Can the pilot of national park system promote industrial agglomeration? double machine learning analysis taking Wuyishan tea industry as an example

The national park system pilot program is a significant initiative in China's ecological civilization institutional reform, and its impact on regional industrial development warrants further exploration. This study utilizes county-level data on tea industry enterprise registration from 2000 to 2022 and innovatively employs Double Machine Learning (DML) model to assess the impact of the national park system pilot program on tea industry aggregation within Wuyishan National Park, examining both horizontal and vertical aggregation dimensions. The research findings indicate the following: (1) The national park system pilot program has significantly promoted horizontal clustering and vertical integration of the tea industry in Wuyishan National Park. After removing outliers, resetting the sample division ratio, and changing the machine learning algorithm, the model results remain robust and reliable. (2) The national park system pilot program can promote tea industry clustering by strengthening fiscal support, technological innovation, and ecological protection. (3) The pilot program has a stronger impact on tea industry clustering in counties with higher economic levels and higher informatization levels. (4) The spatial distribution of horizontal and vertical agglomeration in the tea industry exhibits significant spatial autocorrelation. This study contributes to research on the economic effects of the national park system pilot program and provides theoretical support and practical insights for the green clustering development of agricultural industries.

1 Introduction

In the context of escalating ecological crises worldwide and the ongoing promotion of sustainable development goals, how to balance ecological conservation and agricultural development has become an important issue of widespread concern among academics and policy-makers. As a major institutional innovation in the reform of the ecological civilization system, the construction of national parks has not only achieved comprehensive protection of important ecosystems, but also enhanced the ecological civilization awareness of the entire population and promoted the harmonious coexistence of humans and nature.

However, the role of national parks extends beyond ecological functions; they carry strategic significance in driving regional green transformation and development. Especially in ecologically marginal areas that are dominated by agriculture, the implementation of the national park system may reshape industrial spatial patterns and development pathways through various mechanisms, such as ecological brand building, institutional supply optimization, and strengthened spatial governance. Internationally, the relationship between national parks and agricultural development has been a focus of academic attention (Welteji and Zerihun, 2018; Podawca and Pawat-Zawrzykraj, 2019). The European model of protected areas places greater emphasis on balancing conservation and development (Farkas and Kovács, 2021). In recent years, scholars have begun to focus on the impact of protected areas on the concentration of surrounding agricultural industries (Wang et al., 2023; Jofrey et al., 2025). While existing literature sheds light on international cases, the mechanisms through which the national park system influences agricultural industrial agglomeration—specifically fiscal support, technological innovation, and ecological protection—remain insufficiently explored.

Since China launched the national park system pilot program, ten national park pilot zones, including Wuyishan, Sanjiangyuan, Giant Panda, and Northeast Tiger and Leopard, have been established, covering a total area of over 230,000 square kilometers; this area represents a national ecological security barrier. Wuyishan National Park, which was one of the first national park pilot zones to be established, the birthplace of black tea, and a UNESCO World Heritage Site, has become a crucial experimental field for China to explore the coordinated development of ecological protection and regional economy, especially the coordinated development of the agricultural industry. By 2024, Wuyishan City will have 148,000 mu of tea gardens, a total raw tea output of 30,000 tons, a total industrial chain output value of 15 billion yuan, a tea tax contribution of 228 million yuan, and approximately 120,000 people employed in the tea industry (Wuyishan City Bureau of Statistics, 2025). With the advancement of the national park system pilot program, the development path of the Wuyishan tea industry has undergone significant changes. On the one hand, the pilot policy has driven spatial concentration among tea enterprises through stringent ecological regulations. For instance, establishing ecological standards such as pesticide usage and fertilization protocols has compelled enterprises to enhance the environmental sustainability of their cultivation and processing operations. This has prompted compliant enterprises to cluster spatially, thereby strengthening the horizontal agglomeration of the tea industry (Sheng et al., 2020). On the other hand, the pilot policy has fostered vertical integration within the tea industry chain through targeted incentive mechanisms. For instance, local governments have linked fiscal subsidies to metrics such as green certification and industrial chain coordination, directing resources toward higher-value segments of tea production. This incentivises upstream and downstream enterprises to strengthen collaboration, thereby promoting vertical clustering within the tea sector (Wang et al., 2022). This dual impact exemplifies the complex interplay between ecological conservation and industrial development, mediated by three critical mechanisms: fiscal support (government policies offering financial incentives to industries), technological innovation (advancing agricultural efficiency and product quality), and ecological protection (through park regulations that guide land and resource use). These mechanisms together shape both horizontal and vertical industrial agglomeration, and their roles deserve further systematic examination. National park pilot zones face structural tensions between ecological conservation and industrial transformation. Can the national park system pilot promote horizontal and vertical aggregation of the tea industry? Through what channels can tea industry aggregation be promoted? What heterogeneous characteristics does the aggregation effect exhibit? Answering these questions will provide valuable experience and guidance for promoting agricultural industrial development under ecological institutional constraints. Addressing these questions will provide a more comprehensive understanding of the causal pathways through which national parks influence agricultural industrial agglomeration, offering valuable insights for promoting agricultural industrial development under ecological institutional constraints.

Agricultural industrial agglomeration refers to an organic agglomeration zone that is formed by farmers and enterprises and is centered on the cultivation and management of certain agricultural products in regions with comparative advantages in agricultural natural resources (Wang et al., 2024). The continued concentration of agricultural enterprises in clusters generates horizontal and vertical linkages, ultimately forming a specialized market organization network that integrates planting, processing, and distribution (Du et al., 2020). Scholars have conducted extensive research on issues such as the measurement, influencing factors, and effects of agricultural industrial agglomeration. (1) In terms of measurement methods, scholars commonly use the E-G index (Ellison and Glaeser, 1997a) and location entropy method (Wang et al., 2024) to measure the relative concentration of a particular industry in a specific region. Additionally, some scholars use distance-based industrial agglomeration measurement methods, such as Ripley's K, the M index, and the DO index, which provide new perspectives for studying agricultural industrial agglomeration at the micro level. (2) From the perspective of influencing factors, the formation and development of agricultural industrial agglomeration are comprehensively affected by multiple factors. Traditional theories explore the causes of agglomeration formation from factors such as natural resource endowment, location and transportation, and market scale (Szabó and Ujhelyi, 2024). Research has shown that the degree of infrastructure improvement significantly promotes agricultural industrial agglomeration and that infrastructure, such as transportation and communication, can reduce costs and improve efficiency (Dickens and Lagerlöf, 2023), thereby attracting more agricultural enterprises to form clusters. In addition, government policy support is also an important force that drives the agglomeration of agricultural industries. The government guides agricultural enterprises to cluster in specific areas by formulating preferential policies, providing financial support and strengthening industrial planning. These fiscal policies often operate in synergy with the national park system, offering subsidies and tax incentives that stimulate industry concentration. Technological innovation can increase agricultural production efficiency and product quality, strengthen the competitiveness of agricultural enterprises, and attract more enterprises to join a cluster (Yao et al., 2024). The aggregation of talent provides intellectual support and innovative impetus for the agglomeration of agricultural industries, promoting the upgrading and development of agricultural industries. The national park system's ecological protection efforts, while imposing certain restrictions on industrial activities, can also foster opportunities for the adoption of sustainable agricultural practices, which in turn promote technological innovation and attract investment (He and Jiao, 2022). In addition, some scholars have paid attention to the impact of factors such as the dynamic mechanism of enterprise entry and exit, the diversity of agglomeration types, and the synergistic drive of the policy environment on industrial agglomeration (Guo et al., 2025). (3) Scholars have explored the effects of aspects such as spatial spillover (Ranran and Jingsuo, 2024), nonlinear relationships (Pongpattananurak, 2018), and the shock response. Research has revealed that agricultural industrial agglomeration has a significant positive spatial spillover effect on the efficiency of agricultural green development (Pongpattananurak, 2018), but this effect exhibits regional heterogeneity (Xue et al., 2020). An inverted U-shaped relationship exists between agricultural production agglomeration and environmental efficiency (Repi et al., 2023). The agglomeration of agricultural industries also affects their resilience. For example, when facing external shocks, clustered agricultural industries can better disperse risk and enhance their stability and adaptability.

Overall, although the current research achievements on agricultural industrial agglomeration are abundant, there is still room for further study. First, most studies have only examined agricultural industrial agglomeration from a horizontal agglomeration perspective, lacking a multidimensional analytical framework that integrates both horizontal and vertical agglomeration. Second, empirical research on the impact of national park system pilot programmes on agricultural industrial agglomeration is scarce, particularly lacking spatial autocorrelation tests, causal relationship identification based on quasi-natural experiments. Third, the mechanisms through which national parks influence agricultural industrial agglomeration, particularly in terms of fiscal support, technological innovation, and ecological protection, have not been systematically elucidated.

Therefore, the main contributions of this paper include: first, a dual-dimensional measurement system for horizontal and vertical agricultural industrial agglomeration has been established, enriching the research perspective on agricultural industrial agglomeration; Second, employing Double Machine Learning methods to identify the causal effects of national park system pilot programmes on agricultural industrial agglomeration, providing new methodological support for related policy evaluations; Third, constructing a three-dimensional mechanism framework of “fiscal support—technological innovation—ecological protection” to systematically explain the intrinsic logic of how national park system pilot programmes influence agricultural industrial agglomeration.

2 Literature review and hypothesis development

As a highly systematic and coordinated ecological institutional reform, the national park system pilot program often promotes agricultural industrial agglomeration in a manner that is influenced by many factors. Taking Wuyishan National Park as a representative example, this paper constructs a three-dimensional mechanism transmission model of “fiscal support—technological innovation—ecological protection” to analyse how the national park system pilot program affects the horizontal agglomeration (geographical concentration of enterprises) and vertical agglomeration (industrial chain integration) of the tea industry and then proposes the following research hypotheses.

2.1 Fiscal support promotes tea industry agglomeration

(1) Special funds support ecological transformation and industrial development. The national park system pilot program has driven the introduction of multiple fiscal policies, such as ecological compensation mechanisms, special support funds for tea industry development, tax exemptions (Wang and Zeng, 2022), and loan interest subsidies, enhancing the capacity of tea enterprises in ecological tea garden construction and investments in green processing facilities. For example, the policy titled “Four Major Actions for Financial Support of the Construction of the Wuyishan National Park Protection and Development Belt” provides fiscal guarantees for loan and collateral business innovations for tea enterprises, effectively reducing operational costs for small and medium-sized enterprises (Wang et al., 2023). (2) Fiscal resources drive improvements in infrastructure and service systems. According to the theory of new economic geography, the formation of industrial agglomeration is influenced by both “market potential” and “factor endowments” (Krugman, 1991). Cluster theory further points out that policy support is an important external condition for the formation of industrial agglomeration (Porter, 2000). Domestic and international experience shows that fiscal policy significantly influences the spatial layout of industries through measures such as reducing corporate costs and improving infrastructure. Government investments have upgraded infrastructure, such as roads, communications, and water conservancy, connecting the “capillaries” of tea cultivation, processing, and sales, shortening the logistics cycle of the industrial chain, and enhancing the attractiveness of industrial clusters. Public fiscal investments in platforms such as tea trading markets, ecological processing zones, and warehousing and logistics centers (Weichun and Erling, 2019) have increased resource concentration and service supply capabilities, fostering the formation of spatially concentrated, resource-integrated industrial clusters. Additionally, targeted subsidies and financial support have reduced the entry costs for small and medium-sized agricultural enterprises, encouraging them to cluster around key regions or zones and form a spatially dense network of enterprises. (3) Fiscal performance-oriented incentives promote enterprise clustering and innovation. The government links fiscal subsidies to green production, brand standardization, and production scale, encouraging enterprises that meet high-standard development criteria to receive greater incentives, thereby establishing a mechanism that shifts resources toward the best performers (Pal et al., 2022) and accelerating the concentration of high-value-added enterprises within industrial clusters. (4) Fiscal policies play a pivotal role in connecting resources within the agricultural value chain. Within agricultural industrial clusters, based on the cumulative effects that are generated by forwards and backwards linkages, a vertically specialized market organization network integrating planting, processing, and distribution can be formed (Miao et al., 2023). This industrial chain includes various types of enterprises, which can not only obtain the necessary raw materials and infrastructure from nearby sources, reducing financial and human resource expenditures related to transportation, but also increase the effective supply and enhance the resilience of the agricultural industry. The state has promoted the establishment of agricultural processing industry development funds and supply chain financial tools, significantly increasing investment capacity in the primary processing and deep processing stages and providing financial support for upstream-downstream collaboration. Based on the above analysis, the following hypothesis is proposed:

Hypothesis 1: National park construction can promote agricultural industrial agglomeration through fiscal support.

2.2 Technological innovation promotes the clustering of the tea industry

(1) Collaboration among research institutions drives technological upgrades in the tea industry. The national park system has encouraged research institutions, such as Fujian Agriculture and Forestry University and the Chinese Academy of Tea Sciences, to intensely engage in tea research and development in Wuyishan, resulting in the development of new varieties and green pest control technologies that meet national park ecological standards. Ecological and digitalized tea garden cultivation methods (such as “tea-forest” and “tea-grass” mixed cropping models) have enhanced the sustainability of tea gardens and attracted upstream and downstream enterprises to jointly build an ecological chain. (2) Digitalization and smart manufacturing promote full-chain synergy. Driven by both policy and technology, tea enterprises are accelerating the adoption of intelligent equipment such as IoT-enabled harvesting, AI sorting, and blockchain traceability, improving production and processing efficiency and product quality consistency. This promotes close functional integration and clustered development among upstream and downstream enterprises. (3) Technological innovation decreases the distance between various links in the industrial chain. Within industrial clusters, complementary industries upstream and downstream in the industrial chain can facilitate the vertical dissemination of specialized knowledge, decrease the technological distance between various links in the industrial chain, and promote technological exchange and collaboration between industries, thereby driving the generation of new knowledge and technologies in various links of the industrial chain (Castellini et al., 2025). Technological innovation can promote an increase in the variety of agricultural products and improve product quality, thereby enhancing the resilience of the agricultural industry at the product level. The participants in agricultural industrial clusters leverage advanced technological means to engage in supply chain cooperation through multiple channels, thereby expanding pathways for industrial integration, facilitating the generation and spillover of knowledge, and forming new development pathways, which in turn increase the innovative capacity of agricultural industry resilience. (4) Green innovation catalyzes industrial agglomeration. In technology innovation practices that are guided by ecological principles, the concentrated promotion of green planting technologies, biological pest control, mechanical harvesting technologies, etc., helps enterprises form a “neighborhood clustering” effect based on technological compatibility. Additionally, the establishment of research institutions and technology platforms, such as tea industry science and technology parks, has strengthened the “proximity effect” among enterprises centered on innovation, promoting industrial spatial restructuring and collaborative development among enterprises. Based on the above analysis, the following hypothesis is proposed:

Hypothesis 2: National park construction can promote agricultural industrial clustering through technological innovation.

2.3 Ecological protection promotes tea industry aggregation

(1) Ecological red lines drive the optimization and aggregation of planting space. National parks delineate ecological functional zones, restrict disorderly expansion and extensive planting, and encourage tea gardens to aggregate in compliant areas. Moreover, the government organizes ecological renovations to achieve the “downsizing” and “upgrading” of tea garden space. Additionally, the demarcation of ecological protection zones constrains enterprises' vertical relocation space to some extent (Javeed et al., 2024), reinforcing regional collaboration and vertical integration within the supply chain. (2) Optimized ecological environments enhance brand premium and appeal. National parks provide a “national-level ecological origin” endorsement, significantly enhancing the ecological value perception of the Wuyi Rock Tea brand, thereby creating an ecological “brand magnetism” effect that attracts enterprises to concentrate their operations in areas seeking ecological origin certification (Pizzichini et al., 2020). (3) Ecological environment governance reduces enterprise environmental governance costs. Agricultural industrial clusters can reduce pollution control costs by expanding the production scale, decreasing input costs for various production factors and energy, promoting farmers' adoption of green production technologies, reducing pesticide application intensity, and minimizing environmental pollution (Lanre Ibrahim et al., 2022). This, in turn, influences the governance of agricultural production environments in surrounding areas, thereby promoting agricultural industrial clustering from an environmental perspective. (4) Ecological environmental protection plays a dual role in filtering and incentivizing the horizontal clustering of agricultural industries (D'Alberto et al., 2024). Ecologically sensitive agriculture, such as tea cultivation, often relies on high-quality mountainous terrain, climate, water sources, and other natural resources (Steyn et al., 2023). This spatial dependency naturally leads enterprises to form clusters centered on ecological resources. National ecological protection policies, green access mechanisms, and environmental performance evaluations further reinforce the logic of enterprise clustering in specific ecological locations, making regional ecological environments an important consideration for the selection and layout of enterprise sites (Javeed et al., 2024). Additionally, the promotion of ecological labels and green brands through policy initiatives has led enterprises to preferentially cluster in ecologically prioritized regions to gain certification advantages and market recognition (Taecharungroj et al., 2024), further strengthening the ecological foundation of horizontal clustering. Based on the above analysis, the following hypotheses are proposed:

Hypothesis 3: National park construction can promote agricultural industrial agglomeration through ecological protection.

In summary, the Wuyishan National Park pilot program has addressed the issues of fragmentation, inefficiency, and environmental pressure in the traditional tea industry through fiscal policy investment, ecological protection system construction, and technological innovation system promotion. It has also established a new model of aggregation centered on ecological value. Fiscal investment provides agricultural enterprises with a concentrated public foundation, technological innovation lowers the threshold for synergy, and ecological protection policies set spatial boundaries for concentration. Together, these three factors constitute practical driving force for horizontal concentration in the agricultural industry (see Figure 1).

Figure 1

Figure 1. Mechanism of action.

3 Methodology and research data

3.1 Empirical model

3.1.1 Double machine learning (DML) model

Based on the quasinatural experiment of the national park system pilot program and drawing on existing research (Farbmacher et al., 2022; Zhang and Li, 2023), this study constructs the following partially linear Double Machine Learning (DML) model to explore whether the pilot program can promote the clustering of the tea industry in Wuyishan National Park.

$\begin{array}{ll}{Y}_{it+1}^h={\theta }_{0}Even{t}_{it}+g(X_{it})+U_{it}& (1)\end{array}$

$\begin{array}{ll}{Y}_{it+1}^v={\theta }_{0}Even{t}_{it}+g(X_{it})+U_{it}& (2)\end{array}$

$\begin{array}{ll}E(U_{it}\mid Even{t}_{it},X_{it})=0& (3)\end{array}$

In this equation, i represents the city; t represents the year; $Y_{it+1}^h$ and $Y_{it+1}^v$ represent the horizontal and vertical agglomeration of the tea industry in the county where Wuyishan National Park is located, respectively; Event_it is the policy variable for the national park system pilot program, set to 1 if the pilot program is implemented and 0 otherwise; and θ₀ is the disposal coefficient of primary concern. X_it is a set of high-dimensional control variables whose specific form ĝ(X_it) must be estimated using machine learning algorithms. U_it is the error term with a conditional mean of 0. Estimating the above equation yields:

$\begin{array}{ll}{\widehat{\theta }}_0={\left(\frac{1}{n}({\displaystyle \sum _{i\in I,t\in T}Ev\tilde{e}n{t}_{it}^2}\right))}^{-1}\frac{1}{n}{\displaystyle \sum _{i\in I,t\in T}Ev\tilde{e}n{t}_{it}{\tilde{Y}}_{it+1}}& (4)\end{array}$

$\begin{array}{ll}Ev\tilde{e}n{t}_{it}=Even{t}_{it}-\widehat{m}\left(X_{it}\right),{\tilde{Y}}_{it+1}=Y_{it+1}-\widehat{g}\left(X_{it}\right)& (5)\end{array}$

where n is the sample size. Further examination of its estimation bias is as follows:

$\begin{array}{ll}\sqrt{n}({\widehat{\theta }}_0-{\theta }_0)=a+b& (6)\end{array}$

$\begin{array}{ll}a={\left(\frac{1}{n}{\displaystyle \sum _{i\in I,t\in T}}Even{t}_{it}^2\right)}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i\in I,t\in T}}Even{t}_{it}{U}_{it}& (7)\end{array}$

$\begin{array}{ll}b={\left(\frac{1}{n}{\displaystyle \sum _{i\in I,t\in T}Even{t}_{it}^2}\right)}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i\in I,t\in T}Even{t}_{it}\left[g\left(X_{it}\right)-\widehat{g}\left(X_{it}\right)\right]}& (8)\end{array}$

where a follows a normal distribution with a mean of 0. To accelerate the convergence speed and ensure that the treatment coefficient estimator is unbiased in small samples, the following auxiliary regression is constructed:

$\begin{array}{ll}Even{t}_{it}=m(X_{it})+V_{it}& (9)\end{array}$

$\begin{array}{ll}E(V_{it}\mid X_{it})=0& (10)\end{array}$

where m(X_it) is the regression function of the treatment variable on the high-dimensional control variables and machine learning algorithms are also needed to estimate its specific form $\widehat{m}(X_i)$. V_it is the error term with a conditional mean of 0. The unbiased coefficient estimates are obtained as follows:

$\begin{array}{ll}{a}^{\prime }={\left[E\left(V_{\dot{u}}^2\right)\right]}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i\in I,t\in T}{V}_{it}{U}_{it}}& (11)\end{array}$

$\begin{array}{ll}{b}^{\prime }={\left[E\left(V_{it}^2\right)\right]}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i\in I,t\in T}\left[m\left(X_{it}\right)-\widehat{m}\left(X_{it}\right)\right]}\left[g\left(X_{it}\right)-\widehat{g}\left(X_{it}\right)\right]& (12)\end{array}$

where a′ follows a normal distribution with a mean of 0 due to the use of two machine learning estimates. Therefore, the overall convergence speed of b' depends on the convergence speed of $\widehat{m}(X_i)$ to m(X_i) and ĝ(X_it) to g(X_it), i.e., $n^{-({\phi }_g+{\phi }_m)}$. Compared with Equation (5), $\sqrt{n}({\widehat{\theta }}_0-{\theta }_0)$ converges to 0 at a faster rate, thereby enabling the estimation of an unbiased treatment coefficient.

3.1.2 Global spatial autocorrelation

Global spatial autocorrelation reflects the overall spatial connectivity of the evaluation object (Ord and Cliff, 1981). This metric captures the spatial characteristics of geographically adjacent regional units, categorized into three types: spatial clustering, spatial dispersion, and spatial randomness. Referring to relevant research (Anselin, 1995), Moran′s I is defined as:

$\begin{array}{ll}Moran\text{'}sI=\frac{{\displaystyle {\sum }_{i=1}^n{\displaystyle {\sum }_{j=1}^n{W}_{i,j}(Y_i-\overline{Y})(Y_j-\overline{Y})}}}{S^2{\displaystyle {\sum }_{i=1}^n{\displaystyle {\sum }_{j=1}^n{W}_{i,j}}}}& (13)\end{array}$

In the formula, Y_i and Y_j represent the observed values for regions i and j, respectively. n denotes the number of counties. W_{i, j} represents the spatial weight matrix, set as a 0–1 adjacency matrix. When the global Moran's I value falls between −1 and 0, it indicates spatial negative correlation among observations; When the value lies between 0 and 1, it indicates spatial positive correlation among observations; If the value equals 0, it indicates no spatial autocorrelation among observations.

3.1.3 Local spatial autocorrelation

Local spatial autocorrelation reflects the inter-unit spatial relationships of the evaluation object. It not only analyzes the degree and types of clustering within each spatial unit (Anselin, 1995; Cressie, 2015).

$\begin{array}{ll}LocalMoran\text{'}sI=\frac{(Y_i-\overline{Y}){\displaystyle {\sum }_{j=1}^n{W}_{i,j}(Y_j-\overline{Y})}}{S^2}& (14)\end{array}$

In the formula, W_{i, j} represents a 0–1 spatial matrix; Local Moran's I > 0 indicates that the attribute values of this spatial unit are similar to those of its neighboring units, signifying “high-high” clustering or “low-low” clustering; Local Moran's I <0 indicates that the attribute values of this spatial unit differ from those of its neighboring units, signifying “high-low” distribution or “low-high” distribution.

3.2 Descriptions of variables and data sources

3.2.1 Explained variable

First, horizontal industrial agglomeration is analyzed. The degree of horizontal industrial agglomeration is usually determined by comparing the proportion of a specific sector's output value in a region's total industrial output value with the proportion of that sector's output value in the national total industrial output value (Sun et al., 2019). The formula is as follows:

$\begin{array}{ll}P{Q}_{it}=\frac{p_{ij}/p_i}{P_j/P}& (15)\end{array}$

where PQ_it refers to the horizontal agglomeration of industry j in county i in year t; p_ij refers to the output value of sector j in county i; p_i refers to the total industrial output value of county i; P_j refers to the total output value of sector j in the region; and P refers to the total national industrial output value.

Second, vertical industrial agglomeration is analyzed. Assuming that there are several companies in the tea industry and that these companies continue to produce within the industrial agglomeration zone (Ellison and Glaeser, 1997b), the degree of industrial agglomeration in this region can be represented by EG_i.

$\begin{array}{ll}E{G}_i=\frac{G_i-\left[1-{\displaystyle {\sum }_j}{(x_j)}^2\right]H_i}{\left[1-{\displaystyle {\sum }_j}{(x_j)}^2\right](1-H_i)}& (16)\end{array}$

where x_j represents the ratio of the number of employed persons in region E_j to the total number of employed persons nationwide, i.e., x_j = (E_j/E), which essentially reflects the regional distribution of employed persons in Fujian and Jiangxi Provinces. G_i is the spatial Gini coefficient of the number of employed persons¹, whereas H_i is the level of competition in industry i, which is calculated as $H_i=\sum _f{(z_i^f)}^2$, where $z_i^f=E_i^f/E^i$ represents the proportion of employment in industry i for enterprise f relative to the total employment in industry i across Fujian and Jiangxi Provinces, and $E^i=\sum _f{E}_i^f$, which essentially reflects the scale of employment in enterprise f within industry i (Bai et al., 2019).

3.2.2 Mainly explanatory variable

The implementation of the Wuyishan National Park System Pilot Policy is used as the explanatory variable. In June 2016, the National Development and Reform Commission approved the “Implementation Plan for the Wuyishan National Park System Pilot Zone,” marking the official launch of the Wuyishan National Park System Pilot Program, which became one of China's first 10 national park system pilot zones. Therefore, the Wuyishan National Park System Pilot Program (DID_it) is used as a policy shock to measure its impact on the horizontal and vertical agglomeration of the tea industry.

3.2.3 Control variables

The factors that influence industrial agglomeration include the regional gross domestic product, number of employed workers, average wage, crop planting area, grain production, and number of industrial enterprises. The data are sourced from the China County Statistical Year book, local city statistical year books, and the EPS database. To avoid fluctuations caused by different units and eliminate possible heteroscedasticity, the corresponding variables are treated with natural logarithms in this paper.

3.2.4 Intermediate variables

(1) Fiscal Support (FS_it). Fiscal support is crucial for the effective implementation of national park policies (Helena Torres Da Rocha et al., 2024). This study selects local fiscal autonomy, that is, the ratio of local general budget expenditures to local general budget revenues, as a proxy variable for fiscal support to measure whether national park-related fiscal policies have a resource-driven effect on the clustering of the tea industry.

(2) Technological Innovation (TI_it). Technological innovation is a key driver of the upgrading and clustering of green industries. This paper uses the number of patent applications related to the tea industry as a proxy to indicate technological innovation levels.

(3) Ecological Conservation (EC_it). The level of ecological conservation directly impacts the achievement of the core objectives of national park system reform. This paper employs the frequency of terms related to the environment in government work reports to characterize the intensity of ecological conservation.

3.3 Data sources and processing

This study utilizes county-level business registration data and tea production data from Fujian and Jiangxi Provinces to examine the impact of the national park system pilot program on the horizontal and vertical aggregation of the tea industry. Specifically, these data include the following: (1) Tea production data were manually collected and compiled from 108 counties, drawing mainly on the National Bureau of Statistics, the China County Statistical Year book, and official county websites. Tea production data were manually collected and compiled from 108 counties. (2) Tea enterprise data were obtained from the National Enterprise Credit Information Publicity System (Source: https://www.gsxt.gov.cn/index.html). Enterprises at the county level in Fujian and Jiangxi Provinces with business scopes related to tea, including enterprise names, establishment dates, number of employees, industry categories, and registered addresses, were collected and organized. Additionally, samples where the registered address could not be determined at the county level due to missing registered addresses or addresses only locatable at the provincial level were excluded, as were enterprises in the “public administration, social security, and social organizations” and “international organizations” categories affected by noneconomic factors. Business registration information has stronger timeliness, enabling more timely reflection of the development of the tea industry. (3) Control variables. For example, Financial Support, Technological Innovation, Ecological Protection, Number of on-the-job employees Fixed-line telephone data volume, Wages of on-the-job employees, Local fiscal general budget revenue, Total sown area of crops, Total grain output, Number of industrial enterprises, etc (see Table 1). The data were derived from “China County Statistical Yearbook”, “Nanping Statistical Yearbook”, “Shangrao Statistical Yearbook”, and CEInet Industry Database.

Table 1

Table 1. Statistical description of variables.

4 Empirical results and analysis

4.1 Benchmark regression model results

This paper employs Double Machine Learning model to estimate the policy effects of the national park system pilot program on both horizontal and vertical industrial agglomeration. The sample division ratio is 1:2, and the random forest algorithm is used to predict and solve the main regression and auxiliary regression. The regression results are presented in Table 2. Model (1) controls for city fixed effects, time-fixed effects, and the first-order terms of other city variables across the entire sample range. The regression coefficients for both horizontal and vertical industrial agglomeration in the tea industry are positive, at 0.529 and 1.068 respectively, and significant at the 1% level, indicating that the pilot project of the national park system can promote the horizontal and vertical agglomeration of the tea industry in Wuyishan National Park.

Table 2

Table 2. Baseline regression.

Additionally, in order to reduce the interference of other factors on the experimental results and ensure the reliability of the conclusion, we add the primary and secondary terms of the control variables on the basis of Model (2) and Model (3). The influence coefficients of the national park system pilot on the horizontal and vertical industrial agglomeration of the tea industry are respectively: 0.503, 0.463 and 1.188, 1.197 further indicating that the national park system pilot program can promote the clustering of the tea industry. The primary reasons are as follows: on the one hand, national parks establish ecological red lines to curb the unregulated expansion of tea plantations, promoting their concentrated layout within compliant zones. Moreover, stringent ecological standards—such as regulations governing pesticide usage and fertilizer application—compel enterprises to enhance the environmental sustainability of their cultivation and processing operations (Chen and Ota, 2017). This creates a natural selection effect under ecological thresholds, encouraging compliant tea enterprises to cluster spatially. On the other hand, local governments within national park areas link subsidies to metrics such as green certification and industrial chain coordination. This guides resources toward high-value-added segments of the tea industry (such as deep processing and brand marketing), encouraging upstream and downstream tea enterprises to form vertically integrated networks and fostering vertical clustering among businesses.

4.2 Robustness test

(1) Eliminating the influence of outliers

Considering the influence of outliers of the explained variable on the robustness of the estimation results, in this paper, all variables in the benchmark regression, except for the disposal variable, were subjected to tail processing at the 1%, 99%, 5%, and 95% quantiles. After excluding these outliers, the conclusions obtained more accurately reflect the regular effects of the policy pilot on tea industry agglomeration, avoiding overestimation of policy effectiveness due to individual extreme cases. Models (1)–(4) in Table 3 indicate that after removing outliers, the national park system pilot program still exerts positive effects on both horizontal and vertical agglomeration in the tea industry, with coefficients of 0.428, 0.371, and 0.113, respectively, at the 0.003 significance level. That is, after excluding the influence of extreme values, the national park system pilot policy promotes the enhancement of horizontal agglomeration in the tea industry, and this effect is not driven by individual extreme observations.

Table 3

Table 3. Robustness analysis: tailing treatment.

(2) Resetting the sample division ratio

To avoid the influence of bias in Double Machine Learning model settings on the conclusions, this paper explores the possible impact of the sample division ratio on the conclusions by changing the sample division ratio of the Double Machine Learning model from the previous 1:2 to 1:3 and 1:4. As shown in Models (1)–(4) of Table 4, even after resetting the sample partitioning ratio, the national park system pilot program continues to exert a significant promotional effect on both horizontal and vertical agglomeration within the tea industry. This indicates that the pilot program's promotional effect on tea industry agglomeration is robust and does not exhibit substantial fluctuations due to different model sample partitioning strategies, thereby enhancing the reliability of the research conclusions.

Table 4

Table 4. Robustness analysis: reset split ratio.

(3) Replacing the machine learning algorithm.

To avoid bias in the Double Machine Learning algorithm settings from interfering with the baseline conclusions, the random forest algorithm used earlier was replaced with LASSO regression, gradient boosting, and ridge regression for re-estimation (Zhang and Li, 2023). The results indicate that after replacing the machine learning algorithm, the baseline conclusions of this paper remain robust. As shown in Table 5, regardless of whether LASSO regression, gradient boosting, or ridge regression is employed, the coefficient indicating the impact of national park system pilot programs on tea industry agglomeration remains consistently positive and statistically significant. Moreover, the coefficient for vertical agglomeration exceeds that for horizontal agglomeration. National park system pilot programs not only promote the horizontal concentration of enterprises in geographical space but also facilitate vertical agglomeration within the tea industry through financial support, technical innovation, and ecological protection.

Table 5

Table 5. Robustness test: change machine learning methods.

4.3 Analysis of the mechanisms of action

To further identify the internal mechanisms by which the national park system pilot policy affects the concentration of the tea industry, this paper introduces three mediating variables, namely, fiscal support, technological innovation, and ecological protection levels, as important channel variables for explaining policy effect differences based on the “economic-technical-ecological” analytical framework discussed earlier. This is done to explore how policy incentives affect horizontal and vertical concentration performance through different pathways.

4.3.1 Fiscal support mechanism

Fiscal support is a key policy tool by which the state promotes ecological transformation and industrial development. Column (1) of Table 6 shows that the impact of fiscal support on horizontal agglomeration in the tea industry is 18.340, which is significant at the 10% level; Column (4) shows that its impact coefficient on vertical agglomeration is 46.222, which is significant at the 1% level. This finding indicates that fiscal support plays a significant role in enhancing the clustering of the tea industry.

Table 6

Table 6. Verification of the mechanism of action.

On the one hand, dedicated fiscal funds alleviate the investment pressure on tea enterprises in areas such as ecological park construction and green processing, thereby lowering the barriers for small and medium-sized enterprises to enter national park industrial zones. For example, the policy titled “Financial Support for the Construction of the ‘Four Major Actions' in the Wuyishan National Park Protection and Development Belt” provides support for enterprise financing, interest subsidies, and guarantees, stabilizing enterprise expectations and stimulating clustering intentions. On the other hand, the government links subsidy funds to indicators such as green certification, brand cultivation, and technological innovation, prioritizing support for enterprises with high ecological standards and strong innovative capabilities. This promotes the concentration of resources toward high-quality entities, reinforcing the “stronger-get-stronger” effect in vertical clustering. Additionally, sustained public fiscal investment in infrastructure, such as park roads, water conservancy facilities, warehousing centers, and trading markets, has significantly improved the industrial operating environment, shortened the logistics radius of the industrial chain, and enhanced spatial clustering benefits.

4.3.2 Technological innovation mechanisms

Technological innovation is the key driving force behind the clustering and upgrading of the tea industry. The data in Columns (2) and (5) of Table 6 show that the coefficient of influence of technological innovation on horizontal clustering is 0.603 (p < 0.01), and that on vertical clustering is 0.247 (p < 0.05), indicating that technological factors play a dual role in the spatial restructuring of the industry. The reasons are as follows. First, the national park system has encouraged local governments to strengthen cooperation with universities and research institutions to build platforms for the conversion of scientific research achievements. For example, many research institutions have continued to focus on green cultivation, pest control, and new variety breeding, promoting the promotion of “tea forest” and “tea grass” mixed cultivation models, which has led enterprises with a green development orientation to cluster around the technological core area, strengthening the trend of horizontal clustering. Second, enterprises within the region achieve horizontal competitive advantages and vertical supply chain extensions through process innovation, optimized production workflows, and digital processing technologies (Wu et al., 2023), thereby enhancing cohesion among enterprises and strengthening the guiding effect of the national park system pilot policy on industrial clustering.

4.3.3 Ecological protection mechanism

Ecological protection is the institutional core of the national park system and an indispensable regulatory mechanism for promoting the clustering of the tea industry. The results in Columns (3) and (6) of Table 4 show that the coefficient of the impact of ecological protection on horizontal clustering is 267.163 (p < 0.05), and that on vertical clustering is 327.451 (p < 0.1), indicating its profound influence on the spatial organization of the industry. Specifically, first, ecological protection significantly enhances the brand ecological value of Wuyi Rock Tea. The “National Ecological Origin” label has become an important market identifier for Wuyi Rock Tea, enhancing consumer trust in the product. The brand's “magnetic effect” attracts enterprises to concentrate in the ecological zone or its surrounding areas, forming a spatial aggregation effect that is driven by ecological value. Second, ecological institutions promote standardized management of the industry. The national park uniformly regulates pesticide use, fertilization methods, harvesting management, and processing procedures, enhances the green consistency and collaborative efficiency of the entire industrial chain, and forms vertical coupling under institutional norms. To adapt to strict ecological requirements, enterprises are more inclined to concentrate their operations on compliant platforms, forming a clustered structure within the industrial chain characterized by a clear division of labor and close cooperation. Third, high-level ecological protection not only enhances the quality foundation and ecological brand value of tea and other ecological products but also improves the regional natural environment, providing sustainable ecological support for industrial clustering.

4.4 Heterogeneity analysis

To further explore the differences in the agglomeration effects of the national park system pilot program on the tea industry in different regional development contexts, this study introduces economic level² and information level³ as grouping variables to analyse and identify the heterogeneity impact path.

4.4.1 Economic level heterogeneity

As shown in Table 7, (1) in regions with higher economic levels, the impact coefficients of the national park system pilot program on the horizontal and vertical agglomeration of the tea industry are 0.388 and 0.414, respectively, and both pass the 10% significance test, indicating that the policy has a certain promotional effect in these regions. (2) In regions with lower economic levels, the coefficient of influence of the national park system pilot program on vertical agglomeration is 0.237, which passes the 1% significance test, but the coefficient of influence on horizontal agglomeration is 0.003, which does not pass the significance test, indicating that the policy has limited effectiveness in promoting vertical supply chain integration. As such, the national park system pilot program has a significantly greater impact on the horizontal and vertical agglomeration effects of the tea industry in economically developed regions than in economically underdeveloped regions.

Table 7

Table 7. Economic development heterogeneity.

The reasons are as follows. Firstly, there are differences in resource and platform foundations. Economically stronger regions already possess mature market mechanisms, industrial platforms, and strong organizational coordination capabilities. In economically weaker regions, there are still gaps in primary processing, deep processing, brand marketing, and research and development services, which limit the impact of the pilot program on vertical aggregation. Secondly, there are differences in policy absorption and conversion capabilities. Enterprises and local governments in economically developed regions have greater policy response capabilities and institutional enforcement capacity (Rousell, 2021), enabling them to integrate national park policies with their own industrial upgrading pathways and strengthen the momentum of green horizontal development.

4.4.2 Information level heterogeneity

As shown in Table 8, the findings were as follows: (1) In regions with higher levels of informatization, the impact coefficients of the national park system pilot program on the horizontal and vertical agglomeration of the tea industry are 0.525 and 3.629, respectively, which are significant at the 1% and 5% significance levels. (2) In regions with lower levels of information technology, the coefficients for the impact on horizontal and vertical agglomeration in the tea industry are 1.029 and 1.495, respectively, both of which pass the 1% and 5% significance tests. These results indicate that information technology conditions significantly increase the effectiveness of national park system policies, particularly in terms of supporting supply chain collaboration, resource integration, and value chain extension.

Table 8

Table 8. Information level heterogeneity.

The reasons are as follows. First, regions with higher levels of informatization possess stronger capabilities for the digital integration of industrial chains, facilitating the formation of an integrated development model combining “ecology + industry + data.” In the tea industry, horizontal clustering relies on information sharing among enterprises, brand synergy, and market networks, whereas vertical clustering requires systematic coordination across all stages, from cultivation and processing to sales and services. Regions with high levels of informatization are better able to integrate upstream and downstream resources through digital platforms, supply chain management systems, and big data analysis, thereby achieving deep extension and coordinated development of the value chain. Second, informatization has facilitated the digital assetization of ecological resources, injecting new momentum into industrial aggregation. As the transformation of “green mountains and clear waters” into “golden mountains and silver mountains” increasingly relies on data support, mechanisms for realizing the value of ecological products have been implemented first in highly informatized regions. For example, emerging mechanisms such as tea garden carbon sink monitoring, ecological environment credit evaluation, and green certification digital platforms have enhanced the market appeal of ecological industry development (Zang et al., 2021), indirectly promoting industrial agglomeration effects.

4.5 Analysis of spatial dependence

The preceding section examined the impact of the Wuyishan National Park pilot program on tea industry agglomeration within a non-spatial framework. This section draws upon the concepts of global spatial autocorrelation and local spatial autocorrelation from geography to further explore the spatial characteristics of both horizontal and vertical agglomeration within the tea industry.

As shown in Table 9, both the horizontal and vertical agglomeration Moran's I indices for the tea industry are greater than zero and pass the significance level test. This indicates that horizontal and vertical agglomeration in the tea industry exhibit a relatively significant positive spatial correlation. That is, spatial distribution is not random: counties with high horizontal agglomeration are surrounded by other counties with high horizontal agglomeration, while counties with low horizontal agglomeration are surrounded by other counties with low horizontal agglomeration. Similarly, for horizontal agglomeration in the tea industry, counties with high agglomeration tend to be adjacent to other counties with high agglomeration, while counties with low agglomeration tend to be adjacent to other counties with low agglomeration.

Table 9

Table 9. Global Moran's I values of tea industry agglomeration.

To more intuitively illustrate the spatial clustering characteristics of horizontal and vertical agglomeration in the tea industry, a Moran's dot plot for 2022 was constructed (see Figure 2). It is evident that most counties exhibit high-high and low-low clustering in economic growth, demonstrating pronounced spatial dependency. Specifically, in horizontal agglomeration, 11% and 65% of counties are located in the first and third quadrants, respectively. In vertical agglomeration, 13% and 77% of counties are situated in the first and third quadrants, respectively. In summary, both horizontal and vertical agglomerations of the tea industry exhibit significant spatial autocorrelation in their spatial distribution (Luo, 2021).

Figure 2

Figure 2. Local Moran's I scatter plot of horizontal and vertical tea industry agglomeration (2022).

5 Conclusions and policy recommendations

5.1 Conclusions

Based on a quasinatural experiment of the national park system pilot program, this paper combines county-level economic data and enterprise registration data from Fujian and Jiangxi Provinces from 2000 to 2022 and uses Double Machine Learning model to study the impact, heterogeneity, and mechanism of the Wuyishan National Park system pilot program on the horizontal and vertical agglomeration of the tea industry. The main conclusions include the following:

First, the national park system pilot program has significantly improved the horizontal and vertical agglomeration of the tea industry in the Wuyi Mountain region. The benchmark regression results show that the national park system pilot program has a positive effect on both horizontal and vertical agglomeration, and this impact passes the significance test. This conclusion remains valid under various robustness tests (including removing outliers, adjusting the sample division ratio, and changing the machine learning algorithm), demonstrating strong robustness.

Second, fiscal support, technological innovation, and ecological protection are the key intermediary pathways by which national park policies influence agricultural industrial aggregation. Mechanistic analysis indicates that the national park system pilot program has significantly enhanced local fiscal autonomy, the number of relevant technology patents, and ecological governance investments. These factors have synergistically promoted the horizontal and vertical aggregation of the tea industry through mechanisms such as reducing enterprise entry costs, promoting technology diffusion and standardization, and strengthening ecological branding and environmental constraints.

Third, the national park system pilot program has reshaped the spatial organization logic of the agricultural industry by establishing a three-dimensional synergistic mechanism of “ecology-economy-technology,” providing a new paradigm for regions endowed with ecological resources to achieve high-quality development of characteristic industries. Fiscal policies have provided infrastructure and financial support for enterprise aggregation, technological innovation has enhanced the connectivity efficiency and professional collaboration capabilities of the industrial chain, and ecological protection systems have established spatial boundaries and green constraints for industrial development. Together, these elements constitute the institutional ecosystem for the aggregation of the tea industry in Wuyishan National Park, offering valuable experience for constructing a green development path that harmonizes human activities with nature.

5.2 Policy recommendations

Based on the above research findings, this paper proposes the following policy recommendations from the perspective of synergistic development between national park system construction and the tea industry:

(1) Enhancing the Precision and Sustainability of Fiscal Support

On the one hand, optimize the mechanisms for ecological compensation and industrial support. Establish a special fiscal fund to prioritize support for ecological tea garden renovations, green processing technology upgrades, and brand development for tea enterprises within national park pilot zones. Reduce corporate transformation costs through interest-subsidized loans, tax exemptions, and other measures. Implement a “performance-oriented” fiscal subsidy policy, linking fund allocation to enterprise indicators such as ecological certification, technological innovation, and supply chain integration, to incentivise the clustering of high-quality enterprises. On the other hand, improve infrastructure and public service platforms. Strengthen the construction of transportation, logistics, and digital networks in areas surrounding national parks to reduce spatial distances between supply chain segments and lower collaboration costs for enterprises (He and Wang, 2024). Finally, establish regional tea trading centers, ecological processing parks, and scientific research incubation platforms to promote the concentrated allocation and efficient use of resources.

(2) Promoting technological innovation and digital empowerment

Firstly, establish a collaborative innovation system between industry, academia, and research. Encourage universities, research institutions, and enterprises in national park pilot zones to collaborate on key areas such as ecological cultivation, green processing, and study tour to accelerate the commercialization of scientific and technological achievements (Sæþórsdóttir et al., 2022). Secondly, deepen digital transformation. Support the digital transformation of the tea industry, optimize production processes through the Internet of Things, big data, and artificial intelligence technologies, and achieve data interoperability and collaborative management across the entire chain of cultivation, processing, and sales. Build a regional industrial cloud platform to integrate upstream and downstream enterprise resources, promote information sharing and market connectivity, and enhance the collaborative efficiency of vertical aggregation.

(3) Improve ecological protection and value conversion mechanisms

For one thing, implement spatial planning under strict ecological constraints. Implement differentiated management of tea enterprises, guide tea gardens to concentrate in compliant areas, and avoid the destruction of ecosystems caused by disorderly expansion. For another, activate the market value of ecological brands (Aprylasari and Azizah, 2025). Create a regional public brand of ‘national park ecological tea,' establish a strict certification and traceability system, and enhance product premium capabilities (Hosseini et al., 2021). Explore paths for realizing the value of ecological products (such as carbon trading and ecological credit incentives) to attract social capital to participate in green industrial chain investment.

This study empirically examines the causal effects and underlying mechanisms of the national park system pilot program on the clustering of Wuyishan's tea industry through a dual machine learning model, providing micro-level evidence for understanding the relationship between ecological institutional innovation and agricultural industrial development. However, this study has limitations: First, it lacks comparative research across multiple national parks with different ecological types. This study primarily relies on the single case of Wuyishan National Park, potentially limiting the external validity of its conclusions. China's national park pilot zones exhibit significant differences in resource endowments (e.g., forests, grasslands, wetlands), dominant industries (e.g., ecotourism, specialty agriculture, animal husbandry), and regional socioeconomic foundations. Therefore, future research will select multiple heterogeneous national parks for cross-case comparative studies to identify common patterns in how national park development promotes industrial growth, thereby enhancing the theoretical framework and the generalizability of research conclusions. Second, insufficient examination of long-term dynamic policy effects. Given the recent establishment of the first batch of national parks, this study primarily relies on data from the pilot phase. It can only effectively capture the short-term effects of policy implementation during the pilot period and cannot yet reveal the long-term dynamic effects of national park development. Future efforts will focus on establishing a long-term tracking database (covering multiple dimensions including industrial economy, ecological environment, and socio-demographics). As research data continues to accumulate, we will conduct in-depth analysis of the nonlinear temporal evolution of national park development impacts, thereby providing scientific foundations for the long-term evaluation and dynamic adjustment of national park policies.

Data availability statement

Publicly available datasets were analyzed in this study. This data can be found at: https://www.epsnet.com.cn/index.html#/Index.

Ethics statement

The manuscript presents research on animals that do not require ethical approval for their study.

Author contributions

QC: Conceptualization, Methodology, Investigation, Writing – original draft. BL: Conceptualization, Methodology, Investigation, Writing – review & editing, Supervision.

Funding

The author(s) declare that financial support was received for the research and/or publication of this article. This research was supported by the National Natural Science Foundation of China (42471249) and Research Project of Humanities and Social Sciences of the Ministry of Education (24YJC790013).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declare that no Gen AI was used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

1. ^$G_i=\sum _j{(x_j-s_j^i)}^2$, where $s_j^i=E_j^i/E_i$ represents the proportion of employment in industry i in region j relative to employment in industry i in Fujian and Jiangxi provinces.

2. ^Taking the per capita regional GDP as the indicator to measure the economic level, the average value within the current period was sampled for grouping, and the regions were divided into two categories: “relatively high economic level” and “relatively low economic level”.

3. ^Referring to the level of information infrastructure and the indicators of government digitalization in various regions, the proportion of internet users in the population was selected as the proxy variable, and the average value within the sample period was taken as the boundary, dividing into two groups: “high information level” and “low information level”.

References

Anselin, L. (1995). Local indicators of spatial association—LISA. Geogr. Anal. 27, 93–115. doi: 10.1111/j.1538-4632.1995.tb00338.x

Crossref Full Text | Google Scholar

Aprylasari, D., and Azizah, S. (2025). Coastal tourism's impact on local livestock farming at Bilik Sijile Beach, Baluran National Park. Bulet. Peternakan 49, 366–375. doi: 10.21059/buletinpeternak.v49i1.99355

Crossref Full Text | Google Scholar

Bai, D., Zhang, Y., and Wang, J. (2019). Industrial agglomeration and Chinese enterprises' export behavior: a study based on labor costs of enterprises. Res. World Econ. 46–64. doi: 10.13516/j.cnki.wes.2019.11.004

Crossref Full Text | Google Scholar

Castellini, G., Romanò, S., Merlino, V. M., Barbera, F., Costamagna, C., Brun, F., et al. (2025). Determinants of consumer and farmer acceptance of new production technologies: a systematic review. Front. Sustain. Food Syst. 9:1557974. doi: 10.3389/fsufs.2025.1557974

Crossref Full Text | Google Scholar

Chen, B., and Ota, I. (2017). Stakeholder conflicts under the current nature protection policy in China a case study of the Wuyishan national nature reserve. J. Econ. 63, 48–57. doi: 10.20818/jfe.63.1_48

Crossref Full Text | Google Scholar

Cressie, N. (2015). Statistics for Spatial Data. Hoboken: John Wiley and Sons.

Google Scholar

D'Alberto, R., Zavalloni, M., and Pagliacci, F. (2024). The environmental and socioeconomic impacts of the Italian National Parks: time and spillover effects across different geographical contexts. Glob. Environ. Change. 86:102838. doi: 10.1016/j.gloenvcha.2024.102838

Crossref Full Text | Google Scholar

Dickens, A., and Lagerlöf, N. P. (2023). The long-run agglomeration effects of early agriculture in Europe. Econ. Inq. 61, 629–651. doi: 10.1111/ecin.13140

Crossref Full Text | Google Scholar

Du, J., Xie, J., and Liu, B. (2020). China's agricultural industry agglomeration and agricultural labor productivity: an empirical study based on data from 275 cities. Fin. Econ. Res. 46, 49–63. doi: 10.16538/j.cnki.jfe.2020.06.004

Crossref Full Text | Google Scholar

Ellison, G., and Glaeser, E. L. (1997a). Geographic concentration in US manufacturing industries: a dartboard approach. J. Polit. Econ. 105, 889–927. doi: 10.1086/262098

Crossref Full Text | Google Scholar

Ellison, G., and Glaeser, E. L. (1997b). Geographic concentration in U.S. manufacturing industries: a dartboard approach. J. Polit. Econ. 105, 889–927. doi: 10.1086/262098

Crossref Full Text | Google Scholar

Farbmacher, H., Huber, M., Lafférs, L., Langen, H., and Spindler, M. (2022). Causal mediation analysis with double machine learning. Econom. J. 25, 277–300. doi: 10.1093/ectj/utac003

Crossref Full Text | Google Scholar

Farkas, J. Z., and Kovács, A. D. (2021). Nature conservation vs agriculture in the light of socio-economic changes over the last half-century–Case study from a Hungarian national park. Land Use policy 101:105131. doi: 10.1016/j.landusepol.2020.105131

Crossref Full Text | Google Scholar

Guo, Q., Sun, W., and Sun, H. (2025). Measuring methods and spatio-temporal evolution of the collaborative agglomeration between manufacturing and producer services in Beijing. Acta Geogr. Sin. 80, 415–432. doi: 10.11821/dlxb202502010

Crossref Full Text | Google Scholar

He, S., and Jiao, W. (2022). Adapting traditional industries to national park management: a conceptual framework and insights from two Chinese cases. J. Clean. Prod. 367:133007. doi: 10.1016/j.jclepro.2022.133007

Crossref Full Text | Google Scholar

He, S., and Wang, B. (2024). Driving mechanism for farmers to adopt agroecology for biodiversity conservation: inspiration from tea management in a China National Park. J. Appl. Ecol. 61, 1841–1866. doi: 10.1111/1365-2664.14709

Crossref Full Text | Google Scholar

Helena Torres Da Rocha, A., Guedes-Santos, J., Alexandre Santos Vieira, F., Claudia Mendes Malhado, A., Ladle, R. J., Lizandro Schmitt, J., et al. (2024). What threats do Brazilian National parks face? J. Nat. Conserv. 84:126813. doi: 10.1016/j.jnc.2024.126813

Crossref Full Text | Google Scholar

Hosseini, S., Oladi, J., Amirnejad, H., and Hens, L. (2021). The evaluation of environmental, economic and social services of national parks. Environ. Dev. Sustain. 23, 9052–9075. doi: 10.1007/s10668-020-01011-6

Crossref Full Text | Google Scholar

Javeed, B., Huang, D., Shangguan, D., Ahsan Mukhtar, M., Sajjad, W., Banerjee, A., et al. (2024). Assessing the effectiveness of national park's policies and laws in promoting biodiversity conservation and ecological development in Pakistan. Front. Environ. Sci. 11:1333650. doi: 10.3389/fenvs.2023.1333650

Crossref Full Text | Google Scholar

Jofrey, O. J., Manyama, F. F., and Elisante, F. (2025). Drivers of land use and land cover change around Mkomazi National Park and its conservation implications. Rangel. Ecol. Manag. 102, 37–47. doi: 10.1016/j.rama.2025.05.002

Crossref Full Text | Google Scholar

Krugman, P. R. (1991). Geography and trade. South. Econ. J. 1, 1–16.

Google Scholar

Lanre Ibrahim, R., Bello Ajide, K., Usman, M., and Kousar, R. (2022). Heterogeneous effects of renewable energy and structural change on environmental pollution in Africa: do natural resources and environmental technologies reduce pressure on the environment? Renew. Energy 200, 244–256. doi: 10.1016/j.renene.2022.09.134

Crossref Full Text | Google Scholar

Luo, F. (2021). Research on the impact of spatial agglomeration of tea industry on farmers' income increase: an empirical study based on dynamic panel data model. J. Leshan Normal Univ. 32, 92–96. doi: 10.16069/j.cnki.51-1610/g4.2017.08.015

Crossref Full Text | Google Scholar

Miao, X., Han, J., Wang, S., and Li, X. (2023). Spatial effects of market integration on the industrial agglomeration of agricultural products: evidence from China. Environ. Sci. Pollut. Res. 30, 84949–84971. doi: 10.1007/s11356-023-28303-4

PubMed Abstract | Crossref Full Text | Google Scholar

Ord, A. C. J., and Cliff, A. (1981). Spatial Processes: Models and Applications. London, UK: Pion.

Google Scholar

Pal, R., Panwar, A., Goyal, S. P., and Sathyakumar, S. (2022). Space use by Woolly Wolf Canis lupus chanco in Gangotri National Park, Western Himalaya, India. Front. Ecol. Evol. 9:2021. doi: 10.3389/fevo.2021.782339

Crossref Full Text | Google Scholar

Pizzichini, L., Temperini, V., and Gregori, G. L. (2020). Place branding and local food souvenirs: the ethical attributes of national parks' brands. J. Place Manag. Dev. 13, 163–175. doi: 10.1108/JPMD-06-2019-0043

Crossref Full Text | Google Scholar

Podawca, K., and Pawat-Zawrzykraj, A. (2019). The agriculture and forestry lands level of changes in planning process as a sustainable development measure of municipalities located within national parks impact areas. J. Ecol. Eng. 20, 153–164. doi: 10.12911/22998993/110791

Crossref Full Text | Google Scholar

Pongpattananurak, N. (2018). Impacts from tourism development and agriculture on forest degradation in Thap Lan National Park and adjacent areas. Agric. Nat. Resour. 52, 290–297. doi: 10.1016/j.anres.2018.09.013

Crossref Full Text | Google Scholar

Porter, M. E. (2000). Location, competition, and economic development: local clusters in a global economy. Econ. Dev. Q. 14, 15–34. doi: 10.1177/089124240001400105

Crossref Full Text | Google Scholar

Ranran, C., and Jingsuo, L. (2024). The impact of agricultural production agglomeration on agricultural economic resilience: based on spatial spillover and threshold effect test. Front. Sustain. Food Syst. 8:1464732. doi: 10.3389/fsufs.2024.1464732

Crossref Full Text | Google Scholar

Repi, T., Biahimo, S. D. J., Yasin, I. A., Mokoginta, M. M., Ervandi, M., Mokoolang, S., et al. (2023). Agriculture system and bioresources utilization in villages around Bogani Nani Wartabone National Park: a case study of Ilomata village, Gorontalo. IOP Conf. Ser. Earth Environ. Sci. 1230:012061. doi: 10.1088/1755-1315/1230/1/012061

Crossref Full Text | Google Scholar

Rousell, D. (2021). Cosmopolitical encounters in environmental education: becoming-ecological in the intertidal zones of Bundjalung National Park. J. Environ. Educ. 52, 133–148. doi: 10.1080/00958964.2020.1863313

Crossref Full Text | Google Scholar

Sæþórsdóttir, A. D., Wendt, M., and Ólafsdóttir, R. (2022). Tourism industry attitudes towards national parks and wilderness: a case study from the icelandic central highlands. Land. 11, 1–22. doi: 10.3390/land11112066

Crossref Full Text | Google Scholar

Sheng, C., Wang, Y., and Ye, S. (2020). Discussion on the national park management mode under China's National Park System: a case of Wuyishan National Park System Pilot Area. Int. J. Geoheritage Parks 8, 230–234. doi: 10.1016/j.ijgeop.2020.11.004

Crossref Full Text | Google Scholar

Steyn, C., Jacobs, A., Summerell, B., and Venter, E. (2023). Fusarium diversity from the Golden Gate Highlands National Park. Front. Microbiol. 14:2023. doi: 10.3389/fmicb.2023.1149853

PubMed Abstract | Crossref Full Text | Google Scholar

Sun, J., Xia, T., and Hu, A. (2019). Research on the spatial scale of industrial agglomeration in the Guangdong-Hong Kong-Macao Greater Bay Area. J. Sun Yat-sen Univ. 59, 178–186. doi: 10.13471/j.cnki.jsysusse.2019.02.018

Crossref Full Text | Google Scholar

Szabó, A., and Ujhelyi, G. (2024). National parks and economic development. J. Public Econ. 232:105073. doi: 10.1016/j.jpubeco.2024.105073

Crossref Full Text | Google Scholar

Taecharungroj, V., Vasiljević, Đ., and Pattaratanakun, A. (2024). Snapshots of nature: harnessing flickr data to frame sustainable brand positioning strategies for Thailand's national parks. J. Outdoor Recreat. Tour. 46:100765. doi: 10.1016/j.jort.2024.100765

Crossref Full Text | Google Scholar

Wang, B., He, S., and Min, Q. (2022). Experience and optimisation of the traditional industrial development and transformation in protected areas: a case study of tea industries in Wuyishan National Park. J. Ecol. Rural Environ. 38, 1229–1238. doi: 10.19741/j.issn.1673-4831.2022.0244

Crossref Full Text | Google Scholar

Wang, B., He, S., Min, Q., and Yang, X. (2023). Conflict or coexistence? Synergies between nature conservation and traditional tea industry development in Wuyishan National Park, China. Front. Sustain. Food Syst. 6:991847. doi: 10.3389/fsufs.2022.991847

Crossref Full Text | Google Scholar

Wang, H., and Zeng, Y. (2022). Policy optimization for Hainan Tropical Rainforest National Park based on quantitative comparison of regional policies of free trade port areas. Front. Environ. Sci. 10, 1–14. doi: 10.3389/fenvs.2022.891432

Crossref Full Text | Google Scholar

Wang, S., Xu, Y., Wang, L., and Lv, J. (2024). Research on the impact of agricultural industry agglomeration on agricultural industry resilience: from the dual perspectives of horizontal agglomeration and vertical agglomeration. Econ. Longitude Latitude 41, 1–12. doi: 10.15931/j.cnki.1006-1096.20241218.004

Crossref Full Text | Google Scholar

Weichun, H. E., and Erling, L. I. (2019). Empirical Study on the evolution of industrial agglomeration of China's agricultural products processing industry. Econ. Geogr. 39, 94–103. doi: 10.15957/j.cnki.jjdl.2019.11.012

Crossref Full Text | Google Scholar

Welteji, D., and Zerihun, B. (2018). Tourism–agriculture nexuses: practices, challenges and opportunities in the case of Bale Mountains National Park, Southeastern Ethiopia. Agric. Food Secur. 7:8. doi: 10.1186/s40066-018-0156-6

Crossref Full Text | Google Scholar

Wu, L., Wan, X., Jahanger, A., Li, M., Murshed, M., and Balsalobre-Lorente, D. (2023). Does the digital economy reduce air pollution in China? A perspective from industrial agglomeration. Energy Rep. 9, 3625–3641. doi: 10.1016/j.egyr.2023.02.031

Crossref Full Text | Google Scholar

Xue, L., Shen, Y., and Xu, C. (2020). Agricultural industry agglomeration and green agricultural development: efficiency measurement and impact effects. Econ. Longitude Latitude 37, 45–53. doi: 10.15931/j.cnki.1006-1096.2020.03.002

Crossref Full Text | Google Scholar

Yao, R., Ma, Z., Wu, H., and Xie, Y. (2024). Mechanism and measurement of the effects of industrial agglomeration on agricultural economic resilience. Agric. Basel 14:337. doi: 10.3390/agriculture14030337

Crossref Full Text | Google Scholar

Zang, Z., Xu, W., Ouyang, Z., and Null, N. (2021). Exploration on the value realization of ecological products in China's national park system pilots. Biodivers. Sci. 29, 275–277. doi: 10.17520/biods.2021084

Crossref Full Text | Google Scholar

Zhang, T., and Li, J. (2023). Network infrastructure, inclusive green growth and regional disparities - causal inference based on dual machine learning. Quant. Econ. Technical Econ. Res. 40, 113–135. doi: 10.13653/j.cnki.jqte.20230310.005

Crossref Full Text | Google Scholar