Spatial Distribution Characteristics and Influencing Factors of High-Grade Tourism Resources in Henan Province, China

Abstract

Henan Province is one of China’s most culturally resource-rich regions, yet the spatial organization and driving mechanisms of its high-grade tourism resources remain insufficiently understood. This study investigates the spatial distribution characteristics and influencing factors of 2396 high-grade tourism resource sites across 157 county-level units in Henan Province. An integrated analytical framework combining GIS-based spatial analysis (nearest neighbor index, kernel density estimation, and spatial autocorrelation), geographic detector methods, and a spatial error model is employed to examine the roles of natural, economic, social, and cultural factors. The results indicate that high-grade tourism resources exhibit significant spatial clustering, forming a pronounced core–periphery structure with a dual-center pattern centered on the main urban area of Zhengzhou and Dengfeng. Strong positive spatial autocorrelation is observed, characterized by contiguous hot spots in central and northwestern Henan and cold spots in the eastern plains. The spatial differentiation of high-grade tourism resources is driven by multiple interacting factors, among which topography and material cultural heritage are the primary drivers, while economic development and intangible cultural heritage act as secondary facilitators. These findings provide empirical evidence for understanding inland tourism spatial structures and offer practical insights for heritage-led, cluster-based tourism planning, aimed at promoting sustainable regional development.

1. Introduction

Tourism resources constitute the fundamental basis of tourism activities and play a critical role in shaping the spatial structure and development trajectories of regional tourism systems [1]. Their spatial distribution reflects the interplay of natural conditions and socioeconomic development, influencing tourism accessibility, industrial agglomeration, and regional competitiveness. As tourism expands rapidly and sustainable development gains prominence, understanding the spatial characteristics and underlying drivers of tourism resources has become essential for effective planning and balanced regional growth. Spatial analytical approaches offer robust tools for revealing distribution patterns, spatial heterogeneity, and influencing factors across multiple geographical scales.

Despite a growing body of research on tourism resource spatial patterns and drivers, several gaps remain. Most studies focus on national or coastal regions, while empirical investigations in inland provinces and transitional zones, such as Henan, are limited. This spatial bias constrains comprehensive understanding of how diverse socioeconomic and environmental contexts shape resource distribution. Furthermore, many studies emphasize pattern identification without integrating mechanism-oriented analyses, limiting insights into interactions among natural, cultural, and socioeconomic factors.

Henan Province, a central inland province of China, provides an ideal case to address these gaps. It possesses abundant historical and cultural resources, pronounced spatial heterogeneity, and rapidly evolving tourism development influenced by urbanization and infrastructure expansion. While several recent studies have examined tourism resource distribution in Henan using spatial statistical methods [2,3], these studies have primarily focused on rural tourism or A-level tourist attractions rather than high-grade resources specifically. Moreover, existing Henan studies have typically employed single-method approaches (either GeoDetector or spatial econometrics in isolation) and have not systematically compared interaction effects with independent net effects. To address these gaps, this study is guided by the following research questions and hypotheses:

RQ1: What are the spatial distribution characteristics (aggregation, density structure, and spatial dependence) of high-grade tourism resources in Henan Province?

RQ2: Which natural, economic, social, and cultural factors significantly influence the spatial differentiation of these resources?

RQ3: How do these factors interact to shape spatial patterns?

H1: 

High-grade tourism resources exhibit significant spatial clustering rather than random distribution.

H2: 

Cultural factors (particularly material heritage) and natural factors (particularly topography) demonstrate stronger explanatory power than economic and social factors.

H3: 

Factor interactions produce synergistic enhancement effects that exceed the explanatory power of individual factors alone.

This study examines the spatial distribution characteristics and influencing factors of tourism resources in Henan Province from a multi-scale, mechanism-oriented perspective. An integrated analytical framework combining spatial statistical methods and explanatory models is applied. The nearest neighbor index, kernel density estimation, and spatial autocorrelation analyses are employed to identify spatial patterns and clustering characteristics, while geographic detectors and spatial regression models quantify the effects and interactions of multiple factors. The contributions of this study are threefold: (1) providing empirical evidence from an inland province in central China; (2) advancing substantive knowledge by demonstrating that in culturally dense inland provinces, heritage endowment and topographic heterogeneity exert stronger spatial structuring effects than contemporary economic forces, challenging the infrastructure-led and market-driven narratives dominant in coastal tourism studies; and (3) advancing methodological integration that informs policy and planning for sustainable tourism development.

2. Literature Review

Tourism resources are conventionally understood as natural or human-made attractions capable of satisfying visitor needs [1]. However, contemporary tourism geography increasingly conceptualizes them as relational assets whose value emerges through the interaction of site characteristics, visitor practices, and institutional framing [4,5]. From this perspective, “high-grade” resources represent not merely objective physical entities but socially constructed repositories shaped by heritage discourses, administrative grading systems, and development priorities [4]. This conceptualization has important implications: the distribution of graded resources reflects not only natural endowment and historical accumulation but also the spatially uneven institutional capacity of regions to document and promote sites within official systems.

The spatial distribution of tourism resources has long been recognized as a central topic in tourism geography. Early research primarily examined individual attractions and local-scale tourist movements, emphasizing spatial interactions between supply and demand [1,6]. With advances in GIS and the increasing availability of geospatial datasets, studies have expanded from point-based analyses to regional and systemic investigations, revealing structural patterns and distribution hierarchies across broader spatial scales [7]. More recently, under a sustainable development framework, attention has shifted toward assessing spatial rationality and coordinated resource allocation, particularly in culturally rich regions. Nevertheless, empirical research has exhibited a pronounced spatial bias: most studies concentrate on coastal or economically advanced areas [8], while investigations in inland basins such as the Yellow River have primarily focused on rural destinations or specific attraction types [9], leaving provincial-scale high-grade resource endowments in transitional inland provinces such as Henan underexplored.

A substantial body of research indicates that tourism resources are spatially aggregated and unevenly distributed, influenced by natural endowments, historical processes, and socioeconomic contexts. Attractions often cluster in areas with favorable accessibility and economic foundations, forming core–periphery structures [10,11]. Recent studies in China report pronounced spatial heterogeneity and agglomeration, emphasizing that high-density clusters often correspond with historic or economically central regions [11,12], with similar patterns documented for national-scale rural tourism resources [13,14]. However, the relative influence of driving factors remains contested. Some evidence highlights the primacy of economic conditions and infrastructure investment, whereas other scholarship underscores the persistent spatial anchoring effects of cultural heritage and topographic heterogeneity, which transcend conventional market-driven determinants [5,15]. This divergence suggests that the spatial logics of tourism resource distribution vary substantially across regional contexts—coastal versus inland, mountainous versus plain—necessitating context-sensitive empirical investigations [4,16]. Moreover, the literature frequently treats resources as homogeneous, overlooking differences across grades (e.g., high-grade versus ordinary resources), which may mask distinct spatial mechanisms [17].

Methodologically, GIS-based spatial analysis has become standard practice, enabling the mapping of clustering tendencies, core–periphery structures, and spatial dependence through techniques such as the nearest neighbor index, kernel density estimation, and spatial autocorrelation [11,18]. While these approaches provide valuable visualization and quantitative characterization, they are limited in explaining underlying drivers due to, for example, bandwidth sensitivity and stationarity assumptions [19]. To overcome this, recent studies increasingly combine descriptive GIS analysis with explanatory models, recognizing that natural conditions, socioeconomic attributes, and cultural assets jointly shape spatial differentiation [2,10,20,21,22,23]. Current trends further emphasize mechanism-oriented analysis, integrating GIS spatial statistics with models such as geographic detectors and spatial econometrics [2,11,18,24,25], alongside composite spatial indicators that capture tourism complexity [19]. Nonetheless, studies rarely integrate both interaction-sensitive and regression-based approaches in a single analytical framework, leaving a methodological gap that this study addresses for Henan Province—a culturally rich inland region where spatial patterns and mechanisms remain underexamined.

3. Analytical Framework

Tourism resources are conventionally classified into eight categories—physiographic landscapes, water landscapes, biological landscapes, astronomical and climatic landscapes, buildings and facilities, historical sites, tourist souvenirs and shopping, and cultural activities—which are broadly grouped into natural and cultural types. This classification system, institutionalized through official resource censuses, inherently shapes what is recognized as a “tourism resource” and thereby constrains the spatial inventory under investigation. Consequently, the observed spatial distribution of high-grade resources reflects not only intrinsic formation conditions but also the administrative capacity to identify, grade, and promote sites within official frameworks. This distribution constitutes a multidimensional, dynamically interactive system shaped by the joint influence of natural, economic, social, and cultural factors across spatial scales.

Theoretically, the framework is grounded in two interrelated perspectives. First, tourism resource endowment theory posits that spatial differentiation reflects the uneven distribution of natural and cultural capital stocks, which serve as necessary yet insufficient conditions for resource formation [1,5]. Natural capital is embodied in topographic and hydrological conditions, while cultural capital is manifested in tangible heritage, intangible heritage, and traditional settlements. Second, agglomeration theory suggests that initial concentrations trigger self-reinforcing socioeconomic processes—encompassing investment capacity, consumption potential, urbanization dynamics, and institutional provision—that amplify spatial disparities over time [4,26]. These mechanisms correspond to regional economic scale, urban disposable income, urbanization rates, and public budget expenditure in the empirical model. Synthesizing these perspectives yields three propositions: (a) natural and cultural endowments establish the baseline spatial template; (b) economic and social factors modify this template by enabling or constraining resource recognition, infrastructure provision, and market access; and (c) factor interactions produce emergent patterns irreducible to the sum of individual factor effects.

As illustrated in Figure 1, this study operationalizes these premises through an analytical framework that integrates four dimensions of influencing factors—natural, economic, social, and cultural—and links spatial pattern identification with mechanism-oriented analysis via spatial statistical methods, geographic detectors, and spatial regression models.

3.1. Natural Factors

Natural factors provide the fundamental environmental background for the formation of tourism resources. The aesthetic, recreational, rarity, and uniqueness attributes of natural landscapes directly contribute to the emergence of distinctive tourism resources. Topographic relief determines the vertical differentiation and visual complexity of landscapes, thereby influencing the development potential of characteristic landforms such as mountains, canyons, and waterfalls. Areas with pronounced relief tend to exhibit stronger landscape attractiveness, whereas regions with relatively gentle terrain are generally more conducive to large-scale tourism development and accessibility.

River network density reflects the abundance of surface water resources and the development status of hydrological systems. Regions with dense river networks usually display higher landscape heterogeneity and ecological service functions, which enhance environmental attractiveness and support the formation of water-related tourism resources. Together, topography and hydrological conditions define the natural suitability and spatial constraints of tourism resource distribution.

3.2. Economic Factors

Economic factors represent essential conditions for the emergence, recognition, and upgrading of tourism resources. The level of economic development determines regional investment capacity, infrastructure provision, and tourism market maturity, thereby influencing both the supply and demand sides of tourism resources. Gross Domestic Product (GDP) serves as a comprehensive indicator of regional economic scale and development level. Economically developed regions typically possess stronger fiscal capacity to support tourism infrastructure construction, heritage protection, and resource development.

From the demand perspective, the per capita disposable income of urban residents reflects regional tourism consumption potential. Higher income levels are associated with stronger travel willingness, greater purchasing power, and increased demand for high-quality tourism experiences. This demand pressure encourages destinations to enhance resource quality and service standards, reinforcing a feedback mechanism between tourism consumption and resource upgrading.

3.3. Social Factors

Social factors are theorized to influence tourism resource distribution through two primary mechanisms: market-demand effects (population as a proxy for potential consumption scale) and institutional capacity (government expenditure as public goods provision) [4,26]. Permanent resident population size reflects the potential scale of the tourism market. Densely populated areas more readily attract tourism activities and sustain the demand necessary for tourism-related services, supporting the continuous development and upgrading of tourism resources.

Government public budget expenditure indicates the capacity of local governments to invest in public services and infrastructure. The formation and recognition of high-grade tourism resources depends not only on intrinsic resource endowments but also on supporting facilities such as utilities, communication systems, and visitor services. By improving regional accessibility and reception conditions through fiscal investment, governments can reduce development constraints and facilitate the agglomeration of tourism resources.

3.4. Cultural Factors

Cultural factors are conceptualized as both tangible and intangible capital stocks that generate tourism attractiveness. Following Richards [5] and Li et al. [27], tangible heritage provides spatially fixed attractions, while intangible heritage contributes to experiential diversity and destination differentiation. The inclusion of both dimensions allows the model to test whether spatial fixity moderates the influence of cultural capital on resource distribution.

Intangible cultural heritage represents living cultural practices, including traditional crafts, folk festivals, and oral traditions. Although intangible heritage exhibits weaker spatial fixity than material heritage, it enriches tourism experiences and enhances cultural diversity within destinations. Traditional villages function as integrated carriers of local culture, architectural heritage, and rural landscapes. Villages listed in national or provincial protection inventories generally preserve relatively intact historical features and traditional lifestyles and therefore constitute important high-grade rural tourism resources.

4. Research Methods and Data

4.1. Study Area

Henan Province is located in central China (Figure 2), within the transitional zone between the country’s second and third topographic steps, as well as between the warm temperate and northern subtropical climate zones. It spans four major river basins and exhibits pronounced natural and cultural diversity, providing a complex spatial background for tourism resource distribution analysis.

4.2. Data Sources

To support the spatial analysis of tourism resource distribution and its influencing factors, multiple datasets were collected from official and authoritative sources. These datasets include high-grade tourism resource data and a set of natural, economic, social, and cultural indicators. All data were processed and integrated at the county-level administrative unit to ensure spatial consistency and analytical comparability.

4.2.1. Tourism Resource Data

The data of high-grade tourism resources in Henan Province come from the resource census database of Henan Culture and Tourism Department. The high-grade tourism resources referred to in this study are those rated as “five-grade” by the Department of Culture and Tourism of Henan Province. There are a total of 2396 such resources in Henan Province. Among them, there are 616 natural tourism resources, including 309 physiographic landscapes, 157 water landscapes, 98 biological landscapes, 52 astronomical and climatic landscapes, and 1780 cultural tourism resources, including 775 buildings and facilities, 352 historical sites, 209 tourist purchases and 444 cultural activities.

4.2.2. Influencing Factor Data

Influencing factors were grouped into four categories: natural, economic, social, and cultural.

Natural factors include topographic relief and river network density. Topographic relief was calculated using ASTER GDEM 30 m resolution data obtained from the Geospatial Data Cloud. River network density was derived from river vector data provided by the National Geographic Information Resources Directory Service System.

Economic and social factors were sourced from the Henan Statistical Yearbook (2025), including gross domestic product (GDP), per capita disposable income of urban residents, total retail sales of consumer goods, permanent resident population, urbanization rate, and general public budget expenditure.

Cultural factors were collected from multiple official sources. Representative intangible cultural heritage items were obtained from the Henan Intangible Cultural Heritage Information System; cultural heritage protection units were collected from the official website of the Henan Provincial People’s Government; and traditional Chinese villages were obtained from the Ministry of Housing and Urban–Rural Development. Geographic coordinates of cultural sites were extracted using the Baidu Map coordinate picker.

4.2.3. Data Quality and Limitations

Several data quality issues merit critical reflection. First, the high-grade tourism resource inventory represents an official administrative census conducted by the Henan Provincial Department of Culture and Tourism in 2023. While comprehensive, the grading process involves expert panels whose criteria may incorporate subjective judgments regarding historical significance and market potential, introducing potential bias: politically prominent regions may receive disproportionate attention.

Second, the dataset captures only five-grade resources, excluding lower-grade or ungraded resources that may nonetheless possess tourism value. This truncation could systematically underrepresent emerging resources in peripheral counties.

Third, cultural heritage data represent snapshots of official recognition status and may not fully capture the dynamic vitality of living heritage practices.

Fourth, tourism demand, transport, investment, and policy variables are omitted not primarily due to data unavailability but because they are theoretically conceived as mediators or derivatives of our core explanatory variables, not as exogenous controls. County-level data inconsistencies further hinder inclusion. If correlated with predictors, these omissions may bias our estimates for topography, economy, and heritage. Future work should refine the causal structure and data quality.

4.3. Analytical Methods

This study employed a two-scale analytical design, covering both point-based and areal spatial levels. At the micro-scale, the nearest neighbor index and kernel density estimation were applied to 2396 tourism resource sites to characterize their spatial patterns. At the macro-scale, spatial autocorrelation analysis, GeoDetector, and the spatial error model were employed at the county level to examine the influences of natural, economic, social, and cultural factors on the spatial differentiation of tourism resources. The county was selected as the analytical unit for two reasons: first, socioeconomic and environmental indicators are most complete at this level; second, the county serves as the fundamental decision-making unit within China’s tourism planning system. It should be noted that scale transitions entail the risk of ecological fallacy—statistical associations observed at the county level may not necessarily reflect patterns at the individual site level, and vice versa. To assess the sensitivity of our results to the measurement of the dependent variable, comparative analyses were conducted using three indicators: raw counts, resource density (per 100 km²), and resource count per million residents. Detailed results are presented in Appendix A.1.

4.3.1. Nearest Neighbor Index

The nearest neighbor index (NNI) was employed to assess the spatial aggregation of high-grade tourism resources. The NNI compares the observed mean distance between each point and its nearest neighbor with the expected mean distance under a hypothetical random distribution. Values of the NNI greater than 1 indicate a dispersed distribution, values less than 1 indicate clustering, and values equal to 1 suggest a random distribution.

4.3.2. Kernel Density Estimation

Kernel density estimation (KDE) was used to generate a continuous density surface of tourism resource points. The density value decreases with increasing distance from each resource location, allowing for the identification of areas with high and low resource concentration. Higher density values indicate stronger spatial aggregation, while lower values indicate sparser distribution.

KDE was implemented in ArcGIS 10.8 using the Kernel Density tool. The search radius was determined by the default Silverman’s rule-of-thumb implemented in the software, and the output cell size was set to the default value of approximately 2.19 km (2186.24 m). A quartic kernel function was applied by default in the ArcGIS Kernel Density tool.

4.3.3. Spatial Autocorrelation Analysis

Spatial autocorrelation analysis was conducted to evaluate the spatial dependence of tourism resources across administrative units. Both global and local measures were applied. Global Moran’s I was calculated to determine whether the overall spatial distribution exhibits significant clustering or dispersion. Local Moran’s I was used to detect local clusters (e.g., high–high or low–low) and identify specific hotspot and cold spot locations.

4.3.4. GeoDetector

GeoDetector is a tool for detecting and utilizing spatial heterogeneity. In this study, GeoDetector was used to explore the explanatory power of various influencing factors for the spatial differentiation of high-grade tourism resources in Henan Province. The results were measured by the q value, and the expression was:

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

In this equation, $N$ represents the total number of county-level units (N = 157); ${\sigma }^2$ denotes the global variance of the dependent variable (high-grade tourism resource count); $L$ is the number of strata into which factor X is discretized; $N_h$ is the number of units in stratum h; and ${\sigma }_h^2$ is the local variance within stratum h. The q-statistic ranges from 0 to 1, where higher values indicate stronger explanatory power of factor X for the spatial differentiation of the dependent variable.

To assess the sensitivity of GeoDetector results to discretization choices, continuous variables were discretized using four alternative classification methods—natural breaks (Jenks), quantile, equal interval, and geometric interval—with category numbers set to 5, 6, and 7 for natural breaks and 5 for the others. The natural breaks (5-class) specification was retained as the primary model because it maximizes between-class variance and is most widely applied in GeoDetector studies. Sensitivity results across all alternative schemes are reported in

Table A2. Sensitivity analysis of GeoDetector q-statistics to alternative discretization methods and category numbers.

Factor	Natural Breaks (5)	Natural Breaks (6)	Natural Breaks (7)	Quantile (5)	Equal Interval (5)	Geometric Interval (5)	Rank Stability
X1	0.247	0.238	0.256	0.240	0.200	0.223	High
X2	0.043	0.061	0.035	0.043	0.019	0.070	Low
X3	0.117	0.115	0.163	0.073	0.161	0.105	Medium
X4	0.125	0.145	0.135	0.129	0.120	0.105	High
X5	0.056	0.087	0.080	0.035	0.052	0.038	Low
X6	0.103	0.095	0.131	0.017	0.111	0.094	Medium
X7	0.117	0.134	0.141	0.131	0.106	0.117	High
X8	0.054	0.075	0.060	0.036	0.108	0.067	Low
X9	0.297	0.315	0.306	0.228	0.350	0.271	High
X10	0.097	0.143	0.141	0.081	0.138	0.090	Medium
X11	0.124	0.225	0.305	0.207	0.115	0.201	Low

Note: Rank stability assessed as High (consistently ranked in top 3 across all schemes), Medium (ranked in top 5 with minor variation), or Low (substantial rank changes). The core finding that material culture (X9) and topography (X1) are the dominant explanatory factors is robust across all discretization methods and category numbers. Rural culture (X11) shows higher sensitivity to classification schemes, particularly when the number of categories increases.

.

4.3.5. Spatial Error Model

To account for spatial dependence arising from unobserved or omitted factors, the spatial error model (SEM) was applied. The SEM corrects estimation bias associated with traditional regression when spatial autocorrelation is primarily contained in the error term.

In this study, the SEM was used to estimate the independent effects of influencing factors on the spatial distribution of high-grade tourism resources. The model is expressed as:

$$Y=X\beta +u,u=\lambda Wu+\epsilon$$

In this specification, Y is the dependent variable, X is the matrix of independent variables, β is the parameter vector, u is the spatially autocorrelated error term, λ is the spatial autocorrelation coefficient, W is the spatial weights matrix, and ε is the independently and identically distributed error term.

Prior to spatial regression, the Lagrange Multiplier (LM) test and its robust forms (LM-err and LM-lag) were applied to OLS residuals to discriminate between spatial error and spatial lag specifications. The spatial weights matrix (W) was constructed using first-order Rook contiguity, defining neighbors as counties sharing a common border. Alternative spatial weights—including Queen contiguity, distance-based thresholds (50–200 km), inverse distance weighting, and K-nearest neighbor matrices (K = 4, 6, 10)—were also generated to test specification sensitivity. Because alternative spatial weights encountered matrix singularity or convergence failures under the given sample size and data structure, non-spatial robustness checks were additionally implemented, comprising (1) OLS with heteroskedasticity-consistent standard errors (HC3), (2) negative binomial regression, and (3) subsample analyses split at the median of topographic relief and resource endowment. Detailed diagnostics are reported in Section 5.4.2 and Appendix A.3.

5. Results

This section presents empirical results regarding the spatial distribution characteristics and influencing factors of high-grade tourism resources in Henan Province. First, spatial pattern analysis is conducted to identify overall aggregation, density structure, and spatial dependence. Subsequently, the explanatory power and interaction effects of influencing factors are examined using the GeoDetector model, followed by spatial overlay analysis to unpack the underlying mechanisms. Finally, a spatial error model is employed to assess the independent net effects of key factors while accounting for spatial dependence, and the results are compared with those of the GeoDetector model.

5.1. Spatial Distribution Characteristics of High-Grade Tourism Resources

5.1.1. Overall Spatial Aggregation

The nearest neighbor index (NNI) was applied to evaluate the overall spatial aggregation of high-grade tourism resources in Henan Province. The calculated NNI value was 0.313, which is significantly lower than 1 (p < 0.01), indicating a pronounced clustered distribution pattern (

Table 1. Average nearest neighbor analysis results.

Average Observed Distance (m)	Expected Average Distance (m)	Nearest Neighbor Index	p-Value
1731.15	5532.74	0.313	0.000

). This result confirms that high-grade tourism resources are not randomly distributed and exhibit strong spatial concentration at the provincial scale.

5.1.2. Density Structure and Spatial Heterogeneity

Kernel density analysis was performed using ArcGIS 10.8 to examine local agglomeration patterns of high-grade tourism resources in Henan Province. The results indicate a pronounced dual-center polarization structure, with major clusters in Zhengzhou’s main urban area and Dengfeng City, accompanied by a distinct core–periphery ring pattern (Figure 3). A large, continuous high-density area extends across Zhengzhou, Luoyang, Kaifeng, Jiyuan, Jiaozuo, Xinxiang, Hebi, and Anyang, with overall higher densities in the northwest and lower densities in the southeast. Two very-high-density areas are identified in Zhengzhou and Dengfeng, encompassing prominent resources such as the Yellow River “hanging river,” Henan Museum, Mount Song, and the historic monuments of Dengfeng in The Centre of Heaven and Earth. High-density areas are found in the urban areas of Luoyang (Laocheng, Xigong, Jianxi, Luolong, and Chanhe Hui districts) and Kaifeng (Longting, Gulou, Shunhe Hui, and Yuwangtai districts), as well as in the South Taihang Mountain area (Xiuwu County and northwestern Xinxiang). Representative attractions include Luoyang peonies, the Longmen Grottoes, the Mangshan imperial mausoleum cluster, the Qingming Riverside Landscape Garden, Bian embroidery, and the South Taihang canyons and waterfalls. Additional high-density pockets occur in the urban area of Anyang (Yindu, Beiguan, Wenfeng, and Long’an districts), southwestern Luoyang (Luanchuan County), southern Jiaozuo (Wen County), and eastern Hebi (Xun County), featuring resources such as the Yin Ruins, Laojun Mountain, Tai Chi heritage, and Dapi Mountain. Medium-density ring-shaped areas are distributed in northern, central, and southwestern Henan, represented by Wangwu Mountain, the Red Flag Canal, and Baotianman.

When disaggregated by resource type, natural resources concentrate predominantly in northwestern Henan, whereas cultural resources cluster mainly in the north-central region (Figure 3b,c).

Natural resources exhibit type-specific hotspots. Geomorphological landscapes form multiple high-density zones, with very-high-density cores at Yuntai–Baoquan (South Taihang) and high-density clusters at Mount Wangwu, Mount Laojun, and Mount Baiyun. Water landscapes are concentrated along the Yellow River corridor (Jiyuan–Kaifeng) and at Yuntai–Baoquan. Biological diversity peaks in Luanchuan–Xixia–Neixiang, anchored by the Baotianman National Nature Reserve. Meteorological and climatic attractions cluster at Laojieling–Baotianman and in Dengfeng–Xiuwu.

Cultural resources display a broader multi-nuclei pattern. Historical sites are strongly concentrated in Zhengzhou and Dengfeng (very-high-density), with secondary clusters in Yanshi District, Jiaozuo, Anyang, Kaifeng, and southern Xinyang. Tourism commodities and commercial sites form nine high-density pockets, led by Zhengzhou, Kaifeng, Yanshi, and Xixia. Human activities create a continuous high-density corridor supported by the four ancient capitals (Zhengzhou, Luoyang, Kaifeng, and Anyang). Overall, the kernel density analysis highlights marked spatial heterogeneity, reflecting the combined influence of natural endowments, historical accumulation, and human activities.

5.1.3. Spatial Autocorrelation Analysis

The spatial autocorrelation of high-grade tourism resources in Henan Province was examined using the global Moran’s I. The calculated Moran’s I value was 0.233, with a p-value of 0.001, reaching statistical significance at the 1% level. This result suggests that the spatial distribution of high-grade tourism resources across the 157 county-level units (including counties, cities, and districts) in Henan Province exhibits significant positive spatial autocorrelation.

To gain further insight into local-scale spatial clustering and outlier patterns, we conducted a local spatial autocorrelation analysis using Moran’s I. The analysis identified 32 county-level units with statistically significant local spatial autocorrelation at the p < 0.05 level, comprising 20.38% of all study units. Of these significant units, 24 were homogeneous clusters (HH and LL), accounting for 75% thereof, suggesting that homogeneous agglomeration is the dominant feature of the spatial distribution of high-grade tourism resources across Henan Province.

As shown in Figure 4, hot spots are concentrated in Zhengzhou and Luoyang, forming a contiguous high-value core. Cold spots are located in the eastern Henan Plain, extending from southern Kaifeng to northern Xinyang and constituting a contiguous zonal “low-value corridor.” These results demonstrate the coexistence of spatial agglomeration and regional imbalance in the distribution of high-grade tourism resources.

5.2. Influencing Factors and GeoDetector Results

Based on the research framework, a four-dimensional influencing factor model encompassing natural, economic, social, and cultural dimensions was constructed. The dependent variable in the GeoDetector and spatial regression analyses was the raw count of high-grade tourism resources per county-level unit. Eleven indicators were selected as independent variables: topography (X1), hydrological conditions (X2), GDP (X3), per capita disposable income (X4), total retail sales of consumer goods (X5), permanent resident population (X6), urbanization rate (X7), general public budget expenditure (X8), material culture (X9), intangible culture (X10), and rural culture (X11), measured by traditional Chinese villages. The GeoDetector model was employed to examine the factors influencing the spatial distribution of high-grade tourism resources.

The single-factor detection results indicated that eight variables—material culture (X9), topography (X1), rural culture (X11), per capita disposable income (X4), GDP (X3), urbanization level (X7), population (X6), and intangible culture (X10)—had significant explanatory power for the spatial differentiation of high-grade tourism resources (

Table 2. Single-factor GeoDetector results.

Factor Layer	Indicator	Indicator Interpretation	Q-Value	p-Value	Rank
Natural	Topography (X1)	Relief Amplitude	0.246	0.000	2
	Hydrology (X2)	River Network Density	0.043	0.167	/
Economic	GDP (X3)	Regional GDP	0.117	0.062	5
	Personal income (X4)	Per Capital Disposable Income of Urban Residents	0.125	0.002	4
	Consumption scale (X5)	Total Retail Sales of Consumer Goods	0.055	0.531	/
Social	Population (X6)	Permanent Resident Population	0.103	0.095	7
	Urbanization level (X7)	Urbanization Rate	0.116	0.002	6
	Public expenditure (X8)	General Public Budget Expenditure	0.053	0.124	/
Cultural	Material culture (X9)	Number of Cultural Relic Protection Units at Provincial Level and Above	0.296	0.000	1
	Intangible culture (X10)	Number of Intangible Cultural Heritage Items at Provincial Level and Above	0.096	0.047	8
	Rural culture (X11)	Number of Traditional Chinese Villages	0.148	0.000	3

). Among them, cultural factors exhibited the highest q-values, suggesting stronger explanatory power than natural, economic, and social factors. In contrast, hydrological conditions (X2), consumption scale (X5), and public expenditure (X8) were not statistically significant.

To verify whether the factor rankings were robust to discretization choices, we compared results across five classification methods and three category numbers (Table A2). Material culture (X9) and topography (X1) consistently ranked as the top two explanatory factors across all 11 tested schemes (X9 q-value range: 0.228–0.350; X1 q-value range: 0.200–0.256), confirming that their dominance was not an artifact of the classification method. Per capita disposable income (X4) and urbanization (X7) also demonstrated relatively stable rankings. However, rural culture (X11) exhibited higher sensitivity (q-value range: 0.035–0.305), particularly under quantile classification and with increased category numbers, likely reflecting the highly skewed distribution of traditional villages concentrated in a small number of counties.

Interaction detection results further demonstrated that all factor pairs exhibited two-factor enhancement effects (Figure 5), indicating that the spatial distribution of high-grade tourism resources was shaped by the combined influence of multiple factors. Among the interaction pairs, combinations involving topography (X1), population (X6), GDP (X3), material culture (X9), and intangible culture (X10) showed particularly high explanatory power (q > 0.5), highlighting the importance of synergistic interactions among natural, economic, social, and cultural factors.

5.3. Mechanisms and Spatial Overlay Analysis of Key Factors

The GeoDetector results (Section 5.2) identified material culture (X9), topography (X1), rural culture (X11), per capita disposable income (X4), GDP (X3), urbanization (X7), population (X6), and intangible culture (X10) as statistically significant drivers of spatial differentiation. To unpack the mechanisms underlying these statistical associations, this section presents spatial overlay analyses that link factor distributions to the observed density patterns of high-grade tourism resources.

5.3.1. Natural Factors

Topography ranked as the second strongest single-factor driver (q = 0.246, p < 0.001), confirming that physical relief was a fundamental constraint on resource formation. Henan comprises five primary landform types: platforms (30–70 m), hills (70–200 m), low mountains (200–500 m), medium mountains (500–1000 m), and high mountains (1000–2000 m). Among high-grade resources, 34.60% occurred in hilly areas, 24.50% in low mountains, and 22.79% in medium mountains, compared with only 11.02% on platforms and 7.10% in high mountains. This inverted-U relationship between elevation and resource abundance indicated that moderate relief—neither too flat for scenic differentiation nor too steep for accessibility—optimized tourism resource formation. The spatial overlay confirmed that the dense northwestern clusters (South Taihang, Wangwu, Mount Song) aligned closely with hilly-to-low-mountain terrain.

5.3.2. Economic Factors

GDP (q = 0.117, p < 0.1) and per capita disposable income (q = 0.125, p < 0.01) emerged as significant economic facilitators. Figure 6a showed that GDP followed a concentric ring structure centered on central Henan, with very-high-value and high-value areas covering 13 county-level units. These high-value zones exhibited strong spatial coupling with dense resource agglomerations in Guancheng Hui District, Jinshui District, and Xinzheng City. Resource counts declined steadily as GDP categories decreased, suggesting that local economic scale provided the fiscal and infrastructural conditions necessary for resource discovery, protection, and accreditation.

Figure 6b revealed a similar concentric pattern for urban per capita disposable income, with very-low-value zones forming in the eastern Henan Plain. Zhengzhou and Luoyang, which possessed the highest income levels, also hosted the densest resource clusters. This alignment supported the interpretation that social consumption capacity directly shaped tourism participation and, by extension, the economic viability of maintaining high-grade attractions. The spatial correspondence implied that income-driven demand acted as a reinforcing mechanism: affluent markets attracted investment in tourism infrastructure, which in turn elevated the grade designation of resources.

5.3.3. Social Factors

Population (q = 0.103, p < 0.1) and urbanization (q = 0.116, p < 0.01) were significant in the GeoDetector but operated through nuanced spatial mechanisms. Figure 7a showed that very-high-population zones (e.g., Jinshui District) coincided with dense resource clusters; however, the average number of resources per county dropped sharply from very-high to high population tiers, then stabilized across medium-to-very-low tiers. This pattern suggested a threshold effect: beyond a critical population mass, additional population no longer increased resource endowment, likely because extreme urban density displaced heritage preservation space.

Urbanization exhibited a more complex non-linear relationship (Figure 7b). Moderate-to-high urbanization zones showed strong spatial correspondence with resource clusters, whereas very-high and very-low urbanization zones were resource-poor. This implied an inverted-U or “Goldilocks” mechanism: counties with moderate urbanization retained sufficient rural hinterland and historic fabric to generate tourism resources, while highly urbanized areas may have lost heritage assets to development, and under-urbanized areas lacked the infrastructure to support resource valorization.

5.3.4. Cultural Factors

Cultural factors dominated the GeoDetector rankings, with material culture (X9, q = 0.296), rural culture (X11, q = 0.148), and intangible culture (X10, q = 0.096) all significant. Their spatial overlays revealed distinct mechanisms tied to the spatial fixity of heritage.

Material culture, represented by 2015 national and provincial cultural relics protection units, showed the strongest gradient: the 8 very-high-value counties averaged 49.75 high-grade resources, compared with 6.02 in the 52 very-low-value counties (Figure 8a). The near-linear relationship indicated that tangible heritage stock acted as a direct, spatially anchored seedbed for tourism resources—each protected site represented a potential or actual high-grade attraction.

Intangible culture (944 national and provincial heritage items) displayed a similar but attenuated gradient: very-high-value counties averaged 35.86 resources versus 9.33 in very-low-value counties (Figure 8b). The weaker slope relative to material culture reflected the lower spatial fixity of intangible heritage, which was transmitted through practitioners rather than fixed monuments. Nevertheless, the positive association confirmed that intangible heritage generated tourism resources by providing experiential and thematic content.

Rural culture, measured by 274 traditional Chinese villages, showed a positive correlation with resource endowment (Figure 8c). The three most village-rich counties held 73 high-grade resources collectively, while the five most village-poor counties held only 10. Traditional villages functioned as preserved spatial containers of vernacular heritage, concentrating multiple resource types (architecture, folk customs, landscape) within bounded settlements. Their high GeoDetector ranking (q = 0.148, 3rd overall) underscored that immovable rural heritage was a stronger spatial predictor than intangible heritage alone.

Collectively, the overlay analyses demonstrated that the spatial distribution of high-grade tourism resources was governed by multi-scalar mechanisms: natural topography set the physical template, economic and social factors determined the capacity to valorize resources, and cultural factors—especially spatially fixed heritage—provided the raw material for resource formation.

5.4. Spatial Regression Analysis and Model Comparison

5.4.1. Collinearity Diagnosis

Prior to conducting spatial regression, Pearson correlation analysis and variance inflation factor (VIF) diagnostics were performed to assess potential multicollinearity among the independent variables. The Pearson correlation results (

Table 3. Results of correlation analysis.

	X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	X11
X1	1.000	−0.242	−0.068	0.110	−0.135	−0.212	−0.019	0.051	0.240	0.070	0.440
X2	−0.242	1.000	−0.039	0.104	−0.003	−0.084	0.177	−0.184	−0.146	−0.062	−0.194
X3	−0.068	−0.039	1.000	0.497	0.899	0.768	0.251	0.378	0.187	0.084	−0.071
X4	0.110	0.104	0.497	1.000	0.513	0.179	0.734	−0.199	0.085	−0.224	−0.020
X5	−0.135	−0.003	0.899	0.513	1.000	0.704	0.330	0.219	0.047	0.016	−0.138
X6	−0.212	−0.084	0.768	0.179	0.704	1.000	−0.141	0.657	0.219	0.229	−0.108
X7	−0.019	0.177	0.251	0.734	0.330	−0.141	1.000	−0.529	−0.150	−0.358	−0.109
X8	0.051	−0.184	0.378	−0.199	0.219	0.657	−0.529	1.000	0.382	0.350	0.107
X9	0.240	−0.146	0.187	0.085	0.047	0.219	−0.150	0.382	1.000	0.505	0.346
X10	0.070	−0.062	0.084	−0.224	0.016	0.229	−0.358	0.350	0.505	1.000	0.164
X11	0.440	−0.194	−0.071	−0.020	−0.138	−0.108	−0.109	0.107	0.346	0.164	1.000

) showed absolute correlation coefficients ranging from 0.001 to 0.899. The highest correlation occurred between GDP (X3) and consumption scale (X5, r = 0.899). Strong correlations were also observed between GDP (X3) and population (X6, r = 0.768) and between per capita disposable income (X4) and urbanization rate (X7, r = 0.734), indicating notable linear associations among certain predictors.

VIF analysis (

Table 4. Results of VIF analysis.

X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	X11
1.551	1.137	7.730	3.232	6.671	5.386	3.676	3.323	1.817	1.588	1.377

) showed values ranging from 1.137 to 7.730, with an average of 3.410. GDP (X3), consumption scale (X5), and population (X6) exhibited moderate multicollinearity, with VIFs of 7.730, 6.671, and 5.386, respectively, but none exceeded the critical threshold of 10. Given that X5 and X6 were highly correlated with X3 and theoretically subsumed by it as measures of economic scale and market size, and considering the theoretical primacy of GDP as a comprehensive economic indicator, X5 and X6 were excluded from the regression model. All VIFs in the resulting nine-variable model were below 5, confirming suitability for spatial regression analysis.

5.4.2. Spatial Error Model

Robustness checks under alternative model specifications and subsamples. Because alternative spatial weights encountered matrix singularity or convergence failures (see Section 4.3.5), we validated the SEM findings using non-spatial alternative specifications.

Table A3. Robustness checks under alternative model specifications and subsamples.

Specification	X1 (Topography)	X3 (GDP)	X9 (Material Culture)	X10 (Intangible Culture)	Model Fit
Primary SEM (Rook)	26.73 ***	0.014 ***	0.718 ***	0.569 **	AIC = 1262.78
OLS with HC3	23.37 ***	0.014	0.782 **	0.486	R2 = —
Negative binomial	1.275 ***	0.0004	0.042 ***	0.028	AIC = 1093.5
High-relief subsample	—	—	0.923 **	—	—
Low-relief subsample	—	—	0.310	—	—
High-resource subsample	21.16 **	—	—	—	—
Low-resource subsample	6.26 *	—	—	—	—

Note: *** p < 0.01, ** p < 0.05, * p < 0.10. OLS-HC3 and negative binomial models retain the same nine-variable specification as the primary SEM. Subsample models report coefficients for the theoretically central variables only. Alternative spatial weights (Queen contiguity, distance bands, inverse distance) encountered matrix singularity due to boundary geometry and multicollinearity; therefore, non-spatial and subsample robustness checks were adopted.

reports results from OLS with HC3 standard errors, negative binomial regression, and subsample analyses. Across all specifications, topography (X1) and material cultural heritage (X9) consistently emerged as the most significant predictors, confirming that the primary findings are robust to model specification.

Spatial weight matrix and alternative specifications. Given the geometric properties of county-level administrative boundaries—where adjacent units invariably share common borders—Rook and Queen contiguity produced identical neighbor structures. Distance-based and inverse-distance weight matrices encountered near-singularity attributable to the combination of moderate sample size (n = 157) and structural multicollinearity, preventing convergence of the spatial lag model (SLM) and spatial Durbin model (SDM). Therefore, non-spatial robustness checks were employed as alternative validation strategies: (1) OLS with heteroskedasticity-consistent standard errors (HC3), and (2) negative binomial regression, which was theoretically appropriate for count-dependent variables. Additionally, subsample analyses were performed by splitting the sample at the median of topographic relief and resource endowment. Across all alternative specifications, topography (X1) and material culture (X9) consistently emerged as the most significant predictors (Table A3), confirming that the primary findings were robust to model specification.

SEM results and interpretation.

Table 5. Regression results of spatial error model.

Variables	Coefficient	Standard Error	Z-Value	p-Value
Constant	−26.2711	11.3563	−2.313	0.021 **
X1	26.7271	4.864	5.495	0.000 ***
X2	1945.48	1936.52	1.005	0.315
X3	0.01356	0.00417	3.249	0.001 ***
X4	0.00031	0.00033	0.952	0.341
X7	0.06651	0.09681	0.687	0.492
X8	−0.02622	0.08526	−0.308	0.758
X9	0.71814	0.15411	4.660	0.000 ***
X10	0.56922	0.27887	2.041	0.041 **
X11	0.29825	0.30558	0.976	0.329
Lambda	0.20134	0.11145	1.806	0.071 *

Note: * p < 0.1, ** p < 0.05, *** p < 0.01.

presents the SEM estimates. The spatial autoregressive coefficient Lambda (λ = 0.201, p = 0.071) was marginally significant, confirming the presence of spatial dependence in the error term and validating the choice of the SEM over OLS. This implied that unobserved factors influencing tourism resource distribution—such as regional policy regimes, geological continuity, or shared historical administrative boundaries—spilled over across neighboring counties.

Among the predictors, topography (X1) exerted the strongest independent effect (β = 26.727, p < 0.001). Given that relief amplitude was standardized prior to estimation, a one-standard-deviation increase was associated with an increase of approximately 26.7 high-grade resources, holding other factors constant. This large coefficient reaffirmed that physical terrain was not merely a backdrop but an active generator of scenic and geomorphological attractions.

Material culture (X9) was the second strongest predictor (β = 0.718, p < 0.001). Each additional provincial-level and above cultural relics protection unit corresponded to an increase of roughly 0.72 high-grade tourism resources. The magnitude was economically meaningful: a county with 50 such units (e.g., Dengfeng) was predicted to have approximately 36 more high-grade resources than a county with none, ceteris paribus. This quantified the direct conversion of tangible heritage stock into accredited tourism resources.

GDP (X3) was also significant (β = 0.0136, p < 0.01), though its coefficient appeared small in absolute terms. Given that GDP was measured in billions of yuan, the coefficient implied that a 10-billion-yuan increase in county-level GDP was associated with approximately 0.14 additional high-grade resources. While the marginal effect at the mean was modest, the significance confirmed that economic scale facilitated resource investment and accreditation independently of other mechanisms.

Intangible culture (X10) achieved significance at the 5% level (β = 0.569, p = 0.041). Each additional provincial-level and above intangible heritage item was associated with an increase of approximately 0.57 high-grade resources. The significance of X10 in the SEM—unlike per capita disposable income (X4), urbanization (X7), and rural culture (X11)—suggested that intangible heritage carried a detectable independent net effect once spatial autocorrelation was controlled, likely because it provided thematic content that complemented tangible assets.

Hydrology (X2), public expenditure (X8), per capita disposable income (X4), urbanization (X7), and rural culture (X11) were not statistically significant in the SEM. The insignificance of X4, X7, and X11 was particularly noteworthy because these factors were significant in the GeoDetector. This divergence implied that their influence on tourism resource distribution operated primarily through interaction with other variables (e.g., urbanization moderating the effect of GDP; traditional villages interacting with topography) rather than as standalone determinants. In contrast, X1, X3, X9, and X10 retained significance in both frameworks, identifying them as core independent drivers.

In summary, the SEM isolated the net effects of each factor while filtering out spatial spillovers. The results confirmed a hierarchical driver structure: topography and material culture were the primary independent engines, GDP and intangible culture were secondary facilitators, and hydrology, public spending, income, urbanization, and rural culture exerted influence primarily through synergistic interactions rather than direct independent pathways.

5.4.3. Comparison of GeoDetector and SEM

Comparative analysis between the GeoDetector and SEM provided complementary insights. The GeoDetector assessed each factor’s explanatory power for spatial differentiation via the q-statistic, capturing both independent and interaction effects. In contrast, the SEM quantified the net effect of each variable while controlling for spatial autocorrelation and confounding influences.

Both methods consistently identified topography (X1) and material culture (X9) as primary drivers, with GDP (X3) and intangible culture (X10) as secondary facilitators. Hydrology (X2) and public expenditure (X8) were insignificant in both approaches. Notably, per capita disposable income (X4), urbanization rate (X7), and rural culture (X11) were significant in the GeoDetector but not in the SEM, implying that their influence manifested primarily through interactions with other factors rather than as independent determinants. This highlighted the value of combining interaction-sensitive methods (GeoDetector) with regression-based models (SEM) to comprehensively capture the spatial mechanisms of tourism resource distribution.

6. Discussion

The following sections interpret the empirical findings in relation to theoretical frameworks, the existing literature, and policy contexts, with emphasis on mechanisms and critical implications rather than result restatement.

6.1. Comparison with Existing Studies

Our findings both extend and differentiate from recent Henan-specific studies. Du et al. [2] examined rural tourism in Henan using GeoDetector and identified economic factors as dominant drivers, whereas our analysis of high-grade resources reveals cultural and natural factors as primary determinants. This divergence likely reflects the different spatial logics of rural tourism (more market- and accessibility-driven) versus high-grade heritage resources (more endowment- and history-driven). Kang et al. [3] applied spatial econometrics to tourism development in the Central Plains but focused on tourism economic output rather than resource endowment. By linking resource distribution directly to factor interactions, our study provides a more fundamental understanding of the spatial prerequisites for tourism development in the region.

These comparisons clarify that the present study’s contribution lies not merely in applying established techniques to a new dataset but in redirecting the analytical focus and theoretical framing. While Du et al. and Kang et al. have respectively employed GeoDetector and spatial econometrics in the Henan context, their analytical targets—rural tourism demand and tourism economic performance—operate under market-driven spatial logics. By contrast, our integration of GeoDetector and the SEM targets the spatial prerequisites of high-grade resource endowment itself, revealing an endowment-driven logic that is theoretically distinct from market-accessibility or output-maximization frameworks. This shift in focus—from economic performance to resource formation, and from single-method application to theoretically complementary integration—constitutes the study’s specific contribution to the existing body of Henan literature.

Our findings both confirm and diverge from recent international scholarship. The observed core–periphery structure aligns with Saarinen et al.’s [4] conceptualization of hierarchical spatial organization in tourism systems, where resource concentration creates self-reinforcing agglomeration economies. However, unlike coastal regions where infrastructure-led development dominates spatial patterns [20], Henan’s distribution reflects what Richards [5] termed ‘heritage-led path dependency’—the persistent spatial imprint of historical capitals and sacred landscapes on contemporary tourism geographies.

Regarding spatial heterogeneity, our results support recent calls for context-sensitive tourism planning [19], demonstrating that uniform policy frameworks are ill-suited for inland provinces with pronounced natural and cultural diversity. The strong explanatory power of material culture resonates with Li et al.’s [27] findings on the spatial correlation between traditional villages and rural tourism but extends this logic to high-grade resources at the provincial scale. Notably, our finding that economic factors act as secondary facilitators rather than primary drivers diverges from studies in highly commercialized coastal zones [11]. This suggests that, in culturally dense inland contexts, endowment matters more than market forces for spatial differentiation.

6.2. Driving Mechanisms of Spatial Differentiation

The dominance of topography and material culture in explaining spatial differentiation invites deeper theoretical interpretation. Rather than merely confirming that ‘mountains and heritage sites attract tourism resources,’ the results suggest a historically co-evolved spatial system. In Henan, moderate topographic relief (hills and low mountains) provided both defensive advantages and sacred landscape qualities that facilitated the siting of temples, monasteries, and historic settlements over millennia. These sites subsequently became the spatial anchors for contemporary tourism resource grading, creating a path-dependent distribution where historical geography constrains present-day patterns. This interpretation aligns with Richards’ [5] concept of ‘heritage-led path dependency’ but extends it by demonstrating that topographic conditions acted as the original spatial filter for cultural accumulation.

The secondary role of economic factors (GDP) challenges the prevailing assumption in coastal tourism studies that market forces and infrastructure investment are the primary spatial structuring mechanisms [11,20]. Similar patterns have been documented in sports tourism, where regional imbalances and urban agglomeration effects persist despite economic expansion, suggesting that endowment-driven spatial structures may characterize multiple tourism sectors in China [28]. In Henan—a province where cultural capital is extraordinarily dense but contemporary economic development is spatially uneven—resource distribution appears to reflect ‘endowment-first, development-second’ logic. This finding advances current knowledge by identifying a distinct spatial regime in culturally rich inland provinces, where the stock of historical assets overrides the flow of contemporary economic investment in shaping tourism geographies. It suggests that policy narratives emphasizing infrastructure-led development may be less effective in such contexts unless accompanied by heritage activation strategies.

The interaction effects further reveal that tourism resource agglomeration is not simply additive but emergent. The synergistic enhancement between topography and material culture implies that landscape settings amplify the tourism value of heritage sites beyond their intrinsic historical significance—a mechanism that has been theorized in landscape aesthetics [29] but rarely quantified at the provincial scale. Similarly, the interaction between population and intangible culture suggests that ‘living heritage’ requires demographic thresholds to sustain practice communities, a finding that contributes to ongoing debates about the spatial sustainability of intangible cultural heritage [23].

6.3. Methodological Contributions

By combining GIS-based spatial analysis, GeoDetectors, and spatial error models, this study provides an integrated framework that captures both interaction effects and independent net effects of influencing factors. GeoDetector quantifies nonlinear interactions and synergistic contributions [2,30,31], whereas spatial econometrics provides robust estimates while addressing spatial autocorrelation [32,33]. The spatial error model further clarifies independent net effects while accounting for spatial dependence, confirming that unobserved regional spillovers contribute to clustering patterns [27]. This highlights the importance of incorporating spatial econometric controls to obtain unbiased estimates in tourism geography studies.

Furthermore, the convergence failure of the SLM and SDM underscores the importance of diagnostic-driven model selection, while the robustness of the primary SEM specification enhances confidence in the net-effect interpretation.

The complementarity between GeoDetector’s interaction detection and the SEM’s net-effect estimation addresses a methodological gap identified in recent reviews. Such methodological integration aligns with recent recommendations for multi-method spatial tourism research and is particularly valuable for analyzing culturally and geographically complex inland regions [2,32,33].

6.4. Planning and Policy Implications

First, for local government authorities and tourism bureaus in core counties (Zhengzhou, Luoyang, Dengfeng, and contiguous high-density units), the priority should be quality enhancement and brand consolidation across both cultural and natural assets rather than quantity expansion. Specific actions include: (1) establishing unified interpretation standards and multi-lingual signage systems at World Heritage sites and national geoparks; (2) implementing visitor flow management and carrying capacity monitoring at high-density cultural sites and ecologically sensitive mountain areas (e.g., Mount Song); and (3) creating integrated ticketing and transport pass systems to facilitate multi-site visits within the Zhengzhou–Luoyang corridor, linking cultural landmarks with adjacent natural scenic areas.

Second, for Destination Management Organizations in medium-density ring areas (e.g., Jiyuan, parts of northern Xinxiang, and the Wangwu Mountain region), cluster-based thematic routing should integrate natural and cultural assets. DMOs should develop thematic itineraries—such as the ‘Yellow River Ecological Corridor’ (linking water landscapes and riparian wetlands), the ‘South Taihang Mountain Eco-Trail’ (connecting geomorphological and biological landscapes), and the ‘Tai Chi Culture Trail’—that link secondary attractions to core hubs. These routes should be supported by wayfinding infrastructure, ecological interpretation panels, and digital storytelling platforms that narrate both geological formation and human history.

Third, for heritage institutions, cultural bureaus, and natural resource management agencies, activation programs should target underutilized resource endowments. In counties with high intangible cultural heritage density but low tourism resource counts, living heritage demonstration centers and inheritor-training programs should be established. Concurrently, in counties rich in natural tourism resources (e.g., biological landscapes in Xixia–Neixiang, meteorological landscapes in Laojieling), ecological value conversion strategies should be pursued, including nature education trails, geotourism products, and low-impact eco-tourism certification, ensuring that natural capital is developed without compromising ecological integrity.

Fourth, for peripheral and low-density counties (eastern plain regions and southern mountain counties), planning should focus on connectivity improvement and niche specialization based on local resource endowments rather than direct competition with core areas. Specific measures include: (1) upgrading rural road connections to nearest high-speed rail stations; (2) developing agritourism, eco-tourism, and wellness tourism products tailored to local bioclimatic and topographic conditions; and (3) leveraging traditional village preservation funds and natural conservation programs to create small-scale, dispersed rural tourism nodes that complement rather than replicate the core–periphery model.

6.5. Limitations

Several methodological limitations warrant careful consideration beyond the data-specific concerns already noted in Section 4.2.3. First, the analysis relies on county-level administrative units, which are subject to the Modifiable Areal Unit Problem (MAUP). The spatial aggregation of point-level resources into county counts may mask intra-county heterogeneity and produce different results under alternative zoning schemes. Future research could employ geographically weighted regression or multi-scale spatial models to explicitly address these scale effects. Second, the analysis is cross-sectional and confined to a single time point (2023), which precludes the examination of temporal dynamics, path dependencies, and the causal sequencing of resource formation vis-à-vis economic development. Longitudinal data would be necessary to establish whether tourism resources drive local economic growth or merely co-evolve with it. Third, the findings are context-specific to Henan Province, a culturally rich inland region with distinctive historical and geographic characteristics; generalization to other provinces or countries requires caution. Fourth, the dataset relies on officially recognized and graded resource inventories, which may reflect administrative priorities, political considerations, and uneven survey intensity rather than purely objective resource quality (see Section 4.2.3 for detailed discussion). These methodological caveats should be borne in mind when interpreting the magnitude and generalizability of the reported effects.

7. Conclusions

This study makes three contributions to tourism geography and spatial analysis. First, substantively, it demonstrates that in culturally dense inland provinces, the spatial distribution of high-grade tourism resources follows an ‘endowment-first’ logic, where topographic heterogeneity and material cultural heritage act as primary spatial structuring forces, while economic development plays a secondary facilitating role. This challenges the infrastructure-led and market-driven narratives that dominate coastal tourism studies, suggesting the need for context-specific theoretical frameworks in tourism spatial analysis.

Second, methodologically, the integration of GeoDetector and spatial error modeling within a unified framework demonstrates how interaction-sensitive and regression-based approaches can be complementary rather than substitutive. GeoDetector captures synergistic effects that the SEM cannot detect, while the SEM isolates net effects controlling for spatial dependence. This methodological pairing is particularly valuable for analyzing spatially complex, multi-factor systems in tourism geography.

Third, practically, the findings suggest that heritage-oriented planning strategies—rather than uniform infrastructure investment—are more appropriate for spatially differentiated inland regions. Core areas should focus on quality consolidation and brand integration, while peripheral regions require connectivity improvement and niche specialization. However, these recommendations are context-bound to Henan Province and should be adapted to local institutional capacities and resource endowments in other regions.

For limitations and future directions, see Section 6.5.