces a high ecological safety risk. Moreover, the “boxes’” heights have increased continuously, indicating that the differences in the TES value have gradually widened. The scatter plots show that the levels of TES always manifest in a “pyramid” structure, indicating that most cities have low TES values, and only a few have broken out of “low-level equilibrium”. Furthermore, Figure 5b shows that the curve moves to the right, the peak body gradually widens, and the peak height gradually decreases, reflecting the phenomenon of “club convergence” of TES, which further confirms the conclusion that the difference in TES gradually increases. In the “tail” of the nuclear density curve, small waves are always present and tend towards higher values, indicating that the few cities with higher values of TES are in an excellent position to develop.

In addition, according to the classification standard of the TES level, the TESs of 78 cities were graded, and the changing trend of the number of each level was analyzed. As shown in Figure 6, there were six categories of TES in the Yellow River Basin during 2004–2019, which included I (Extreme Insecurity), II (Insecurity), III (Relative Insecurity), IV (Critical Insecurity), V (Critical Security), and VI (Relative Security). In general, the number of I decreases significantly, the number of II shows a trend of increasing first and then decreasing, the number of III gradually increases, and the cities in categories IV, V, and VI have grown out of nothing but account for a small proportion. Specifically, the proportion of I dropped from 86% to 14%, indicating that the improvement in TES in the YRB achieved initial results. The continuous deterioration of the ecological environment in most cities has been reversing since 2004, when the Ministry of Ecology and Environment of the PRC issued the “Construction Indicators for Ecological Counties, Ecological Cities and Ecological Provinces”. The number of IIs showed a continuous upward trend from 2004 to 2013, especially in 2013, and the proportion increased rapidly from 41% in the previous year to 79%. This is because the Chinese government made the strategic decision of ecological civilization construction at the end of 2012, which led urban governments to adopt strict environmental regulations under the pressure of political assessment, resulting in a rapid improvement in the ecological environment in a short time [62]. Moreover, in the same year, the China National Tourism Administration (CNTA) established seven national-level ecotourism demonstration areas in the YRB, including the Yuntai Mountain Demonstration Zone in Jiaozuo, the Yao Mountain–Giant Buddha Demonstration Area in Pingdingshan, the World Expo Park Demonstration Area in Xi’an, the Dangzhou Grassland Demonstration Area in Gannan, the Xinglong Mountain Demonstration Area in Lanzhou, and the Shapotou Demonstration Area is in Zhongwei. They play a key role in the improvement of TES in the YRB. However, within China’s politically competitive system, the effectiveness of policies has decreased over time [63], and many IIs reverted to category I by 2014. From 2014 to 2019, the proportion of IIs showed a fluctuating decline due to many cities upgrading to higher tiers. The share of IIIs gradually increased from 4% in 2004 to 24%, and they ranked second in 2009. Moreover, since Jinan, Zhengzhou, TaiCNY, and Xi’an became stage IV cities in 2014, the proportion of stage IV cities gradually increased to 9% in 2019. Xi’an was promoted to V in 2017, and Zhengzhou and TaiCNY were also promoted from IV to V the following year. In 2018, Xi’an became the only city in VI, which shows that Xi’an has achieved great success in improving TES.

4.2. Spatial Evolution Characteristics of Tourism Ecological Security

4.2.1. Spatial Distribution Pattern

On the spatial scale, to clarify the spatial distribution characteristics of TES in the YRB, the years 2004, 2007, 2010, 2013, 2016, and 2019 were selected and visualized using ArcGIS 10.8. As shown in Figure 7, the overall situation of TES in the YRB is poor, with the high-value cities showing a “scattered” spatial distribution, most of which are along the river. However, the development trend is positive during the study period. Specifically, the TES status of most cities from 2004 to 2010 was extremely insecure, and only the cities located in the Tao River Basin, Hetao Plain, and the Yellow River Delta had a relatively good safety status. On the one hand, the ecological environment of the YRB is fragile, with the SanjiangCNY Reserve and Qilian Mountains in the upper reaches, the Loess Plateau in the middle reaches, and the Yellow River Delta in the lower reaches being highly vulnerable to ecological degradation. On the other hand, the ecological environment has been seriously damaged by over-tourism, which has brought ecological problems, such as carbon emissions, soil consolidation, and soil erosion. From 2010 to 2013, the TES in the YRB rapidly improved, and the number of I cities decreased rapidly from 48 to 11. This is because the Chinese government started to adopt the strategy of ecological civilization construction, and emphasized the importance of developing low-carbon green tourism. From 2013 to 2019, the TES status of the Yellow River basin showed significant spatial divergence, in which the cities located in the Hexi Corridor, Wei River basin, and the East Henan Plain still exhibited a status of extreme insecurity. However, the cities located in the Fen River basin and the Yellow River delta generally became better. This is due to the rapid development of low-carbon tourism and ecotourism in Shanxi Province, located in the Fen River Basin, and Shandong Province, covering the Yellow River Delta under the guidance of the Eco-Province Construction Program. In contrast, the cities with statuses of extreme insecurity still focus on sightseeing tours, which cause significant damage in the ecological environment.

4.2.2. Spatial Agglomeration Characteristics

To investigate the spatial clustering characteristics of tourism ecological safety in the YRB, the Moran’s I of TES was calculated using Stata16, as shown in

Table 2. Evaluation index system of tourism ecological security.

Year	Moran’s I	p-Value	Z-Value	Spatial Pattern
2004	0.016	0.693	0.394	Random
2007	0.014	0.713	0.368	Random
2010	0.015	0.707	0.376	Random
2013	0.158	0.024	2.264	Clustered
2016	0.118	0.082	1.741	Clustered
2019	0.164	0.019	2.352	Clustered

. The results show that the global Moran’s I was positive from 2004 to 2019. However, the p values greater than 0.1 from 2004 to 2010 indicate that the global spatial autocorrelation of TES was insignificant, and the spatial pattern of TES showed a random distribution before 2010. However, the global Moran’s I increased significantly and passed the 10% robustness test after 2013, indicating that TES presented a significant positive global spatial correlation. This is because different provinces adopted differentiated tourism development strategies after 2013 [64]. Among them, where TES gradually improved, Shanxi and Shandong Province vigorously developed their ecotourism, which helped to form high-level TES agglomeration areas.

To further analyze the clustering characteristics of TES in local space, its Getis-Ord Gi* index was calculated according to Formula (11), and the distribution of hot and cold spots was visualized using ArcGIS 10.8 (Figure 8). There are three hot spots of TES, located in the upper reaches, the intersection of the middle and upper reaches, and the lower reaches of the Yellow River. Specifically, hot spots in the upper reaches only appeared in 2004, including Baiyin, Lanzhou, Longnan, and Zhongwei. The hot spot at the intersection of the upper and middle reaches showed a trend of expansion, extending southwards from the Hetao Plain to the entire Fen River Basin, including TaiCNY, Datong, Xinzhou, Jinzhong, and other cities. Moreover, the scope of the hot spots located in the eastern part of the Shandong Peninsula has shown a trend of increasing first and then decreasing since 2010. In 2013, it covered most cities, including Weihai, Yantai, Qingdao, Weifang, and other famous seaside tourist destinations in China, but after that, only Weihai was left. This may be due to the rapid development of tourism after 2013, which threatened the TES once again. The scope of cold spots of TES is small and mainly includes two areas. The cold spot located at the intersection of the middle and lower reaches mainly includes cities such as Zhengzhou, Xuchang, and Xinxiang, which had the most extensive range in 2004 but almost disappeared after then. Another one located in the eastern Henan Plain mainly includes cities such as Zhoukou and Xinyang, which have generally suffered from the problem of high population density and unreasonable tourism development. In addition, there are a small number of cold spots in a “sporadic” distribution pattern, such as Yan’an, Qingyang, Wuzhong, and GuCNY.

4.2.3. Spatial Evolutionary Trends

The spatial trend surface analysis tool of ArcGIS10.8 was used to analyze the dynamic evolutionary trends of TES according to Formula (12), and the results of critical years were visualized. As shown in Figure 9, the arrow of the X-axis is oriented to the east, the Y-axis is oriented to the north, the Z-axis represents the development level of TES, the north–south fitting curve of TES is blue, and the east–west fitting curve is green. Overall, no absolute high or low values existed in the north–south or east–west directions during 2004–2019, suggesting that the level of TES is not “spatially locked”. However, the spatial trend surface of TES in the east–west direction is relatively flat, while that in the north–south direction is steeper. Specifically, in the east–west direction, the spatial pattern of TES showed that “the upstream was higher, and the midstream and downstream were slightly lower” in 2004, which gradually changed to “increasing from downstream to upstream” after 2007. Only in 2016 did the spatial pattern of “collapse in the middle, the bulge in the upstream and downstream” appear briefly. In the north–south direction, the spatial pattern gradually changed from “decrease from north to south” to “bulge in the middle and collapse in the south and north”. However, the worst TES situation in the southern Yellow River Basin has not changed, indicating that cities, mainly located in the Wei River Plain and Eastern Henan plains, are still facing a crisis in TES.

4.3. The Dynamic Transfer Characteristics of Tourism Ecological Security

The change in the TES level of cities in the YRB is a dynamic process. The questions of whether cities at lower ranks can transfer to the next rank and whether cities with higher rankings are at risk of downgrading need to be answered. To answer the above questions, according to Formulas (13) and (14), traditional Markov chains and spatial Markov chains were constructed, respectively, to analyze the dynamic transfer law of TES. In addition, according to the classification standard of TES, cities were classified as showing extreme insecurity in I, insecurity in II, and relative insecurity in III. Furthermore, since the numbers of cities in states of critical insecurity, critical security, and relative security are small, they were uniformly classified as IV to avoid outliers.

As shown in

Table 3. Markov matrix of tourism ecological security from 2004 to 2019.

ti/ti+1	n	I	II	III	IV
I	630	0.8714	0.1286	0	0
II	424	0.0613	0.8656	0.0684	0.0047
III	85	0	0.0706	0.8235	0.1059
IV	31	0	0	0.0323	0.9677

, the elements on the diagonal line represent the probability that the TES type does not change. In contrast, the elements on the off-diagonal line represent the transition probability between different TES types. First, all elements on the diagonal are significantly higher than those on the off-diagonal, suggesting that the TES at all grades tends to remain the same. Among them, the probability of IV remaining unchanged is as high as 96.77%, and the probability of III (with the lowest probability) has reached 82.35%. In addition, the transfer probability is between 0.47% and 12.86%, and the probability of transferring to higher levels is greater than that of transferring to lower levels, indicating that the TES of the cities generally tends to improve. However, even if TES transfer occurs, it is mostly between adjacent levels, and the probability of transition across grades is very low (the only possibility is to transfer from II to IV, with a probability of 0.47%). Furthermore, although TES has a high probability of transitioning to the upper level, it is also necessary to be alert to the risk of regressing from a high to a low level. Specifically, the probability of regression from II to I reached 6.13%, from III to II reached 7.06%, and from IV to III was 3.23%.

With the adoption of collaborative environmental governance measures and the “holistic tourism” policy in the YRB, the spatial connection of TES between cities has become increasingly close [65]. In addition, Tobler’s First Law of Geography proposes that any geographical phenomenon is spatially correlated. We also confirmed that the TES in the Yellow River Basin began showing a significant positive spatial correlation after 2013. Therefore, a spatial Markov chain of TES was constructed by adding a spatial lag term to the traditional Markov shift probability matrix to verify whether the spatial lags influence TES’s rank shift in neighboring cities. As shown in

Table 4. Spatial Markov matrix of tourism ecological security from 2004 to 2019.

Spatial Lag	ti/ti+1	n	I	II	III	IV
I	I	373	0.9169	0.0831	0	0
	II	69	0.0870	0.8551	0.0579	0
	III	31	0	0.0323	0.9677	0
	IV	0	0	0	0	0
II	I	253	0.8142	0.1858	0	0
	II	326	0.0613	0.8712	0.0613	0.0062
	III	41	0	0.1220	0.7073	0.1707
	IV	27	0	0	0.0370	0.9630
III	I	4	0.2500	0.7500	0	0
	II	29	0	0.8276	0.1724	0
	III	12	0	0	0.9167	0.0833
	IV	4	0	0	0	1
IV	I	0	0	0	0	0
	II	0	0	0	0	0
	III	1	0	0	0	1
	IV	0	0	0	0	0

, compared to the traditional Markov transition matrix, the transfer probabilities of TES levels in the Yellow River Basin changed significantly after considering spatial factors. In general, the probability of TES shifting upward becomes more significant, and the probability of shifting downward decreases when adjacent to higher-ranked cities; the opposite is true when adjacent to lower-ranked cities. Specifically, when adjacent to I, the probability of I changing to a higher grade decreases by 4.55%. The probability of II shifting to III decreases by 1.05%, while the probability of downgrading increases by 2.57%. The probability of III shifting to IV decreases to 0, but the probability of downward shifting also decreases by 3.83%, indicating that high-level cities have a self-protection mechanism. When adjacent to II, the probability of I shifting to higher grades increases significantly by 10.27%, indicating a positive spatial spillover effect in high-level cities. The probability of II remaining unchanged increases by 0.56%, with little change in the probability of grade shift. The probability of III shifting to lower grades increased by 5.14%, but the probability of shifting to higher grades also increased by 6.48%, indicating that neighboring cities greatly influence it. The probability of IV transferring to lower grades is almost unchanged, which further verifies the existence of a self-protection mechanism for high-grade cities. When adjacent to III, the probability of I changing to a higher level is significantly increased to 75%, indicating that III’s positive spatial spillover effect is noticeable. The probability of II moving to III is also increased by 10.4%, and it has eliminated the downgrade. The probability of III moving to a higher grade does not increase, but it also eliminates the risk of downgrading. In addition, when III is adjacent to IV, it receives a positive spillover from high-level cities and achieves a level of promotion.

4.4. Driving Mechanism of Tourism Ecological Security

4.4.1. Influencing Factors Analysis

The TES of the YRB has significant spatial and temporal variability. To clarify the driving mechanism of the variability, referring to the study of Ruan et al. [27], a Geo-Detector was used to analyze the influence level of 27 variables derived from the TES index system to provide a basis for the analysis of the driving mechanism (Appendix A). Then, the trend of their influence degree was visualized through Origin2021. As shown in Figure 10, there is significant variability in the influence degree of each element on TES.

(1)

Driving influencing factors. In D1 to D6, the influence degrees of D1 (growth rate of tourism revenue), D4 (natural growth rate of population), and D6 (GDP growth rate) are relatively stable, generally approximately 0.2, which indicates that tourism development, population growth, and economic development have always been an essential driver for TES. The degree of influence of D2 (growth rate of tourists) on TES shows an increasing trend, from 0.099 in 2004 to 0.254 in 2019, indicating that the pressure of tourism development on the ecological environment has gradually emerged as the number of tourists has increased. The q value of D3 (urbanization rate) shows an inverted U-shaped trend, reaching a maximum value of 0.449 in 2014 and gradually weakening. This is because China followed an extensive and rapid urbanization path before 2013, and urban sprawl caused severe damage to the ecological environment of the YRB. After that, the Chinese government began to adopt the strategy of “new urbanization”, which emphasized the harmonious coexistence of urban development and the ecological environment [66]. The q value of D5 (GDP per capita) varies widely but is mostly above 0.3, confirming that economic development is an essential driver of TES.

(2)

Pressure influencing factors. P1 (tourism spatial index) and P3 (tourist density index) were two factors that exerted tremendous pressure on TES, which is consistent with the research conclusion of Ruan et al. [27]. This is because the increased density of tourists exerted significant pressure on the local water bodies, atmosphere, and soil. Moreover, the uncivilized behavior of tourists also aggravated the deterioration of the local ecological environment. The q value of P2 (population density) shows a fluctuating upward trend, which may be because the increase in population density puts pressure on the supply of water and land resources, and affects TES [67]. The q values of P4 (industrial wastewater discharge), P5 (SO2 emission), and P6 (solid waste output) fluctuated slightly by approximately 0.2, which indicates that wastewater, exhaust gas, and waste discharge were also important reasons for the pressure on TES.

(3)

State influencing factors. The q values of S1 (domestic tourism income), S2 (tourism foreign exchange income), and S3 (per capita tourism income) are always large and show an increasing trend, which indicates that the impact of the tourism economy on TES is significant. This is because, although tourism may negatively impact the ecological environment, some of the tourism revenue is used to improve tourist destinations’ ecological environment to maintain the continued attractiveness to tourists, which contributes to the improvement of the TES. In addition, the degree of ecological damage caused by tourism is gradually decreasing as the Chinese government vigorously develops low-carbon tourism [68]. S4 (green coverage rate of the built-up region) and S5 (annual average concentration of inhalable particles) have less impact on TES. However, any influencing factors cannot be ignored according to the view of open systems theory.

(4)

Impact influencing factors. The impact of I1 to I5 on TES is noticeable, and their q values exceeded 0.4 in 2019. Among them, the q values of I1 (proportion of total tourism revenue in GDP), I3 (tourism economic density), I4 (tourism industry cluster), and I5 (employees in accommodation and catering industry) show a fluctuating upward trend. First, as tourism plays an increasingly important role in urban economic development, local governments should pay more attention to tourism and adopt stricter environmental regulations to ensure the healthy development of tourism [69]. In addition, under the scale effect of tourism, tourism companies continue to invest funds to improve the tourism environment and ensure TES. Tourism employees are an essential guarantee for sustainable tourism development, which is also a vital part of the TES system. The q value of I2 (proportion of tertiary industry) showed an inverted U-shaped trend, which is above 0.4, indicating that tertiary industry is an important influencing factor of TES, as it can provide widespread support for tourism development.

(5)

Response influencing factors. R1 (number of students in ordinary colleges and universities) has the most significant effect on TES, with its mean q value reaching 0.632, indicating that talent supply is a critical factor in improving TES, which also validates the findings of Ruan et al. [27] and Liu and Yin [1]. R2 (proportion of fiscal expenditure in GDP) also has a significant effect on TES, which indicates that financial support is an essential guarantee for ecological improvement. However, the q value of R2 has shown a downward trend in recent years, which indicates that the utility of fiscal funds has declined. This may be due to the government’s inefficient ecological environmental governance under China’s “promotion championship” system. In addition, the q values of R3 (sewage treatment rate), R4 (domestic waste treatment rate), and R5 (comprehensive utilization rate of solid waste) are small and fluctuate irregularly. However, they are also essential factors that cannot be ignored in the TES system.

4.4.2. Driving Mechanism Analysis

According to open system theory and tourism ecological security theory, the TES system is an organic unity in which various subsystems are closely linked and interact, reflecting the symbiotic relationship between tourism development and the ecological environment [69]. Based on the above research, this study explored the driving mechanism of TES in the YRB (Figure 11). First, driver (D) is the source of the changes in the TES system. Urbanization and a new population provide sufficient land and human capital for local economic development, promoting the local economy’s rapid growth. At the same time, tourism, as an essential part of the economic system, also achieves rapid development. However, tourism activities bring “by-products” such as pollutant discharge and the massive consumption of resources, and particularly the irreversible damage to the ecological environment caused by over-tourism, which exerts enormous pressure (P) on the ecosystem of the tourist destination. On the one hand, urban ecological quality has declined; on the other hand, since an excellent ecological environment is an essential basis for maintaining the attractiveness of a tourist destination, the decline in the quality of the ecosystem leads to a slowdown in the growth rate of the tourism economy. In this state (S), the development of the tourism industry and related supporting industries is negatively impacted (I), and the industrial economy and tourism employment levels also decline. Finally, to achieve sustainable tourism development, tourism stakeholders implement a series of responses (R). Local governments adopt strict environmental regulations to force the transformation of high-polluting enterprises, which helps restore the tourism destination ecosystem. In addition, cultivating tourism talent provides adequate specialized human capital for high-quality tourism development, which helps transform the tourism industry from an extensive development model to a sustainable one. Furthermore, the government spends monetary funds on ecological management, such as improving domestic sewage and waste treatment technology, which helps enhance the TES level. In summary, according to the cumulative causation model, responses have a feedback effect on the driver, pressure, state, and impact subsystems. They generate a complete circle for the TES system through reconstruction and integration with each other.

5. Discussion

From 2004 to 2019, the comprehensive TES showed an upwards trend, but its level was still low at the end of the study period. Compared with previous studies, the TES level in the YRB has been consistently lower than that of some key economic regions in China, such as the Yangtze River Delta and Beijing [2,40]. It indicates that the TES in the YRB still faces a significant threat, which also explains the necessity of the Chinese government taking “ecological protection and high-quality development in the YRB” as a national strategy since 2021. Moreover, although COVID-19 had a substantial negative impact on China’s tourism industry, we did not find that the TES in the YRB was significantly disturbed in 2019. This may be because the outbreak of COVID-19 in late 2019 had less impact on tourism throughout the year. Spatially, we found that the cities with high TES levels show a “scattered” spatial distribution, which was consistent with the findings of Ma et al. [2] and Ruan et al. [27]. This indicates that TES may have a “polarizing effect”. Cities with high TES levels have positive spillover effects on neighboring regions. This finding is consistent with the conclusion that Tang et al. [40] obtained by taking seven Beijing districts as a case study. Regarding the driving mechanism, we found that indices related to tourism and the economy are important driving factors, and the driving force of tourism-related factors has gradually increased. In contrast, existing scholars have not drawn the above conclusions [1,27] because they pay more attention to the influence of single indicators. In addition, there are gaps in the existing research, such as the lack of a theoretical basis for the TES index system, the deficient analysis of the spatial dynamic evolution of TES, and the lack of refinement according to the results of driving factors.

Therefore, based on the shortcomings of existing research, some innovations were made. First, this study deepened the theoretical interpretation of the TES evaluation model. The open system theory added to the explanation of TES, which states that any system is an organic whole with a specific function composed of multiple subsystems in a particular form. The system view helps to correct the tendency of “ignoring the whole and focusing on the part” in the process of TES research by some scholars. Second, in previous research, scholars paid more attention to TES’s spatial distribution and agglomeration characteristics rather than its dynamic evolutionary characteristics. Therefore, referring to the method of evolutionary analysis in evolutionary economic geography, a spatial trend surface and Markov chain were used to present the dynamic evolution law of TES. It is an innovation in terms of the perspective and methods of TES research, which helps to more accurately reveal the changing characteristics and predict future development trends. Third, in previous studies, scholars mainly used econometric models to reveal the influencing factors of TES, which cannot focus on the changing characteristics of the impact level. Therefore, to address these shortcomings, the influence level of each factor was measured in critical years and their trends were measured using Geo-Detector, which helps provide targeted policy implications. Finally, different from previous studies on the macro or micro scale, the world-famous Yellow River Basin was selected as a case site. This study made up for the lack of practicality in macroscale studies and spatial comparisons in microscale studies, providing a paradigm for studying TES in other large river basins.

5.1. Practical Implications

According to the findings and the development reality of the YRB, some policy implications are provided.

First, all city governments in the YRB should make ecological protection and high-quality development a fundamental goal at the current stage, and vigorously develop ecotourism. Ecotourism, a vital type of sustainable tourism, is defined by the International Ecotourism Association as a tourism activity that has the dual responsibility of protecting the natural environment and safeguarding the rights and interests of local people’s lives [70]. It has been widely promoted worldwide in many ecologically fragile areas [71]. Therefore, the cities’ tourism managers should formulate an overall plan for ecotourism development based on a full investigation of ecotourism resources to achieve the harmonious coexistence of tourism and the ecological environment. For example, in Zibo, the government built a famous ecotourism attraction, Swan Lake International Slow City, on the sediment accumulation landform of the Yellow River. It promotes the growth of the tourism economy, achieves vegetation recovery and soil restoration, and provides a suitable habitat for more than 200 species of birds, such as swans and the Chinese autumn sand duck.

Second, cities with high or low TES values in the YRB have shown a tendency of spatial agglomeration in recent years, which is not conducive to sustainable tourism development and the collaborative governance of the ecological environment. Therefore, the governments of all the cities in the YRB need to break down administrative barriers and build a joint monitoring mechanism for TES across regions [18]. In addition, the cities located in arid zones have more fragile ecosystems and lower TES levels. They require experts to conduct field visits and develop targeted improvements to prevent the collapse of the TES system. Furthermore, cities in the Fen River basin, such as TaiCNY, Datong, and Xinzhou, are gradually becoming high-level areas of TES due to the transformation of the coal processing industry to industrial tourism and ecotourism. Their successful experience could be extended to other regions, especially cities that are undergoing industrial transformation and upgrading worldwide.

Third, according to the results of the spatial Markov chain, TES has spatial spillover effects. Therefore, it is essential to take advantage of the positive spillover effect of cities with high TES levels to neighboring cities. On the one hand, with the development of information and communications technologies (ICTs), the cross-regional dissemination of codified knowledge breaks the barriers to information exchange between cities [72]. The experience of ecological environment governance in tourist destinations can be shared among cities promptly through the use of ICTs. On the other hand, as an integrated industry, tourism requires the cooperation of upstream and downstream industries and neighboring cities [4]. Therefore, tourism managers should promote the integration of ecotourism with agriculture, forestry, marine, and other related industries, and establish a cooperative relationship with neighboring cities. It is conducive to stimulating the positive driving effect of TES high-level cities and helping TES low-level cities escape from the extensive tourism development model [73].

Fourth, the improvement measures of TES should not only be limited to the governance of the ecological environment, but also pay attention to the critical role of tourism and high-quality economic development, which can provide industrial and economic support for TES, respectively. In addition, according to open systems theory, it is necessary to improve TES in terms of Driver (D), Pressure (P), State (S), Impact (I), and Response (R) [1]. Among them, in the Driver (D) subsystem, excessive urbanization and tourism economic development are the root causes of threats to TES. Therefore, governments should shift their focus from high speed to high quality, which is also in line with the idea of sustainable development [74]. Pressure (D), State (S), and Impact (I) are the most sensitive subsystems of the TES, and they can change at any time due to changes in driving factors and responsive measures. Therefore, it is necessary to set up a real-time monitoring system for them. Moreover, the Response (R) is the tip of the TES system, which is the fundamental driving force for improving the TES level. Therefore, stakeholders should take a series of measures, such as training high-quality tourism talent and providing financial support for the environmental improvement, in order to achieve the proper functioning of the TES system [2].

5.2. Limitations and Outlook

Although this study provides theoretical contributions and policy insights into TES, some deficiencies provide directional guidelines for future research. First, due to the lack of tourism statistics after 2019 in some cities, this study did not examine the changing characteristics of TES in the YRB during COVID-19. What impact does China’s epidemic isolation policy have on TES? What are the changes in the influencing factors of TES before and after COVID-19? Other exciting issues are also worthy of study. Future scholars can use tourism big data to evaluate the TES level during COVID-19. Second, the data used in this study are primarily derived from official government statistics, which may make the TES evaluation framework impossible to apply in other countries due to the non-uniform statistical caliber. Therefore, in the future, remote sensing data, such as land use monitoring data and Nighttime light (NTL) data, can be used to optimize existing studies. Finally, although analyses of the time-series dynamic evolution of the influence level of the factors were compared to the results of a previous study, the influence degree of each factor at different quantiles of TES has not been considered. Therefore, in future studies, scholars can use panel quantile regression models to examine the dynamics of the impact degree of factors on TES at different quantile points to avoid the idealization of mean regression.

6. Conclusions

Based on open systems theory and the DPSIR model, an evaluation system for TES was constructed. Moreover, taking the 78 cities in the YRB as a study site, this study analyzed the spatial and temporal distribution pattern and the dynamic evolution pattern of TES from 2004 to 2019. Then, the change in the degree of influence of each factor was explored and the driving mechanisms discussed. The conclusions are as follows:

(1)

From 2004 to 2019, the comprehensive TES value in the YRB showed a steady upward trend but remained at a deficient level, indicating that the TES in the YRB still faces a significant threat. The difference in TES between cities increased over time. Most cities were still at a deficient level until the end of the study period, with only a few cities breaking out of “low-level equilibrium”. Moreover, the proportion of cities with low status levels of TES declined rapidly, while the proportion of cities with high status levels of TES grown slowly;

(2)

Spatially, low-TES value cities have always been in the majority, and the high-value cities show a scattered spatial distribution, most of which are along the river. Moreover, regarding spatial agglomeration, the TES was randomly distributed before 2013, but showed a significant positive spatial clustering feature after that. Specifically, the range of hot spots extends from the intersection of the middle and upper reaches to downstream, while the cold spots are always scattered. Furthermore, there was no absolute high-value or low-value area on the spatial trend surface in all directions, suggesting that the TES was not regionally locked. Concretely, the spatial trend surface in the east–west direction was relatively flat, while the spatial pattern gradually changed from “decreasing from north to south” to “bulging in the middle” in the north–south direction;

(3)

The dynamic transfer of the TES levels in the YRB had remarkable regularities. In general, all TES levels lacked the vitality of transfer, with a probability of remaining unchanged at over 80%. However, the cities were more likely to shift to higher grades than lower grades, suggesting that the cities’ TES tends to improve overall. Moreover, the changes in TES levels were mainly concentrated in adjacent levels rather than across levels. In the spatial Markov chain, we found that the variation in TES levels was closely related to the TES levels of neighboring cities. In general, the probability of TES shifting upward increases and the probability of shifting downward decreases when adjacent to higher-ranked cities. The opposite is true when adjacent to lower-ranked cities;

(4)

Overall, the factors related to tourism and the economy were TES’s most important driving forces. Moreover, the driving force level of tourism-related factors increased significantly over time. On this basis, the driving mechanism of TES in the YRB was constructed. Specifically, tourism, economic and social development are the original Drivers (D) of the TES system. However, tourism activities, especially over-tourism, also put enormous Pressure (P) on the TES. The decline in the quality of the ecological environment reduces the attractiveness of tourist destinations, which leads to a slowdown in the growth rate of the tourism economy. In this State (S), the development of tourism and related supporting industries is also negatively Impacted (I). Finally, tourism stakeholders implement a series of Responses (R) to restore the TES.