The Functional Transformation of Green Belts: Research on Spatial Spillover of Recreational Services in Shanghai’s Ecological Park Belt

Abstract

The establishment of a new green space system based on the green belt has become a new trend in the world. Shanghai’s Outer Ring Ecological Park Belt (formerly the Outer Green Belt) faces challenges of spatial imbalance in recreational service distribution and mismatched supply and demand in functional allocation during its transition from an ecological barrier to a recreational service provider. An approach based on spatial spillover effects serves as a critical solution to address these issues. We integrate RPS and ROS models to build an evaluation framework, map recreational service supply for 2013, 2018, and 2023, delimit core areas via MSPA, and quantify spatial spillovers with models SLM and SEM. The results show that high-value areas of recreational service levels along the ecological park belt have driven the development of neighboring areas through spatial spillovers, with this promoting effect radiating outward from the core zones. As the distance from the core areas increases, the effect weakens, with 400 m as the maximum effect boundary, 1 km as the critical spillover boundary, and unstable effects with decreased significance beyond 2 km. We further conduct localized spatial spillover analysis using representative parks as case studies. The research provides theoretical support and implementation suggestions for the planning and construction of an ecological park belt.

1. Introduction

In 1995, Shanghai officially initiated the construction of its Outer Green Belt as an ecological barrier to curb uncontrolled urban expansion. With rising living standards and increasing demand for recreational activities during leisure time, the green belt—located at the urban–rural interface—has demonstrated unique geographical advantages, diverse landscape resources, and heterogeneous land-use patterns, endowing it with the potential to foster urban–rural integration and provide high-quality recreational services for the entire city [1]. Over three decades of development, the green belt has evolved from a purely protective green corridor into the current “Outer Ring Ecological Park Belt”. Today, the focus lies on creating recreational spaces for high-density urban areas, offering new leisure lifestyles and enhancing public interaction with nature.

The concept of green belts originated from Ebenezer Howard’s “Garden City” theory. After 1900, rapid urban expansion in London prompted the London County Council to propose “Green Girdles” [2,3]. This idea later developed into the globally recognized “green belt” planning paradigm [4,5]. Over recent decades, worldwide green belt development has undergone continuous adjustments and functional transformations amid controversies, including green belt fragmentation [6], declining benefits [7], and urban sprawl bypassing green belt boundaries [8]. This evolution has established a new trend of creating urban outer green systems based on traditional green belts [9,10].

The Ecological Park Belt represents an upgraded urban open-recreation space, evolved from traditional green belts through structural optimization and functional enhancement, aiming to meet urban–rural integration and public recreational demands [11]. Multiple studies in China have demonstrated that ecological park belts in high-density cities, including Shanghai’s, can optimize spatial configurations, improve functional efficiency, and address public recreational needs [12,13]. However, systematic research on ring-shaped ecological park belts remains limited globally, underscoring the need for further theoretical and practical exploration of recreational service enhancement during this transition.

Recreation refers to leisure activities pursued for relaxation beyond work or study [14,15,16]. Widely regarded as a component of cultural ecosystem services, recreational services involve the planning, utilization, and management of recreational resources to meet public demand [17]. Current research focuses on supply-side evaluations, employing frameworks such as the Recreation Potential Supply (RPS) and Recreation Opportunity Supply (ROS) models to quantify service capacity [18,19,20,21]. However, there are still few studies on the recreation service level of urban green space systems, and there is a lack of a systematic evaluation system for green space systems, such as urban green belts. Some researchers introduce the spatial spillover effect (hereinafter referred to as the “spillover effect”) to reveal the remote coupling effect of the short-range spillover mechanism of urban green space recreation service on the surrounding area and believe that the evaluation and optimization of spillover effect is the entry point to alleviate the unfair allocation of green space recreation resources [22,23].

The spillover effect (or externality) describes how organizational activities generate unintended societal impacts beyond their primary objectives. The spillover effect of green space services is generally defined as the radiation effect of the economic, social, and environmental impacts of green space services on the surrounding areas [24]. In recent years, based on the externality of green space as a quasi-public good, some scholars have paid attention to its spillover phenomenon, explored the impact mechanism of green space on surrounding areas through spatial adjacency or spatial distance transmission, and scientifically measured the scope of its external spillover [25,26]. Proving that the construction of a Spatial Correlation Network (SCN) is crucial for strengthening regional cooperation and improving urban green space services is essential [27].

Overall, previous studies have laid a necessary foundation for understanding the spillover of green space services and have made preliminary explorations into recreational-service spillovers, but such research is often confined to the macro scale. Although spillover effects are critical for delineating recreational service ranges and capacities, the spillover effects of recreational services provided by green-space systems—especially green belts or ecological park belts—have not yet been systematically assessed or quantitatively analyzed. This study hypothesizes that, during the pivotal transition from green belt to ecological park belt, a positive spillover of recreational-service supply capacity occurs from the core to the periphery, and this hypothesis will be tested through data analysis.

Shanghai’s Outer Ring Ecological Park Belt faces spatial imbalances in recreational service distribution and supply–demand mismatches. As a hybrid urban–rural interface, it exhibits complex spatial elements, heterogeneous land-use conditions, and varying socioeconomic development levels, complicating planning and construction. Current recreational services fail to meet diverse public demands, while disparities in transportation infrastructure and urban functions along the belt highlight the need for optimization.

Evaluating and leveraging spillover effects offers a critical pathway to address these issues. Spatial-spillover research can guide regional synergy and optimize the allocation of recreational services along and beyond the ecological park belt. At the city scale, by testing and quantifying these spillovers, this study seeks to measure the positive radiative capacity of the ecological park belt and to propose strategies for leveraging these effects to enhance urban resilience and foster a sustainable development pattern. Concurrently, by focusing on representative green spaces within the belt and conducting localized spillover analyses, the study reveals how recreational services influence their surroundings, pinpoints supply–demand mismatches, and offers targeted, actionable recommendations to elevate regional recreational service levels.

2. Data

2.1. Study Area

Shanghai, as the leading city of the Yangtze River Delta, is located in China’s eastern coast (120°52′–122°12′ E,30°40′–31°53′ N). From the end of the 20th century, Shanghai began to build a green belt system based on the green belt of the ring road. In recent years, a number of urban parks have been built along the green belt of the outer ring road, and the function of the greenway has begun to transform in the direction of paying equal attention to ecology and recreation services [28]. This study focuses on the core part of the ecological park belt (Figure 1).

In this study, the buffer zone of the recreational service spillover effect in Shanghai Outer Ring Park was first determined as the research scope. The study refers to the urban 15-min cycling circle and comprehensive travel circle planning experience of Singapore [29] and Shenzhen [30], combined with the current situation and based on Shanghai’s street-level administrative boundaries. The study area is defined as the street units within the ecological park belt and about 4 km outside the belt.

2.2. Data Collection

2.2.1. Data Sources

This study evaluates the recreational services of the ecological park belt using data collected from two assessment indicators: recreation potential (RP) and recreation opportunity (RO). Concurrently, urban economic and population data were obtained for subsequent spatial econometric analysis, establishing a multi-source database as shown in

Table 1. Multi-source geospatial database.

Data	Data Type	Data Source	Description
Land use/cover map	Raster data	Resource and Environment Data Cloud Platform http://www.resdc.cn (accessed on 20 July 2025)	Different types of land use in Shanghai (30 m)
Urban green map	Raster data	Open Street Map https://openmaptiles.org (accessed on 20 July 2025)	Shanghai park green space
Park POI	Vector data	Baidu map and Google Maps https://www.baidu.com/, http://www.google.com (accessed on 20 July 2025)	POI (Natural sensing point): Scenic spot, Plant/animal exhibition area, ecological conservation area
			POI (Shopping/dining point): Restaurant, marina complex, service center, shop in the park
			POI (Dynamic entertainment point): Playground, amusement park, cycling station, camping base
			POI (Cultural/Educational point): Educational base, museum, art gallery, historical building in the park
			POI (Innovative development level): University, innovation/technology industrial district
center of administrative regions	Vector data	Administrative map	The development center of Shanghai administrative regions
road network	Vector data	Open Street Map https://openmaptiles.org (accessed on 20 July 2025)	Different types of roads in Shanghai
subway station	Vector data	Baidu map https://www.baidu.com (accessed on 20 July 2025)	Subway stations in Shanghai
GDP	Vector data	‘government work report’ of each administrative regions	Per capita GDP of Shanghai administrative regions
Population	Vector data	‘government work report’ of each administrative regions	Population of Shanghai administrative regions

. This study collected data for three specific years—2013, 2018, and 2023—to demonstrate the evolution of recreational service patterns and the growth in service levels over the 15-year period.

All spatial data were collected in either vector or raster format and uniformly projected to the same coordinate system (WGS_1984_UTM_ZONE_51N) using ArcMap 10 software for analysis. To unify the precision of data across different years, facilitate analysis, and support the development of targeted and implementable strategies, we chose to process and output the data as raster files with a standardized pixel size of 50 × 50 m. We normalized the spatial data of each indicator. These normalized raster layers were then weighted and overlaid to generate the recreational service evaluation results for the ecological park belt in 2013, 2018, and 2023.

2.2.2. Recreation Service Evaluation

This study constructed an evaluation system for recreational services in the ecological park belt. The research introduced Clark et al.’s assessment method for cultural ecosystem service supply, establishing an evaluation framework through two models: Recreation Potential Supply (RPS) and Recreation Opportunity Supply (ROS) [31,32].

The RPS assessment is based on land use and land cover types to scientifically reflect the spatiotemporal changes in recreational service supply [33]. This study incorporated various ecosystems into the analysis, including urban green spaces, woodlands, water bodies, grasslands, unused land, agricultural land, and urban areas, identifying them as potential providers of recreational services. We employed the evaluation strategy proposed by Crossman et al. [34] to assess and assign values to different land types based on their recreational potential, with adjustments made according to the actual conditions of the ecological park belt.

The ROS was introduced into the evaluation system to assess the opportunities for citizens to enjoy recreational services provided by the ecological park belt by analyzing the types and quantities of recreational service points, green space accessibility, and the surrounding urban environment. Recreational behaviors, as observable responses to public recreational demands and environmental support, can reflect public feedback on recreational services [14]. Through field surveys, the study categorized current recreational activities in the ecological park belt into four types: dynamic entertainment, Natural sensing, culture/education, and shopping/dining, and accordingly classified and spatially located recreational service points, with a focus on open spaces or facilities provided by parks and green spaces for public use. POI data were used to obtain the aforementioned facility information, including names and coordinates, through the POI interfaces of Baidu Maps and Google Maps. Research has shown that distance from the center of each administrative region, distance from water bodies, distance from roads, distance from subway stations, and green space accessibility affect people’s opportunities and efficiency in reaching recreational service points [35,36]. Based on actual hydrological and transportation conditions and considering Shanghai’s polycentric urban development pattern, these indicators were incorporated into the system.

The evaluation of green space accessibility employed a cost–distance analysis method. This method reflects the accessibility of “sources” by calculating the minimum cumulative cost distance from each grid cell to the nearest source [37]. The calculation formula is Formula (1):

$$A=\left\{\begin{array}{c}1/2{\displaystyle \sum _{i=1}^n}\left(C_i+C_{i+1}\right)\\ \sqrt{2}/2{\displaystyle \sum _{i=1}^n}\left(C_i+C_{i+1}\right)\end{array}\right.$$

(1)

where $A$ represents the distance cost from any location to the destination, $C_i$ represents the cost value of the $i$-th grid cell, $C_{i+1}$ represents the cost value of the $i+1$-th grid cell along the movement direction, and n is the total number of grid cells [38].

By establishing a 50 × 50 m grid for the study area and using pedestrian walking time as the cost, with reference to previous studies [39,40] and the actual situation in Shanghai, different cost values were assigned to the grid according to different land use types, as shown in

Table 2. Cost values of different land use types.

Land Use Type	Urban Green/Forests	Water	Grassland	Unused Land	Agricultural Land	Road	Urban
Cost values	5	10	3	2	4	1	2

. Using the cost-weighted distance tool in the spatial analysis module of ArcGIS, the green space accessibility results were obtained.

The datasets of other RO indicator layers were also compiled in raster format, with kernel density analysis and Euclidean distance analysis performed using GIS software, all standardized to a spatial resolution of 50 m.

To more accurately evaluate the spatial differentiation of recreational service levels in the ecological park belt and analyze its internal patterns, we employed the entropy weight method to determine indicator weights based on the degree of variation in each indicator’s values. The entropy weight calculation yielded weights of 49.80% for RP and 50.20% for RO. To simplify the analysis and facilitate subsequent calculations, the study rounded these weights to 50% each. After rounding the weights, the ranking change rate of the evaluation results was 3.63% (less than 5%), indicating that the results were not significantly altered. Additionally, the study subjected the rounded weights to a 5% perturbation, and the change rate remained below 10%, further confirming the robustness of the weights. The complete evaluation framework with all indicator weights is presented in

Table 3. The evaluation index system is composed of recreation service (recreation potential and recreation opportunity).

Component	Variable	Description
recreation potential (0.5)	Land use/cover	Urban green (5)
		Forests (4) Water (3)
		Grassland (4)
		Unused land (3)
		Agricultural land (2) Road (2)
		Urban (1)
recreation opportunity	POI (Natural sensing point) (0.118)	Kernel density analysis
(0.5)	POI (Commercial service point) (0.131)	Kernel density analysis
	POI (Dynamic entertainment point) (0.12)	Kernel density analysis
	POI (Cultural and Educational point) (0.136)	Kernel density analysis
	Distance from the center of each administrative region (0.100)	Euclidean distance
	Distance from water (0.099)	Euclidean distance
	Distance from road (0.099)	Euclidean distance
	Distance from subway station (0.099)	Euclidean distance
	Accessibility of green space (0.098)	Cost distance

.

3. Research Method

3.1. Identification of Recreation Service Core Areas

Following the systematic evaluation, the study obtains recreational-service supply levels for the park-belt area in 2013, 2018, and 2023. It then proceeds to identify the core zones of recreational services within the ecological park belt and its buffer zone—source areas that supply recreational services—and spatially structures these cores to establish the basis for subsequent verification of the core-to-periphery driving effect in the spatial-spillover analysis. The study selects MSPA, a mature spatial-pattern analysis approach developed by scholars such as Vogt based on mathematical morphology principles, because it measures, identifies, and segments the spatial patterns of raster images and can more accurately distinguish landscape types and spatial structures [41,42].

The study first superimposed the raster data of recreational service evaluation results and classified the 10-year integrated recreational service level into low, medium, and high zones using the natural-breaks method. Using Guidos Toolbox 3.2, MSPA analysis was conducted to identify the spatial topological relationship between the target pixel set and the structural elements and to calculate various spatial types, including core, background, branch, edge, perforation, islet, bridge, and loop. The study thereby identified the high-level source areas of recreational services within the study area and delineated the boundaries of the core zones, providing the foundation for subsequent verification and analysis.

To analyze the spillover effects of improved recreational service levels in core areas on surrounding regions while ensuring data operability and scientific rigor, this study employs GIS software to classify grid cells into 100-m interval zones from the core areas. The spillover buffer was precisely delineated within a 2-km radius of core zones, generating 20 concentric data belts (including the core areas themselves) as primary samples for spillover analysis. This approach optimally balances the study’s geographical characteristics with methodological precision.

3.2. Spillover Effects Analysis

After obtaining the recreational-service-level data for each belt, this study will compute and analyze the data. To examine whether the core areas of recreational services in the ecological park belt system have driven the growth of recreational-service levels in other areas through spillover effects, a spatial-econometric model will be established.

This study examines whether the core areas of recreational services in the ecological park belt system have driven the growth of recreational service levels in other areas through spillover effects. The method is based on research in the field of econometrics on the driving effects of inter-regional wealth [43,44] and refers to the analytical framework constructed by scholars such as Huang, who introduced this idea into the study of the driving effects of inter-regional ecotourism comfort [45]. The spatial econometric model is used to explore the impact of an index in a region at the beginning of a period on the growth rate of the index in surrounding regions over a certain time span, so as to verify its driving and spillover effects. Based on the current situation of ecological park belt construction and against the background of the overall growth trend of recreational services from 2013 to 2023, this study takes the growth rate during 2013–2018 and 2018–2023 as the explained variable and the recreational service level at the beginning of the period as the explanatory variable to construct subsequent spatial autocorrelation and spatial econometric models. With reference to related studies [45], population size (P), economic level (GDP), and innovation capability (In) are added as control variables. The control variables include the population of Shanghai’s administrative regions, per capita GDP of Shanghai administrative regions, and the kernel density of universities and innovation/technology industrial zones. The explanatory variables and control variables have passed the Pearson correlation test with a coefficient |r| < 0.7.

3.2.1. Spatial Autocorrelation Analysis

To test the spillover of recreational services in the ecological park belt, it is necessary to pass the global Moran’s I test to prove that the recreational services in the study area have spatial autocorrelation. Spatial autocorrelation analysis can obtain the overall characteristics of the correlation degree of the recreational service growth rate in adjacent spaces. The formula is as follows [46,47]:

$$Mora{n}^{\prime }s I= \left[\sum _{i=1}^n\sum _{j=1}^n{W}_{ij}\left(X_i-\overline{X}\right)\left(X_j-\overline{X}\right)\right]/\left[S^2\sum _{i=1}^n\sum _{j=1}^n{W}_{ij}\right]$$

$$S^2=1/n\sum _{i=1}^n{\left(X_i-\overline{X}\right)}^2$$

(2)

where $n$ is the total number of data zones; $X$ is the growth rate of recreational services; $W_{ij}$ is the spatial weight, where $W_{ij}$ = 1 if data zones are adjacent and 0 otherwise; and $I$ takes values between [−1, 1], when $I$ > 0 it indicates positive spatial correlation of recreational service growth rates, with values closer to 1 showing stronger positive correlation.

3.2.2. Spatial Econometric Models

To verify the influence of explanatory variables on the dependent variable and lay the groundwork for subsequent calculations, this study employs the Ordinary Least Squares (OLS) regression model from spatial econometrics to conduct a preliminary assessment of relationships between variables.

OLS uses the principle of minimum sum of residual squares to determine the position of the line to solve the fitting problem of the relationship between data variables [48]. The formula is as follows:

$$Y=a_0{+a}_{1}X+a_2{X}_p{+a}_3{X}_{GDP}{+a}_4{X}_{In}+\epsilon$$

(3)

where $Y$ represents the growth rate of recreational service levels in a given region during a specific period; $X$ represents the initial recreational service level of the region at the beginning of the period; $X_p$, $X_{GDP}$, $X_{In}$ serves as the control variable; $a_1$ to $a_4$ are the coefficient of the impact factor to be estimated; $a_0$ is a constant; and $\epsilon$ is the error random term.

The spatial lag model (SLM) and spatial error model (SEM) from spatial econometrics [49] were employed to examine the influencing factor model of recreational service level growth rates. The SLM explores whether the explanatory variable Y is influenced by the growth in the level of recreation services in its surrounding area (spillover effects), with the following equation [50]:

$$Y=\rho WY{+a}_{1}X+a_2{X}_p{+a}_3{X}_{GDP}{+a}_4{X}_{In}+\epsilon$$

(4)

where $\rho$ is the spatially lagged auto regressive coefficient, measuring the spillover effect of the growth rate of recreation service level in the geographical neighborhood; $W$ is the spatial weight matrix. The spatial weight matrix $W_{ij}$ follows the same specification method as in Formula (2); ε is the random disturbance term.

The SEM suggests that when growth rates at a specific data tape are influenced by a range of local factors, certain key variables that exhibit spatial correlation across geographic areas may be overlooked (referred to as error terms). SEM captures the impact of explanatory variables of growth rates on spatially interdependent random error shocks in neighboring regions through the following equation [51]:

$$\begin{array}{c} \\ Y=a_{1}X+a_2{X}_p{+a}_3{X}_{GDP}{+a}_4{X}_{In}+\epsilon \\ \epsilon =\lambda W\epsilon +\mu \end{array}$$

(5)

where ε is the spatial autocorrelation error term; λ is the autoregressive coefficient of the spatial error term, quantifying the impact of the error term from sample observations on the explained variable; and μ is the random error vector of the normal distribution.

4. Results

4.1. Evolution of Recreational Service Supply and the Scope of the Core Area

4.1.1. Evolution of Recreational Service Supply and Its Growth Rate

From 2013 to 2023, the recreational services of the ecological park belt showed progressive improvement, with high-value areas becoming more clustered and the overall structure tending toward continuity and completeness. The results are presented in Figure 2 and Figure 3, as follows:

(1)

During the implementation of green belt ecological projects, recreational services in the outer-ring area showed higher levels in the west than east, with significantly better performance along the belt than surrounding areas, exhibiting scattered growth. In 2013, the average service level was 0.234, with few and dispersed parks along the belt. High-value clusters concentrated in the southwest and northeast, while the western section, benefiting from proximity to sub-centers and metro stations, demonstrated more continuous and superior services. By 2018, the green belt was nearly connected, with ecological projects largely completed. The average service level rose to 0.243, representing a relative increase of 3.85% compared to the previous period This improvement was supported by the addition of 15 new parks and expansions, such as Gucun Park, which helped form a preliminary ecological park belt structure. The western area, intersecting with the Huangpu River waterfront green belt, emerged as a strong growth cluster, while the southeastern section showed limited improvement;

(2)

2018–2023: From 2018 to 2023, the enhancement of green belt services and initiation of the ecological park belt construction led to widespread improvements in recreational services along the belt, with continuous high-value areas. By 2023, the average recreational service level reached 0.264, marking a relative increase of 8.64% from 2018. High-growth areas clustered in the southwestern section of the ecological park belt, where the elongated Meilong Ecological Park and Xuhui Riverfrond Green Space formed growth corridors through continuous development, achieving local growth rates exceeding 10%, while newly built parks like Mianqing Park in the southeast jointly boosted recreational services in the southern belt. Meanwhile, newly constructed parks and the Outer Ring Canal Green belt in the northeast connected previously isolated high-value areas, forming additional growth corridors.

4.1.2. Recreational Service Core Area

Through spatial pattern analysis, we identified core supply areas of recreational services. The results reveal that the core areas are primarily located within wider green spaces along the belt, most of which underwent renovation by 2023 with added recreational facilities. Additionally, core zones are distributed at junctions where the green belt connects with other urban green systems (e.g., waterfront spaces), forming perpendicular intersections that extend into both the inner urban and outer suburban areas. A stable recreational service network has begun to take shape. The corridor, composed of the Outer Ring Expressway and adjacent linear green spaces, is updated or constructed concurrently with belt parks to form a slow travel system connecting various parks and recreational areas.

Integrated analysis of service levels and growth rates shows high spatial coincidence between core areas and 2023 high-value recreational zones. Notably, the strongest growth areas across both periods were predominantly located within core zones. This demonstrates Shanghai’s successful utilization of high-potential green spaces along the belt over the decade, with supplemental support from other green systems, resulting in increasingly coherent recreational network patterns. The strategic functional transformation of the green belt has achieved remarkable outcomes.

4.2. Spillover Effects of Recreational Services

4.2.1. Spatial Correlation Analysis Results

This study uses Stata18 to calculate Moran’s I index and subsequently verifies that the growth rate of recreational services exhibits spatial autocorrelation. Under the 2 km threshold condition and the 01 spatial matrix, Moran’s I value is 0.524 in the first five-year period and 0.416 in the second period, both with a significance of 0.000 (<0.05) (

Table 4. Spatial correlation test results.

Moran’s I			OLS			LM
Variable	I	p-Value *	Variables	Coefficient	p-Value *	Test	Statistic	p-Value *
Y2013–2018	0.524 ***	0.000	X	1.497 ***	0.000	LM-ERR	1.700	0.192
			GDP	−0.478 ***	0.008	R-LM-ERR	5.632 **	0.018
Y2018–2023	0.416 ***	0.000	In	−0.493	0.313	LM-lag	22.612 ***	0.000
			p	−0.012	0.968	R-LM-lag	26.544 ***	0.000

Note: ***, **, *, respectively, represent the significance levels of 1%, 5%, and 10%. Table 5 is the same.

). This shows that in two periods, the recreational service growth rates will be subject to a significant global spatial agglomeration effect. The Moran’s I indices are >0, indicating a positive correlation.

The study subsequently conducted a linear OLS regression on the sample data within the 2 km threshold, demonstrating a linear relationship between the initial level of recreational services and the growth rate. The independent variable has a positive impact on the dependent variable, with a significance level of 0.000 (<0.01) (Table 4). This indicates that the higher the initial level of recreational services, the higher the growth rate. The analysis confirms that the sample data exhibit significant spatial autocorrelation and that there is a significant linear relationship between the variables, which allows for further analysis of spillover effects.

4.2.2. Spillover Effect Analysis Results

To examine the spillover effects of recreational services in the ecological park belt, we first performed LM tests to select the appropriate model. The results showed LM-lag and R-LM-lag: p < 0.01; R-LM-ERR: p < 0.05; LM-ERR: p > 0.1. These findings indicate the coexistence of both spatial error effects and spatial lag effects, with the spatial lag model demonstrating relatively stronger explanatory power. To verify the robustness of the model estimates, the study conducted the Hausman test to confirm the appropriateness of the fixed effects specification, with statistically significant results. A likelihood ratio test (LR) was performed, yielding p > 0.05, indicating that the SDM model was not statistically superior to the SLM model. Based on these findings, the analysis primarily employed the SLM model for spillover effect estimation, while SEM results served as supplementary evidence (

Table 5. Spillover effect test results of SLM and SEM.

	SLM		SEM
Variable	Coefficient	p-Value	Coefficient	p-Value
X	6.078 ***	0.000	1.288 ***	0.000
GDP	3.427 ***	0.000	−0.126	0.652
In	0.409 ***	0.008	−0.885	0.234
p	1.663	0.637	−0.052	0.907
ρ/λ	0.347 ***	0.000	0.582 ***	0.000
Number of obs	40		40
R2	0.684		0.856
Log-likelihood	172.837		98.680
Hausman	132.130 ***		132.130 ***

).

(1)

The core areas of recreational services in the ecological park belt can drive coordinated development in other regions through spillovers of service level improvements. The SLM analysis showed ρ > 0 (ρ = 0.347, p < 0.001), indicating excellent model fit and strong explanatory power. Significant positive correlations in growth rates were observed among adjacent data belts within 2 km of core areas (including the cores themselves), confirming that high-value areas can achieve spillovers through recreational service growth. The SEM results showed λ > 0 (λ = 0.582, p < 0.001) and provided additional evidence of significant spillover effects;

(2)

The explanatory variable coefficient was 6.078 (p = 0.000), meaning each unit increase in initial recreational service level corresponded to an average 6.078-unit increase in current growth rate. This suggests that areas with better recreational service foundations demonstrate stronger growth performance during construction and renewal phases against the backdrop of overall regional recreational level improvement.

4.2.3. Spillover Distance Boundary

To further analyze the spillover boundaries of the growth in recreational services in the core area, a method of setting distance thresholds was employed. The Spatial Lag Model (SLM) was utilized to estimate Equation (4), with the spillover coefficient (ρ) serving as the basis for examining the spatial dependence and spillover range among adjacent data bands. When setting the thresholds, a regression was conducted every 200 m from the core area, and the results were sequentially recorded. The findings are illustrated in Figure 4, as follows:

As the distance from the core area increases, the spillover effect shows a weakening trend, forming three distinct distance intervals: 400–1000 m Interval: The spillover coefficient peaks at 400 m and then drops sharply.1000–2000 m Interval: The spillover coefficient significantly rebounds to around 0.4 at 1000 m. From 1000 m to 2000 m, it decreases steadily, with minor increases at 1400 m and 1800 m.

2000–3000 m Interval: At a threshold of 2200 m, an abnormal value occurs. Further analysis reveals that in some areas outside the ecological park belt segment between 2000 m and 2400 m, there are large tracts of farmland. Over the past decade, with the progress of urbanization, the area of farmland has decreased significantly while the area of construction land has increased markedly. This has led to a decline in the recreational service potential of these areas and even a small amount of negative growth in recreational service supply. From 2400 m to 3000 m, the spillover coefficient continues to decline steadily, maintaining a consistent trend with the previous two intervals. Therefore, in the preceding sections of this study, a boundary of 2000 m from the core area was set to verify spatial correlation and spillover effects.

4.3. Spillover Effects and Boundaries of Specific Parks

The study further conducted a spillover analysis of recreational services for representative parks along the belt. We selected a total of 20 local data bands, including the core block where the park is located and its surrounding areas, i.e., sample data within 2 km of the core area of the park, and obtained the local spillover coefficients for these areas using the same method as described above. The parks were categorized into four types: Type A parks were constructed early and underwent multiple expansions over the decade; Type B parks are newly built urban parks along the belt; Type C parks are connected to waterfront green belts or green wedges, forming park clusters together. The results are shown in Figure 5.

(a)

These parks have a relatively large initial area and have continued to expand, maintaining a high level of recreational services. Shanghai Waterside Forest Park has a high spillover coefficient, which increases gradually with the threshold value. Its second phase of construction, based on citizens’ needs for nature therapy and weekend excursions, complements the first phase in terms of recreational service types. Features such as RV campsites and nature education bases effectively highlight the park’s unique characteristics. However, Gucun Park exhibited a negative spatial spillover, where the increase in the level of recreational services within the park did not positively drive the surrounding areas and even suppressed the growth rate of nearby regions. Over the decade, the local population has increased, but the development of surrounding infrastructure, such as transportation, has been relatively slow. The lack of connecting parks to the east of Gucun Park resulted in a discontinuity along the belt, which is not conducive to the positive spillover effect;

(b)

These parks represent the largest proportion along the belt, with rich recreational facilities built. Over the decade, with the construction of high-tech parks and residential areas, the population density around these parks has increased. Recreational demand has driven a rapid increase in both the quantity and quality of supply. The construction of Mianqing Park filled the gap in the southern section of the ecological park belt, serving as the starting point for updating this section. Its distance attenuation curve is similar to the overall trend of the belt. Gaole Park showed the highest spillover effect within the 1500–2000 m threshold. The park has an elongated shape, running parallel to the Pudong Canal and the Outer Ring Canal on its east and west sides, meeting the waterfront recreational needs along the canals;

(c)

These park clusters are formed by the integration of multiple urban green space systems. The grouped layout of multiple parks shows a significant spillover effect, with the maximum positive driving effect occurring at a threshold of 1000 m. At the junction of Xuhui Riverfront and the belt, the land use functions around the park cluster are relatively unified, with active art and leisure sports activities, creating a unique district atmosphere. There are several functionally unique urban parks, such as a nature art park and a sports park. Their distinct characteristics have significantly boosted the surrounding recreational service supply. In contrast, the Biyun Green Wedge park group showed a substantial weakening of the spillover effect and even a negative spatial spillover when the threshold exceeded 1500 m. The area around the park cluster is mainly composed of industrial parks and residential areas. It has a large population density and a high demand for recreational services. However, land use is limited, the green space ratio is low, and the spillover of recreational services is restricted. Currently, there is an imbalance between the supply and demand of recreational services.

5. Discussion

5.1. The Effectiveness of the Transformation of the Green Belt Function

After calculation and analysis, the study demonstrates that over the past decade, the overall level of recreational services in the green belts and their surrounding areas has improved significantly, with a relatively fast growth rate and more balanced development. Meanwhile, the study confirms that within a 2-km radius of the core area, Shanghai’s ecological park belt exhibits spillover of recreational services. This indicates that the functional transformation of green belts has achieved considerable results over the past decade. The ecological park belt has effectively played a positive driving role, leading to an enhancement in the service level of urban green spaces. The foundation of green belts and fully utilizing green space resources to create an ecological park belt that is systematically integrated in space and functionally interlinked is a feasible direction for the renewal of green belts.

5.2. Factors Influencing the Spillover Effect of Recreational Services

After calculation and analysis, the study confirms that Shanghai’s ecological park belt exhibits spillover of recreational services. The analysis shows that there is a significant overlap between the areas with high recreational service growth rates (i.e., the core areas and their surrounding regions) and the areas with high spillover coefficients and high recreational opportunity values across the two five-year periods. Since the kernel density of various types of recreational services is a primary indicator for evaluating recreational opportunities, we argue that the increase in the quantity of recreational service supply and the diversity of service categories are the main driving forces behind the rise in recreational service value and its subsequent spillover to surrounding areas.

Integrating recreational demands to promote the linkage between recreational service points can effectively amplify the spillover effect of recreational services. In the early stages of the green belt’s functional transformation, park construction focused on basic infrastructure, only meeting people’s demands for nature experience and entertainment/fitness-related recreational activities. However, the types of services within the parks were limited, and the recreational systems between parks were relatively independent. Starting from 2018, parks began to pay more attention to cultural education and dining/shopping needs during their construction and expansion. As a result, the number of recreational venues such as nature education bases, art galleries, cultural experience centers, and specialty dining spots along the belt has increased. Parks like Gucun Park, which were constructed early, need to explore recreational motives and link with newly built parks to find new attractions for recreational activities.

5.3. Potential Limitations in the Research

There are a number of weaknesses that limit the quality of our approach. First of all, the precision of the data still needs to be improved. In order to unify the precision across different years, there is a certain degree of error in the evaluation of the RP index in this study. Due to the limitations of data sources, this study selects data with a time span of five years. The potential randomness, to some extent, affects the accuracy of the research results. Therefore, future studies should optimize data collection methods and appropriately reduce the time span to control errors.

In future research, attempts should be made to incorporate the consideration of non-physical factors (such as public perception, activity preferences/satisfaction) into the evaluation of recreational services. The demand side can be quantified through methods such as the collection and analysis of social media data and questionnaire surveys. By integrating both the supply and demand sides of services, the evaluation system can be made more comprehensive, leading to more holistic and in-depth research conclusions that better address the recreational needs of citizens.

In addition, the study still needs to delve deeper into the mechanisms of the spillover effect. First, when conducting a partial analysis of the ecological park belt, we chose to classify the parks based on their construction phases and distribution patterns along the belt and proposed suggestions on how parks along the belt can effectively exert spillover effects. Future research can optimize the classification methods, refine the functional types of different sections of the ecological park belt, explore the impact of recreational service types on the spillover effect and its buffer boundaries, as well as analyze the underlying mechanisms. Second, future research can further investigate the factors and internal mechanisms affecting spillover effects by optimizing models. By setting multiple explanatory variables and using the Spatial Durbin Model, researchers can obtain indirect effects, direct effects, and total effects, thereby revealing spillover mechanisms and heterogeneous impacts and providing targeted planning recommendations for the improvement in green space system services.

6. Conclusions

The study examines the spatial imbalance in the layout of recreational service structures and the imbalance between supply and demand of functions during the functional transformation of Shanghai’s circular green belt into an ecological park belt. Against the backdrop of the “Park City” concept, the study responds to the integration of urban and rural development and the recreational needs of citizens. It also aims to improve the evaluation system for recreational services and assess the ecological park belt and its surrounding areas in 2013, 2018, and 2023. By introducing spatial econometric methods that account for spillover effects, we establish an analytical model to verify the existence of spillovers in the recreational services of the belt. We further conduct boundary analyses of the spillover effects on the entire belt, its local sections, and different types of parks along the belt. The main conclusions are as follows.

(1)

The ecological park belt’s recreational services have increased, with core areas becoming more clustered and structurally continuous, yet remaining somewhat fragmented. Over the decade, Shanghai has effectively enhanced overall service levels by introducing diverse recreational spaces and facilities while preserving urban forest resources and maintaining regional recreational potential. However, the concentration of high-value growth zones within the belt suggests planners should further explore the recreational value of green spaces beyond the belt;

(2)

The initial level of recreational services has a positive impact on the growth rate of the period. The spillover of the growth in recreational service levels in high-value areas drives the development of adjacent low-value areas, with this promoting effect radiating outward from the core zones. This demonstrates that the ecological park belt, as a relatively complete green space system, effectively enhances the service level in the urban–rural integration areas and increases its potential to expand its service scope both toward the urban center and the peripheral suburbs;

(3)

The spatial scope of the spillover effect is examined. As the distance threshold increases, the spillover effect weakens, resulting in three distance intervals. The spillover coefficient is highest at 400 m, then drops sharply, and significantly rebounds at 1000 m; beyond 2000 m, the spillover effect becomes unstable. In planning and construction, efforts should continue to improve the quality of recreational services along the belt and fully utilize the spillover effects of the three boundaries to enhance the overall green space service level in the region. Specifically, the 400-m boundary with the maximum spillover effect should be leveraged by extending the recreational activities of the main parks to the surrounding squares and pocket parks. This will meet the daily recreational needs of citizens, especially the elderly, through short walks and promote the clustered development of recreational services in the area, together with the main parks. The 1000-m key spillover boundary should be utilized by improving the quality of affiliated and vacant green spaces. Based on citizens’ needs, various types of recreational facilities should be added to create more high-value areas. The ecological park belt system should also be linked with waterfronts and green wedges to form a coordinated network. The 2000-m stable spillover boundary, which serves as the buffer zone for spillover effects, should focus on tapping into green space resources outside the belt. This will create a unique recreational atmosphere in the area and comprehensively meet various needs. High-value areas should be connected through cycling greenways and waterfront scenic corridors. Facilities such as cycling stations should be added to integrate small and micro green spaces with urban transportation systems, meeting citizens’ slow travel needs. This can also mitigate the fragmentation and separation of green space resources by built-up areas, enrich the composition of the ecological park belt core, and enhance its structural stability during development and expansion;

(4)

There are differences in the spillover effects of parks and park clusters along the belt. During the multi-phase expansion of old parks, the negative spatial spillover should be guarded against. While optimizing the park itself, efforts should be made to improve the surrounding supporting infrastructure, establish activity linkages with small green spaces to disperse visitors, and deeply explore the unique features of the park to differentiate it from other parks of similar size in terms of functional settings and event planning to avoid competition. New parks along the belt show good initial spillover effects and can gradually fill the gaps in the belt. In design and planning, it is essential to fully understand citizens’ needs and explore regional culture and local landscape characteristics. After construction, updates should focus on the park’s connection with surrounding water bodies and transportation. Park clusters exhibit better spillover effects, with the optimal spillover effect achieved at 1500 m. A group of parks with clear functions and different main types can significantly enhance the surrounding recreational service supply. In planning, attention should be paid to the distribution of park clusters along the belt and the balance of green space service volumes on both sides of the belt. If one side is affected by urban built-up areas, accessible affiliated green spaces should be explored to meet the needs of citizens on that side and balance the driving effects on both sides.

In conclusion, the study analyzes the spillover patterns of recreational services from the green belt of a megacity to surrounding areas after its functional transformation. It concludes that over the past decade, the ecological park belt in Shanghai has effectively driven the overall improvement in the recreational service level in the outer ring area. Rational utilization of the spillover effect to maximize its potential is a key focus. The study provides theoretical support and practical suggestions for the further planning and construction of ecological park belts.